Vous êtes sur la page 1sur 37

Fouling in membrane bioreactors used in wastewater treatment

Pierre Le-Clech , Vicki Chen, Tony A.G. Fane


UNESCO Centre for Membrane Science and Technology, School of Chemical Engineering, The University of New South Wales, Sydney 2052, NSW, Australia

Abstract The membrane bioreactor (MBR) can no longer be considered as a novel process. This reliable and efcient technology has become a legitimate alternative to conventional activated sludge processes and an option of choice for many domestic and industrial applications. However, membrane fouling and its consequences in terms of plant maintenance and operating costs limit the widespread application of MBRs. To provide a better understanding of the complex fouling mechanisms and propensities occurring in MBR processes, this review compiles and analyses more than 300 publications. This paper also proposes updated denitions of key parameters such as critical and sustainable ux, along with standard methods to determine and measure the different fractions of the biomass. Although there is no clear consensus on the exact phenomena occurring on the membrane interface during activated sludge ltration, many publications indicate that the extracellular polymeric substances (EPS) play a major role during fouling formation. More precisely, the carbohydrate fraction from the soluble microbial product (also called soluble EPS or biomass supernatant) has been often cited as the main factor affecting MBR fouling, although the role of the protein compounds in the fouling formation is still to be claried. Strategies to limit fouling include manipulating bioreactor conditions, adjusting hydrodynamics and ux and optimizing module design.
Keywords: Membrane bioreactors; Fouling; Activated sludge; Operating conditions; Cleaning

Contents
1. 2. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1. MBR history . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Fouling mechanisms for complex uids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. Concepts of critical and sustainable ux in mixed species environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2. Effect of operating modes on performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3. Cake structure and the effect of mixed species on cake morphology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4. Effect of membrane morphology and surface chemistry on fouling mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.5. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Roadmap for MBR fouling parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1. Membrane characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.1. Physical parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.2. Chemical parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2. Feedbiomass characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.1. Nature of feed and concentration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.2. Biomass fractionation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.3. Biomass (bulk) parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.4. Floc characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 18 19 19 20 21 23 24 24 25 25 27 28 28 28 29 31

3.

Corresponding author. Tel.: +61 2 93855762; fax: +61 2 93855966. E-mail address: p.le-clech@unsw.edu.au (P. Le-Clech).

4.

5.

3.2.5. Extracellular polymeric substances (EPS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.6. Soluble microbial products (SMP) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3. Operating conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1. Aeration, crossow velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.2. Solid retention time (SRT) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.3. Unsteady state operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4. Fouling mechanisms in MBRs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.1. Constant TMP operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.2. Constant ux operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mitigation of MBR fouling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1. Removal of fouling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.1. Physical cleaning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.2. Chemical cleaning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2. Limitation of fouling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1. Optimization of membrane characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.2. Optimization of operating conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.3. Modication of biomass characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

32 34 36 36 36 37 38 38 38 41 41 41 41 42 42 42 43 44 45 45

1. Introduction Membrane bioreactor (MBR) technology combines the biological degradation process by activated sludge with a direct solidliquid separation by membrane ltration. By using micro or ultraltration membrane technology (with pore sizes ranging from 0.05 to 0.4 m), MBR systems allow the complete physical retention of bacterial ocs and virtually all suspended solids within the bioreactor. As a result, the MBR has many advantages over conventional wastewater treatment processes. These include small footprint and reactor requirements, high efuent quality, good disinfection capability, higher volumetric loading and less sludge production [1]. As a result, the MBR process has now become an attractive option for the treatment and reuse of industrial and municipal wastewaters, as evidenced by their constantly rising numbers and capacity. The current MBR market has been estimated to value around US$ 216 million and to rise to US$ 363 million by 2010 [2]. However, the MBR ltration performance inevitably decreases with ltration time. This is due to the deposition of soluble and particulate materials onto and into the membrane, attributed to the interactions between activated sludge components and the membrane. This major drawback and process limitation has been under investigation since the early MBRs, and remains one of the most challenging issues facing further MBR development [3]. 1.1. MBR history The MBR process was introduced by the late 1960s, as soon as commercial scale ultraltration (UF) and microltration (MF) membranes were available. The original process was introduced by Dorr-Olivier Inc. and combined the use of an activated sludge bioreactor with a crossow membrane ltration loop [4]. The at sheet membranes used in this process were polymeric and featured pore size ranging from 0.003 to 0.01 m [5]. Although the idea of replacing the settling tank

of the conventional activated sludge process was attractive, it was difcult to justify the use of such a process because of the high cost of membranes, low economic value of the product (tertiary efuent) and the potential rapid loss of performance due to fouling. As a result, the focus was on the attainment of high uxes, and it was therefore necessary to pump the mixed liquor suspended solids (MLSS) at high crossow velocity at signicant energy penalty (of the order 10 kWh/m3 product) to reduce fouling. Due to the poor economics of the rst generation MBRs, they only found applications in niche areas with special needs like isolated trailer parks or ski resorts for example. The breakthrough for the MBR came in 1989 with the idea of Yamamoto et al. to submerge the membranes in the bioreactor [6]. Until then, MBRs were designed with the separation device located external to the reactor and relied on high transmembrane pressure (TMP) to maintain ltration. The other key steps in the recent MBR development were the acceptance of modest uxes (25% or less of those in the rst generation), and the idea to use two-phase bubbly ow to control fouling. The lower operating cost obtained with the submerged conguration along with the steady decrease in the membrane cost encouraged an exponential increase in MBR plant installations from the mid 1990s. Since then, further improvements in the MBR design and operation have been introduced and incorporated into larger plants. While early MBRs were operated at solid retention times (SRT) as high as 100 days with mixed liquor suspended solids up to 30 g/l, the recent trend is to apply a lower SRT (around 1020 days), resulting in more manageable mixed liquor suspended solids (MLSS) levels (1015 g/l). Thanks to these new operating conditions, the fouling propensity in the MBR has tended to decrease and overall maintenance has been simplied as less frequent membrane cleaning is necessary. There is now a range of MBR systems commercially available, most of which use submerged membranes although some external modules are available; these external systems also use two-phase ow for fouling control. In terms of membrane congurations, mainly

hollow ber and at sheet membranes are applied for MBR applications [7]. The economic viability of the current generation of MBRs depends on the achievable permeate ux, mainly controlled by effective fouling control with modest energy input (typically 1 kWh/m3 product). More efcient fouling mitigation methods can be implemented only when the phenomena occurring at the membrane surface are fully understood. The plethora of publications dealing with MBR fouling and published within the last 5 years tends to dilute the accessibility of information and may lead to some confusion. This review presents a state-of-the-art assessment of MBR fouling based on the most recent and relevant papers on the subject. After discussion of fouling mechanisms for complex uids, a comprehensive roadmap for MBR foulants and fouling parameters will be proposed. Finally, a review of the current methods for fouling mitigation in MBR systems will detail design options to optimize MBR operation. This review aims to open doors to new ideas and directions for optimized and more sustainable MBR processes. 2. Fouling mechanisms for complex uids Signicant advances in understanding fouling of individual components such as bacteria, yeast, proteins, and colloids have occurred in microltration and ultraltration literature [811]. Much of this literature has focused on the effect of charge (via pH variation or salt concentration), crossow, concentration, membrane hydrophilicity, membrane pore size and ux (constant pressure or constant ux). While some broad trends for simple colloids are valid for macromolecules (the most commonly studied of which are proteins), the labile nature of proteins and range of polydispersity of naturally occurring macromolecules such as polysaccharides and humic substances add a particular complexity to the fouling mechanisms. In addition, the interaction between the suspended colloids or those in the deposited cake in a mixed species environment has the potential to signicantly change the nature of the foulant layer in terms of resistance and reversibility, even for simple model systems. In this section, the heuristics providing insights to fouling in such mixed species environment are considered in the context of broad observations and commonly used tools to decipher them. 2.1. Concepts of critical and sustainable ux in mixed species environment Optimizing ux to control fouling has been pursued since the mid-1980s. Moderating TMP differences within modules had been utilized to reduce excessive localized fouling. While much of the existing literature has been performed under constant pressure conditions, the use of constant ux and monitoring of resultant TMP rise have proved to be particularly useful in the context of monitoring fouling in complex uids and is currently the mode of choice in many MBR applications (Fig. 1). Typically, increasing ux steps are imposed and the TMP monitored for its stability at each step. When the TMP is no longer stable at each ux step and increases rapidly to indicate rapid

Fig. 1. (a) Critical ux determination by the ux-stepping method and (b) resulting data obtained during the study of the effect of membrane state (BW: backwashed, chemically cleaned or new) on the fouling rate (dP/dt) [14].

accumulation of foulants, this is usually referred to as the critical ux. The original critical ux hypothesis for MF assumes that a critical ux exists below which a decline of permeability with time does not occur, and above which fouling is observed [12]. Since this rst denition, the critical ux concept has been rened with numerous different meanings, denitions and methods of determination, reviewed in [13,14]. Two distinct forms of the critical ux concept have been dened, with, respectively, no fouling and little fouling occurring at sub-critical operation for the strong and weak forms. In practice, the ux obtained during sub-critical ux (strong form) equates to the clean water ux obtained under the same conditions. In the alternative weak form, the sub-critical ux is the ux rapidly established and maintained during the start-up of the ltration, but does not necessarily equate to the clean water ux. The critical ux depends on the back transport provided by the crossow or turbulence generated by imposed liquid ow and/or bubbling as well as the specic solutemembrane interactions, which are affected by charge and hydrophobicity. Solute size also plays a signicant role in determining the regime of back transport whether it is diffusive or inertial lift [8]. High local concentrations that promote local aggregation due to concentration polarization will also determine the cohesiveness of the foulant layer.

For single particles, the force balance between convective drag to the membrane and back transport due to crossow can be analyzed with various shear enhanced diffusivity models [15] to predict the critical ux at which deposition occurs at a particular hydrodynamic condition. The reversibility of the deposition has been less well documented but can be assessed by examining the hysteresis of the TMP versus ux prole [16,17]. The hysteresis technique is the most recently formalized technique to determine critical ux. By using this method, Chen et al. [16] studied the transition from concentration polarization to cake formation during the membrane ltration of colloidal silica. Recently, Espinasse et al. [18] published a proposed standard method for the determination of critical ux using hysteresis effects. In this method, critical ux is dened as the lowest ux that creates an irreversible deposit on the membrane. Models proposed by Bacchin and co-workers [1921] to incorporate intermolecular forces with convective transport have been used to predict regimes where particle aggregation begins to dominate on the membrane surface. For macromolecules, the apparent critical ux can also be determined however in these systems, background adsorption occurs even with no convection and a slow increase in membrane resistance is always detected even at low uxes. The kinetics of this adsorption, particularly for proteins, can be crucial as conformational changes of the initial adsorbed layer can change the surface chemistry of the membrane surface with time. This continual adsorption may close off smaller pores that cause redistribution of ow throughout membrane structure. While concentration polarization may not be initially present for these small macromolecules in microltration, the adsorption followed by retention of an increasing fraction of the feed can result in rapid loss in transmission and resultant polarization. High initial ux (typical of constant pressure experiments) can also generate aggregation of some macromolecules such as proteins [22,23]. For complex uid systems, one common practice to experimentally determine the critical ux value is to incrementally increase the ux for a xed duration. This leads to relatively stable TMP at low uxes (indicating little fouling), and an everincreasing rate of TMP rises at uxes beyond the critical ux values [16,2428]. Since zero rate of TMP increase is generally not attained in ltration of complex uids, no critical ux, in its strictest (or strong) denition, can be dened (Fig. 1). In uids with both macromolecules and particulates, membrane fouling takes place even at low ux rates, but changes dramatically when the so-called critical ux (in its weak form) is reached. Although apparently straightforward in principle, the precise identication of the critical ux value from ux-stepping experiments strongly depends upon the conditions used (step duration, step height, initial state of the membrane) [14]. The most important parameter remains the step height, which needs to be kept as small as possible for higher accuracy in the determination of the critical ux value [13]. Unfortunately, no standard protocol exists, such that the comparison of critical ux values reported in the literature is difcult. The determination of the exact value of the critical ux (i.e. the passage from little to severe fouling) is then left to the judgement of the researchers, although rigorous

mathematical expressions have been reported [14]. In spite of the arbitrary aspect of this method, critical ux determination by this short-term experiment remains an efcient approach to assess the fouling behavior of a given ltration system and to compare different operating conditions. Interestingly, this method was recently used as a standard test to assess the fouling propensity of an MBR on a daily basis [29]. This approach allows the plotting of fouling intensity against an MBR parameter such as biomass characteristics. It is also now generally accepted that the short-term determination methods for the critical ux (especially the ux-stepping approach) does not yield predictive absolute permeability data for extended operation of complex uids. For example, the fouling rate (dTMP/dt) values measured for long-term experiments are always signicantly lower than the equivalent values measured for the short-term ux-step experiments [13,14,3032]. In addition, a second phase of TMP increase has been observed when long-term ltration was carried out at sub-critical ux (experimentally determined in short-term experiments) even for simple model feeds, such as alginate solutions. A number of models have been proposed to account for the development of a second phase of TMP increase which may occur after hundreds of hours of operation [17,31,33,34]. Most of the models have focused on the potential for slow pore closure or blockage that results in high local uxes due to redistribution of ow and subsequent rapid fouling. The challenge remains to use short-term experimental data to project long-term fouling characteristics in such mixed systems where foulant inventory and fractionation may play important roles. Thus the focus may be shifted to considering a sustainable ux (see Section 4.2.2) where reversibility of the foulant deposition and global operational constraints for productivity and costs are taken into account. 2.2. Effect of operating modes on performance Constant pressure ltration behavior is typied by a rapid ux decline at the start of ltration followed by a more gradual decrease until a steady-state or a pseudo-steady-state ux is reached. Four ltration models (Table 1), originally developed for dead-end ltration [35], have been proposed to describe the initial ux decline. Comparison between operating modes (constant pressure and constant ux) have been limited [27,36,37]. Constant ux operation avoids excessive fouling of membranes as well as being cost effective for submerged membrane operations [27]. Vyas et al. [38] investigated the performance of different combinations of constant pressure and constant ux crossow microltration of lactalbumin suspensions, since in their case the critical ux is too low to be an economic operation. It was found that operating under constant ux just above the critical ux followed by constant TMP operation causes severe membrane fouling. It appears that during the constant TMP period, small particles continued to permeate through the relatively thinner cake, formed during the low constant ux ltration, into the membrane pores. In contrast, constant TMP operation followed by very low constant ux operation can offer scope to reduce

Table 1 Empirical dead-end ltration equations Law Cake ltration Complete blocking Intermediate blocking Standard blocking Physical cause Boundary layer resistance Pore blocking Long-term adsorption Direct adsorption Description Deposit of particles larger than the membrane pore size onto the membrane surface Occlusion of pores by particles with no particle superimposition Occlusion of pores by particles with particle superimposition Deposit of particles smaller than the membrane pore size onto the pore walls, reducing the pore size Equation t/V = AV + B ln(J/J0 ) = At + B 1/J = At + B t/V = At + B

Where V is the cumulative volume of permeate at time t, J the ux, and J0 is the initial ux.

surface fouling by reducing the convective force towards the membrane. While fouling is generally observed as being slower in constant ux operation, there is some evidence that the deposition under these low fouling conditions may be more irreversible as the resultant mechanism tends to be predominately internal fouling by macromolecular species. Constant ux operation may generate a substantial initial deposit, but its effect on subsequent deposition of macromolecules may be benecial in some circumstances by serving as a prelter for species which may otherwise inltrate more deeply into the membrane pores. In addition, the constraints of productivity in terms of ux and permeation of targeted species for applications such as fermentation redene the optimal ux operational mode. Intermittent ltration combined with continuous crossow may allow deposits to relax as long as the particles are still labile. In appropriate regions suggested by Bacchin et al. [19] where coagulation or aggregation has not occurred, this approach may allow removal of foulant cakes. However, for many biological solids, the cohesive strength of the cake may be signicant and proteins adsorption and gel formation result in strong attractive bonds to the membrane materials. 2.3. Cake structure and the effect of mixed species on cake morphology Once the cake is formed on the membrane surface, the cake layer offers an additional resistance for ltration. The permeability of the cake layer can be affected by ux, electrostatic interactions, and particle size. General observations by Petsev et al. [39] include: When salts do not cause aggregation in the feed, the permeability of the cake layer sharply decreases with the increase in electrolyte concentration. The permeability of the cake layer sharply decreases with the increase in permeate ux because the increased ux results in a more compressed cake layer. The permeability of the cake layer increases with the surface potential of the particles due to the increase in the interparticle repulsion. However, above a certain value of surface potential, a plateau value for the permeability is reached. The permeability of the cake layer passes through a minimum with the increase in the particle size.

The reason for such behavior is that for very small particles, inter-particle repulsion (electrostatic repulsion) exerts a dominant effect on the voidage of the cake layer. In this range of particle size, the inter-particle repulsion decreases with the increase in particle size resulting in the decrease of permeability [40]. After a certain value of particle size, the effect of the electrostatic force becomes negligible and the permeability increases with the increase in particle size. Fane et al. [41] observed a similar dependency of the permeability of the cake layer on particle size. They proposed a different explanation based on the counter balance of the Brownian diffusion (dominating for smaller particles) and particle migration due to hydrodynamic forces (dominating for larger particles). The development of the cake layer during microltration was also studied [4247]. Cakes formed in the crossow mode may have higher specic cake resistances than cakes formed in dead-end ltration and may even increase with the increase in membrane resistance [48]. Many of these observations can be explained by size dependent particle deposition and the dependence of the specic resistance on the particle size. Based on the mass transfer mechanisms, there is a maximum diameter of particle that can deposit on the membrane surface. As the crossow velocity increases, the cut-off diameter decreases, allowing that smaller particles to deposit on the membrane surface. Thus specic resistance may increase. In addition, cake formation during crossow tends to eliminate larger particles from the deposits, leading to cake containing a ner fraction of the particle size distribution. Plugging and catastrophic reduction in permeability of the retained cake is another potential cause of the two stage TMP increase during sub-critical ux operation indicated earlier [34]. However, Keskinler et al. [44] reported that the specic resistance for lower crossow velocities was greater than the one obtained in higher crossow velocities for all yeast cell concentrations tested. In contrast, during the membrane ltration of monodisperse latex particles, no effect of stirring speed was found on the specic resistance values [43]. The compressibility indexes of the cake have been found to be different in crossow and dead-end ltration. Keskinler et al. [44] found that non-living yeast cakes formed in the crossow mode are more compressible than cakes formed in dead-end ltration. The compressibility index was found to be 1 and 0.39 for the crossow and dead-end ltration, respectively. Xujiang et al. [49] found similar trends during microltration of talc suspensions. By contrast, Mota et al. [50] reported that, for spherical particles, the compressibility index (n) both in

dead-end and crossow ltration were similar basing on the studies at relatively low crossow velocity. Tanaka et al. [45] found lower compressibility index in crossow ltration than that in dead-end ltration during microltration of rod-shaped B. subtilis, which are 0.6 and 0.8, respectively. They explained these differences as follows. During the crossow ltration at lower TMP, the cells were arranged by the shear from the beginning of cell deposition on the membrane surface, thus the cake showed higher specic cake resistance than that in the dead-end ltration. While at higher TMP, the cells deposition at the initial stage of crossow ltration in particular tended to deposit in a manner similar to that in dead-end ltration due to the high permeate ux; therefore the specic cake resistance became close to that in the dead-end ltration. This may provide the reason that the compressibility is lower in crossow ltration than dead-end ltration. Hughes and Field [51] recently showed that increasing shear stress reduced the amount of reversible fouling in yeast ltration but the irreversible component remained constant. The potential for size segregation and lateral transport for yeast cells near the membrane wall has also been observed [52]. Foley [53] recently reviewed factors affecting lter cake properties of microbial suspensions. For complex uids such as membrane bioreactors efuent, fermentation broths, and natural organic matter, the fouling interactions of the colloidal component are affected by the potential for small macromolecules to penetrate and adsorb into the membrane structure and foulant cake structure. On the other hand, colloids or particles can affect the initial deposition of the macromolecules by adsorbing them on their surfaces or providing a secondary layer that entraps aggregates of these macromolecules. Studies to elucidate this phenomenon have been carried with yeast and protein mixtures. Davis and colleagues [5456] showed that the presence of yeast actually prevented fouling of bovin serum albumin (BSA) in microltration as the yeast layer on the membrane surface captured the BSA aggregates and prevented them from fouling the internal structure of the membrane. In this case, the cake layer formed by the yeast particles can be considered as a prelter (Fig. 2). They observed higher protein transmission and higher ux in the presence of the yeast cake than in its absence. Recent studies by Ye and Chen [57] showed that the critical ux of the mixtures of yeast and BSA showed little change from critical ux measured for yeast alone; however, the reversibility of the deposited formed by these mixed layer is substantially reduced. Thus the macromolecules can serve to bind the particulates together. Results

Fig. 2. Cake layer as prelter.

with alginate, a microbial polysaccharide, showed increasing specic resistance with time, indicating a consolidation of the cake layer formed which may be due to inltration of small fractions of the alginates among the alginate aggregates initially trapped by the microlter. When both alginate and protein are present, the transmission of both components was reduced while the compressibility of the mixed deposit was increased. Thus the rigidity and compressibility may vary substantially depending on the chemical nature of the extracellular components bound or soluble in MBR or fermentation broths. Some researchers indicated that particles in the mixed feed solution determine the ux behavior during the membrane ltration. Timmer et al. [58] found that the small quantities of silicates completely determined the ux behavior in the crossow microltration of -lactoglobulin solutions. Causserand et al. [59] studied the permeability changes in clay cake due to protein adsorption. A minimum limiting ux was found at the isoelectric point of the clayprotein complex. Interestingly, they found that at higher pH values, the mixture behavior was similar to the protein, whereas below pH 4.5, the limiting ux was similar to those observed for the ltration of clay suspensions alone. By optimizing the electrostatic interactions between proteins and an adsorptive surface like clay, Causserand et al. [60] improved protein fractionation and decreased membrane fouling by the protein, which was attributed to the formation of a secondary membrane by clay particles on top of the original particles. Hwang et al. also showed that capture of BSA in bed of latex particles can be related by standard capture equation for deep-bed ltration. Interesting studies by van Oers et al. showed how the presence of silica sols can reduce rejection of polyethylene glycol (PEG) and dextran by providing a high polarization layer (unstirred) zone near the membrane [61]. In contrast, the rejection of PEG and dextran increased in the presence of BSA. The compressibility of the BSA layer leads to highest rejection occurring at the highest pressure of ltration. The impact of large particles on the fouling process is not easy to gauge. Researchers have indicated that fouling can be reduced by adding suspended solids during UF of organic molecules such as polysaccharides and proteins. Panpanit and Visvanathan [62] investigated the role of bentonite addition in the UF for oil/water emulsions. It was found that the addition of bentonite up to a certain concentration dramatically decreased the membrane fouling. This was because the reduction of oil/water emulsions concentration by bentonite adsorption and the formation of larger particles when oil/water emulsion contacted with bentonite. However, beyond the limiting concentration, the ux improvement gradually declined, possibly due to the formation of packed cake of particles on the membrane surface. This composite cake structure is illustrated in Fig. 3. Recent studies with bentonite and alginate mixtures during constant ux MF showed formation of a bentonite cake near the membrane while the alginate formed a viscous layer above the cake. Particle velocities through this viscous layer dropped steadily as ltration time proceeded, indicating densication of the viscous gel layer (Fig. 4). This may provide insight into the cohesive and transport characteristics of such composite layers. In contrast, compact cellular or particulate cakes which form with swollen macromolecular

Fig. 3. Composite cake structure.

in carbon source for the fermentation of C. glutamicum affected the microltration performance. The specic cake resistance of cells cultivated with sucrose was half as much as those cultivated with glucose at neutral pH, and were almost the same below pH 4.0. The authors attributed these differences in specic cake resistance, as well as their pH dependencies, to the higher hydrophobicity and lower surface charge of cells grown on sucrose. By performing extracellular matrix modication of marine bacteria SW8 with a proteolytic enzyme and a chelating agent, the important role of matrix in resistance was conrmed by the changes of ux and specic cake resistance [68]. 2.4. Effect of membrane morphology and surface chemistry on fouling mechanisms Conventional wisdom generally attributes lower fouling to smooth hydrophilic membranes with high porosity and narrow pore size distribution. This has been supported by extensive work with various biological uids, particularly proteins solutions [10,11,70,71]. Reduction in the macromolecular adsorption with hydrophilic surfaces or by mitigating charge interactions will reduce the rate of pore closure due to this mechanism. Metsamuuronen et al. [72] reported that much lower critical uxes were observed for the ultraltration of bakers yeast when a hydrophobic polysulfone membrane was used as opposed to a hydrophilic regenerated cellulose membrane. This phenomenon is more obvious at pH 6 where both membranes have a zeta potential of zero. By using matrix-assisted laser desorption ionization mass spectrometry (MALDI-MS) for quantitative analysis, Chan et al. [73] studied the membrane fouling by protein mixtures on hydrophilic and hydrophobic 30 kDa molecular weight cut-off (MWCO) UF membranes. It was found that, for the hydrophobic membrane, the deposition exceeded quantities corresponding to a monolayer above and below the apparent critical ux. When a hydrophilic membrane was employed, coverage in excess of a monolayer was only found above the apparent critical ux. Interestingly, while high molecular weight proteins appear to dominate the apparent critical ux, the predominant proteins observed on the membrane by this technique tended to be the lower molecular weight species [74] that penetrate the pores. In mixed species feeds, the surface chemistry of the membrane may be masked by adsorption of the multitudes of macromolecular species thus the benets of hydrophilicity may be obscured during the long-term fouling. At a given xed ux, one would initially expect the pore size of the membrane to be irrelevant to the convective force exerted on the particles and to any back diffusion or shear induced diffusion effect. However, as pore size decreases, hindered transport of macromolecules exacerbates local polarization and the potential for aggregation and fouling. The local porosity and associated local convective velocities as opposed to average ux across the whole membrane surface also need to be considered when comparing membranes with widely varying porosities. The inuences of pore size on the fouling were found to differ in various studies. In the microltration of 0.4 wt.% BSA solutions, Chen [75] found that the critical ux increased with pore size when track-etched membranes of pore size 0.1, 0.2 and

underlayer may disengage spontaneously if the cake build-up is sufcient to create high shear stress due to crossow [63]. In systems with the microorganisms, the likelihood of different cell wall properties precluded researchers from making denitive statements regarding the effects of cell size and shape on ltration characteristics. In this context, Foley and his coworkers used polymorphic microorganisms to conduct a detailed investigation of the effect of cell size and shape on ltration behavior [48,6467]. The shape of this microorganism, ranging from yeast-like to lamentous, could be varied in a controlled way by altering its growth conditions. The structure of the cell wall was reasonably constant and independent of the cell shape. The results clearly showed that the specic cake resistance and compressibility of the microbial lter cakes was strongly related to cell morphology, in particular the mean aspect ratio of the cells. The potential errors in calculating specic resistance and tortuosity is signicant in mixed species cakes and the effect of cell shapes and structure in a compressible media may require better understanding of the associated extracellular material [50,68]. Recent work by Ohmori and Glatz [69] and earlier work by Hodgson et al. [68] have shown that the ltration properties of microbial suspensions were dependent not only on the cell shape but also on the physical characteristic of the cell and associated extracellular matrix. Ohmori and Glatz [69] found that changes

Fig. 4. Velocity prole during ltration of binary model solution with direct observation apparatus (500 mg/l alginate50 mg/l bentonite solution; apparent bulk velocity = 2 mm/s; constant ux of 56 l/m2 h). The background picture shows the fouling layer after 2 h of ltration.

0.4 m were used. In comparison, Wu et al. [76] investigated the effect of membrane pore size (50 kDa, 100 kDa and 0.2 m) on critical ux for three types of feed uids. For all feed uids tested: 0.5% silica, 0.15% BSA and 5% yeast cell suspension, the critical ux decreased with increasing membrane pore size. Narrow pore size distributions reduce the inhomogeneous ow distribution between pores that lead to preferential deposition and blockage of large pores [77,78]. Similarly, high porosity means that local ux at the pore entrance will be reduced. Membranes with interconnecting pore structures also have the advantage that surface blockage can be mitigated [79]. Membrane morphology will determine initial macromolecular transmission and fouling mechanisms, particularly at low ux operation. The transition between pore closure and cake formation is critical in the fouling progression in mixed species feed. As the effective pore size is reduced, the local ux increases, increasing the convective forces to the pore. Larger particles are then pulled in and accelerate the foulant build-up. Typically, membrane blocking laws (constant ux and constant pressure mode) have been used to establish when this transition between pore blockage and cake formation takes place [80]. Ho and Zydney [81] have developed a combined pore blockage and cake formation model, with the cake layer only forming over the regions of the membrane that have already been blocked by the initial deposit in the membrane pores. Unlike most prior pore blockage models, it was assumed that some uid is still allowed to ow through the pores blocked by large aggregates. The model was successfully used to analyze the protein fouling [81], alginate fouling [33] and humic acid fouling [82] during microltration. 2.5. Summary In complex uids, the interactions between the macromolecular and particulate components of the feed can result in unexpected and rapid changes in fouling. The kinetics and inventory of macromolecules adsorbing will dictate the initial fouling phase. Progressive closure of pores or membrane surface results in a change in transmission and species convected to the surface and the foulant cake. While the initial low fouling phase at low ux (or sub-critical ux) features slow progressive adsorption of macromolecules on the membrane surface, a more rapid fouling phase then occurs. During that period, pore closure results in enhanced rejection of macromolecules and deposition of larger particles (Fig. 5). Evolution of this foulant cake and its irreversibility depend on both its composition and the hydrodynamic environment under which it was established. The interaction between particulate and macromolecular fouling needs to be considered with many of the same complexities observed in fouling studies of natural organic matter. Macromolecular fouling can increase particulate adhesion, but particles can affect the transmission and inltration of macromolecules into the membrane pores. Greater understanding of the foulant structure in mixed specie systems will allow better control measures to prevent foulant build-up or to disengage the foulant layer. The lessons learnt from such studies are important for understanding fouling in MBRs.

Fig. 5. Progressive pore blockage leading to rapid TMP increase.

3. Roadmap for MBR fouling parameters All the parameters involved in the design and operation of MBR processes have an inuence on membrane fouling. For the purpose of this review, three categories of factors are dened, i.e. membrane and module characteristics, feed and biomass parameters and operating conditions (Fig. 6). While some of these parameters have a direct inuence on MBR fouling, many others result in subsequent effects on phenomena exacerbating fouling propensity. The complex interactions between these parameters complicate the perception of MBR fouling and it is therefore crucial to fully understand the biological, chemical and physical phenomena occurring in MBRs to assess fouling propensity and mechanisms.

Fig. 6. Factors affecting fouling in submerged MBRs.

3.1. Membrane characteristics 3.1.1. Physical parameters 3.1.1.1. Pore size and distribution. The effects of pore size (and distribution of pore size) on membrane fouling are strongly related to the feed solution characteristics and in particular the particle size distribution. Depending of the pore size and the type of biomass ltered, results reported in the literature have shown opposite trends. If particle size is smaller than pore size, pore blocking and/or restriction is expected. It is therefore expected that large pore membranes like MF would present higher fouling propensity compared to UF membranes. Table 2 reports results obtained in 11 studies during which the pore size effects have been assessed by different fouling parameters. It is quite clear from this table that the pore size alone cannot predict hydraulic performances as no general trend was observed between these two parameters. The complex and changing nature of the biological suspension present in MBR systems and the large pore size distribution of the membrane generally used in MBR are the main reasons for the undened general dependency of the ux propensity on pore size [83,84]. Additionally, the duration of the experiment and other operating parameters like crossow velocity (CFV) and constant pressure or constant ux operation have a direct inuence on the determination of the optimization of the membrane pore size (Table 2). For example, when MF and UF membranes were compared in a similar environment (with a CFV of 0.1 m/s), the MF membrane produced a hydraulic resistance around twice that of the UF membrane. In that same study, the fouling behaviors of the MF and UF membranes were different when operated at higher CFV. This was due to the effect of CFV on critical ux of particulates (Section 3.3.1). Interestingly, the dissolved organic carbon (DOC) rejection of both membranes were similar after 2 h of operation, indicating the creation of a dynamic membrane layer on the MF membrane [85].

The long-term effect of UF membrane pore size on hydraulic performances has been assessed by He et al. for anaerobic MBR operated under constant TMP [87]. The smallest MWCO tested (20 kDa) featured the largest permeability lost within the rst 15 min of ltration when compared to 30, 50 and 70 kDa membranes. However, when operated for extended time (over 100 days) with regular hydraulic and chemical cleaning, the largest MWCO membrane (70 kDa) experienced the greater fouling rate, as 94% of its original permeability was lost, compared to only 70% performance decrease for the other three membranes. As a result, the 30 and 50 kDa membranes provided the best overall hydraulic performances, indicating the possibility of an optimum membrane pore size for a given application and for a given ltration time. These results also revealed that the experiment duration is crucial to fully assess the fouling propensity of a membrane. Similar trends showing the time dependency for large pore MF with the highest initial fouling for the smaller pore and the greater long-term fouling for the larger pore were reported for pore size ranging from 1.5 to 5 m operated at constant TMP [90]. While the quest for the highest steady-state permeability is probably desirable, it is important to be aware that conclusions derived from ux decline data could be sometimes deceptive, as an intrinsically high ux membrane may appear to foul more for the same increment of resistance. It is expected that smaller pore membranes would reject a wider range of materials, and the resulting cake layer features a higher resistance compared to large pore membranes. However, this type of fouling is more reversible and is easily removed during the maintenance cleaning than fouling due to internal pore clogging obtained in larger pore membrane systems. The irreversible fouling, due to the deposition of organic and inorganic materials onto and into the membrane pores is the main cause of the poor long-term performances of larger pore size membranes. However, when testing membranes with pores ranging from 0.4 to 5 m (at constant TMP), Gander et al. observed the opposite results, i.e. higher initial fouling for large pore and

Table 2 Effect of pore size on MBR hydraulic performances Membranes tested 0.1,0.22, 0.45 m 20, 30, 50, 70 kDa 70 kDa, 0.3 m 30 kDa, 0.3 m 0.1, 0.2, 0.4, 0.8 m 200 kDa, 0.1, 1 m 0.3, 1.5, 3, 5 m 0.4, 5 m 0.01, 0.2, 1 m 200 kDa, 0.1, 1 m 0.05, 0.4 m
a

Optimum 0.22 m 70 kDa 50 kDa 70 kDa 30 kDa 0.3 m 0.8 m 1 m 5 m 0.3 m 0.4 m No effect No effect 0.1 m 0.05 m

Test duration 20 h 110 min 110 days 8h 2h Merge n/a 3h 25 min 45 days 1 day From 50 days Few hours n/a n/a

Other High concentrated Feed, anaerobic CFV = 0.1 m/s CFV = 3.5 m/s Based on critical ux test Based on critical ux test Anaerobic

References [86]a [87] [88] [85] [89]a [84]a [90] [91] [92]a [93] [94]

Constant ux operation, non-marked references are constant TMP operation.

signicant ux decline when the small pore membrane was used over an extended period of time [91] (Table 2). Characterization of the molecular weight (MW) distribution of the compounds present in the supernatant of MBRs operated with four pore sizes (ranging from 0.1 to 0.8 m) has also been presented [89]. With apparent lower fouling rate, the 0.8 m pore size MBR featured a slightly higher concentration of most of the macromolecules present in the bioreactor supernatant. However, it seems unlikely that the small differences in MW distribution are the main cause of the various fouling rates observed between the four MBR systems. In another study based on short-term experiments, subcritical fouling resistances and fouling rates increased linearly for membrane resistances ranging from 0.4 to 3.5 1011 m1 , corresponding to membrane pore size from 1 down to 0.01 m, respectively [84]. These results indicated the creation of a dynamic layer of greater overall resistance for more selective membranes under sub-critical conditions. However, it was also postulated that increasing the pore size may decrease the deposition onto the membrane at the expense of internal adsorption. Long-term trials conrmed this theory as progressive internal deposition eventually leads to catastrophic increase in resistance [14,95]. This again emphasizes the importance of test duration in fouling studies. 3.1.1.2. Porosity/roughness. Membrane roughness and porosity were suggested as potential reasons for the different fouling behaviors observed when four MF membranes with nominal pore sizes narrowly ranged between 0.20 and 0.22 m were tested in parallel [96]. The four membranes were operated under the same constant pressure, and therefore produced different initial uxes. The track-etched membrane, with its dense structure and small but uniform cylindrical pores, featured the lowest resistance due to pore fouling. In contrast, the other three membranes presented interwoven sponge-like microstructures and were more prone to pore fouling due to their highly porous network. Although all membranes featured similar nominal pore size, polyvinylidene uoride (PVDF), mixed cellulose esters (MCE) and polyethersulfone (PES) membranes presented different fouling behaviors. While fouling was mainly due to cake formation for the PVDF and MCE membranes, pore blocking was responsible for 86% of the total hydraulic resistance when the PES membrane was used. Overall, the PES membrane showed a 50% higher fouling resistance than the PVDF and MCE membranes. It was suspected that membrane microstructure, material and pore size distribution were all affecting MBR fouling signicantly [96]. Comparison between two microporous membranes prepared by the stretching method revealed the inuence of the pore aspect ratio (mean major axis length/mean minor axis length) on fouling in an MBR. With both membranes the average pore size and pure water ux were identical, but less fouling was observed with the membrane having the higher pore aspect ratio (elliptical pore) rather than with the circular pore membrane [97]. With roughness values (measured by AFM) ranging from 2.4 to 33.2 nm, for 20 and 70 kDa MWCO membranes, respectively, initial fouling was observed to decrease while irreversible resistance increased [87]. However, in this

study based on an anaerobic MBR, membrane morphology and pore size were changing simultaneously, so it was not possible to clearly determine the effect of roughness on MBR fouling. However, an assumption was made that the large lling-in points present on rougher membranes are more prone to the creation of fouling layers, compared to the fewer and smaller crevices observed on smoother membranes [87]. Detailed discussion about the effect of membrane surface properties on cell attachment could be found in [98], while Ho and Zydney gave more details about membrane morphology and MBR fouling [99]. 3.1.1.3. Membrane conguration. The current trend in MBR design tends to favor submerged over sidestream congurations in the majority of the studies dealing with domestic wastewater treatment. As a result, comparison between these two MBR congurations will be discussed only briey in this review, but more details can be found in [100103]. Based on short-term critical ux tests, a direct comparison between submerged and sidestream MBRs showed that similar fouling behavior was obtained when the two congurations operated at supercial gas velocity (UG ) of 0.070.11 m/s and supercial liquid velocity (UL ) of 0.250.55 m/s for submerged and sidestream, respectively [102]. An increase of UG in the submerged MBR was also found to have more effect in fouling removal than a similar raise of CFV (or UL ) in the sidestream conguration (also see Section 3.3.1). This may be due to the benet of unsteady state ow achieved by bubbling. In submerged MBR processes, the membrane can be congured as vertical at plates, vertical or horizontal hollow ne bers (ltration from out-to-in) or, more rarely as tubes (ltration from in-to-out). Although the tubular conguration is generally preferred for sidestream processes, the effect of the lumen size on submerged MBR fouling has been investigated [104,105]. While hollow ber modules are generally cheaper to manufacture, allow high membrane density and can tolerate vigorous backwashing, uid dynamics and distributions may be probably easier to control for at plate and tubular membranes, where the membrane channel width is well dened [106]. As a result, hollow bers may be more prone to fouling and require more frequent washing and cleaning. An interesting discussion of the relative performances of hollow bers and at plate membranes was initiated by Gunder and Krauth [107] and revealed the better hydraulic performance of the at plate in their studies. Two types of submerged MBR of comparable size, operated for the same length of time for sewage treatment have also been compared [108]. The differences observed were mostly due to the different operating and maintenance conditions (see Section 4.1) rather than the module designs per se. Although the price of the at plate MBR is estimated to be 2025% higher than hollowber-based-systems, fouling rate and maintenance operation are generally less for the former conguration. This observation may be due to the design ux at which MBR were operated in this study, i.e. 2027 l/m2 h for at plate and 2333 l/m2 h for hollow ber [108]. The backwashing requirement of the hollow ber MBR (up to 25% of the permeate volume [108]) may also slightly complicate the process. In another study, the effect of

membrane conguration was assessed when hollow ber and at plate MBRs (featuring similar pore size of 0.4 m) were used for high-strength wastewater treatment [109]. Once both systems were operated at similar ux, it was found that the at sheet MBR fouled slightly more and could not recover its original performance after water cleaning. However, chemical cleaning managed to remove most of the fouling (probably due to pore blocking in this specic case). Finally, each conguration has specic footprint, airow requirement, and integrity testing which may favor one process over another one for a given application. More details about these two congurations (with a focus given on aeration intensity) and the effect of their physical parameters are available from Cui et al. [106]. Amongst the numerous membrane manufacturers, Kubota (at plate conguration), Zenon, Mitsubishi and US Filter (hollow ber conguration) are the main membrane suppliers for MBR systems. Few large-scale studies based on comparison of these commercially available MBR systems have been conducted. The city of San Diego, California, and the research consultant, Montgomery Watson Harza, have been evaluating the MBR process through various projects since 1997, including feasibility of using MBR to produce reclaimed water [110,111], optimization of MBR operation, and parallel comparison and cost estimations of the four leading MBR suppliers [112]. MBRs were evaluated for their ability to produce high quality efuent and to operate with minimum fouling. In terms of hydraulic performances, it was shown that all four processes were able to cope with ux rates exceeding 33 l/m2 h and HRTs as low as 2 h. A 6-year-development programme has also been initiated for the introduction of MBR technology in the Netherlands market. Started in 2000, a comparative study of four 750 m3 /dayMBRs carried out by DHV water has been reported [113,114]. Finally, three MBR plants, treating a design ow of 300 m3 /day each, have been operated in parallel during 2003 and 2004 in Singapore. This most recent study reported MBR power consumption of less than 1 kWh/m3 of treated water [115], while energy consumption around 1.9 kWh/m3 was reported for 2001 [116] and up to 2.5 kWh/m3 in 1999 [117]. Although these three studies have been conducted with the MBR systems running in parallel (with the same inuent water), the MBR maximum ux, operating conditions and general design applied were those recommended by the suppliers, and therefore somewhat different for each system. This makes it difcult to make a fair comparison, so it is not possible to classify the MBRs as a function of their relative hydraulic performances, which need to be considered along with the cleaning protocols applied to each system (see Section 4.1). An important parameter for submerged hollow bers is likely to be packing density. The distance between adjacent membranes is suspected to directly impact on mass transfer and therefore the shear and aeration demands. Moreover, increasing the packing density could lead to severe clogging by gross solids and to the slower rise of bubbles, limiting their effect on fouling limitation. Experiments carried out with a model bundle of nine bers revealed the overall module performance to be much worse than that of an individual ber [118,119]. It was also clearly shown that the surrounded bers are less productive than the

outer bers. At high feed concentration and low cross-velocity, surrounded bers become completely blocked and eventually produce negligible ux. Finally, it was advised that the packing density should be lower than 30% in order for the bundle to perform similarly to single bers. In low packing density congurations, cake layers from adjacent bers do not interfere with each other and the effect of CFV may be more evenly distributed, limiting overall fouling [119]. A mathematical model based on substrate and biomass mass balance also revealed the signicant role played by packing density in the overall MBR performance, and the hydrodynamics of the biomass in particular [120]. The effects of other membrane characteristics including hollow ber orientation, size and exibility are discussed in the review of Cui et al. [106]. For hollow ber membranes used for yeast ltration, higher critical uxes were measured for slightly loose membranes (95%), with small diameter (0.65 mm) and greater length (80 cm) [121]. Contradictory results showing slightly higher specic ux for shorter membranes (30 cm cf. 100 cm) has been reported [122]. The pressure drop due to the permeate ow in the lumen of the hollow ber could be the main cause behind the effect of membrane length on rapid and severe fouling. Signicant pressure loss (up to 53 kPa) was measured for long bers (60 cm). Below the critical length of 15 cm, pressure loss was minimal at less than 11 kPa [97]. Further discussion of fouling distribution in hollow bers can be found elsewhere [31,123127]. 3.1.2. Chemical parameters 3.1.2.1. Hydrophobicity. Because of the hydrophobic interactions occurring between solutes, microbial cells and membrane material, membrane fouling is expected to be more severe with hydrophobic rather than hydrophilic membranes [92,128130]. In many reported studies, change in membrane hydrophobicity often occurs with other membrane modications such as pore size and morphology, which make the correlation between membrane hydrophobicity and fouling more difcult to assess. In a recent study for example, the contact angle measurement showed that the apparent hydrophobicity of polyethersulfone (PES) membranes decreased (from 55 to 47 ) with the increase in MWCO (from 20 to 70 kDa membranes, respectively) [87]. The effect of membrane hydrophobicity was studied in detail during comparison of two UF membranes of similar characteristics [131]. Based on the greater solute rejection and fouling and cake resistances reported for the hydrophobic membrane, the authors were able to postulate on the effects of membrane fouling on the removal performances of the MBR process. It was concluded that the greater solute rejection was mainly due to the dynamic layer formed by adsorption and/or sieving in the cake deposited on the membrane, and, to a lesser extent, due to direct adsorption into membrane pores and on the surface. Numerous anti-fouling studies have been based on membrane surface modication, and will be reviewed in Section 4.2.1. Surprisingly, Fang and Shi [96] indicated that membranes of greater hydrophilicity tend to be more vulnerable to deposition of foulants of hydrophilic nature. In MBRs, activated sludge contains substantial amounts of hydrophilic EPS, which has been identied as an important foulant (Section 3.2.5). However, in

this study, the most hydrophilic membrane also featured more open pores, which could be another reason for severe fouling. Notwithstanding the signicance of the membrane hydrophobicity on the early stage of the fouling formation, this parameter is expected to play only a minor role during extended ltration periods. Once initially fouled (i.e. conditioned), the membranes chemical characteristics would become secondary to those of the sludge materials covering the membrane surface. 3.1.2.2. Materials. Although featuring superior chemical, thermal and hydraulic resistances, ceramic membranes are not the preferred option for MBR applications due to their high cost. However, ceramic membranes have been successfully used for several MBR applications, such as treatment of high-strength industrial waste [132,133] and anaerobic biodegradation [134]. Ceramic membranes, in modules which require higher pressure and turbulence, are generally used in sidestream congurations. The benet of turbulence promoters in such MBR systems has been reported [135]. The potential advantage of using ceramic membrane was demonstrated in a test comparing 0.1 m ceramic with 0.03 m polymeric multi-channel membrane modules operated in sidestream air-lift mode. The ceramic MBR did not substantially foul for short-term experiments with uxes up to 60 l/m2 h, while the polymeric membrane critically fouled at around 36 l/m2 h [136]. However, in the same study, the overall cost of the ceramic membrane was reported to be around an order of magnitude more expensive than the polymeric materials. Finally, novel stainless steel membrane modules have recently shown good hydraulic performance and fouling recovery when used in an anaerobic MBR for wastewater treatment [137]. However, the large majority of the membranes used in MBRs are polymeric-based. A direct comparison between polyethylene (PE) and PVDF membranes clearly indicated that the later leads to a better prevention of irreversible fouling and that PE membrane fouled more quickly [138]. In that same study, the authors also mentioned that the composition of the irreversible fouling was dependant of the membrane material, as some fractions of the organic matter present in the biomass presented a higher afnity with certain polymeric materials. 3.2. Feedbiomass characteristics 3.2.1. Nature of feed and concentration Although the effects of wastewater properties on membrane fouling are undeniable for direct wastewater ltration [139141], fouling in the MBR is mostly affected by the interactions between the MBR membrane and the biological suspension rather than wastewater per se [88]. However, in the rare cases of using saline sewage as feed, the resulting higher fouling rate generally leads to a more frequent cleaning [142]. The most striking effect of the wastewater nature is on the physico-chemical changes in the biological suspensions [13,14]. For example, the protein fraction measured in the extracted EPS (eEPSp) has been found to be signicantly lower when biomass was fed with synthetic feed (chemical oxygen demand: COD of 460 mg/l) rather than with real sewage (COD of 140 mg/l). Simultaneously, the fouling rate was higher using synthetically fed MBR [14]. For

these reasons, the fouling propensity of the wastewater is indirectly taken into consideration during the characterization of the biomass (Section 3.2.3). 3.2.2. Biomass fractionation Activated sludge biomass can be fractionated into three idealized components, i.e. suspended solids, colloids and solutes. This approach has often been applied to account for the relative contribution of each biomass fraction on MBR fouling. The methodology applied to appropriately separate the biomass fractions varies from one study to another but remains a crucial step in the denitions of the different biomass fractions and therefore, the interpretation of the results. Unfortunately, no standard method exists. However, Fig. 7 shows a typical protocol where the biomass sample is centrifuged, the resulting supernatant is then ltered with a dead-end membrane cell, with the calculated hydraulic resistance (Rsup ) being attributed to colloidal and soluble species (Rcol and Rsol , respectively). Another portion of the biomass suspension is then ltered by a microlter (with nominal pore size of around 0.5 m). The fouling properties of this coarse-ltered supernatant are attributed solely to the soluble matter with resistance Rsol . Calculations assess the relative fouling contributions of the suspended solids and the colloids [143]. In another approach, the concentration of colloids was also characterized by the difference between the levels of TOC present in the ltrate passing through 1.5 m ltration paper and in the permeate collected from the MBR membrane (0.04 m membrane) [29]. Although fractionation methods may significantly vary for different studies (see references from Fig. 8), results are often reported in terms of hydraulic resistances for suspended solids (Rss ), colloids (Rcol ) and soluble species (Rsol ), the sum of which being the total resistance (Rt ). Although an interesting approach for studying MBR fouling, the fractionation experiment neglects any coupling or synergistic effects which may occur among the different components of the biomass. The interactions between each biomass fraction and the operating conditions are numerous and include the type of feed water used [144], permeability of the membrane, particle size and hydrodynamics conditions [143]. Examples of interactions between suspended and dissolved solids and membrane fouling

Fig. 7. Experimental method for the determination of the relative fouling propensity for the three biomass fractions.

Fig. 8. Relative contributions (in %) of the different biomass fractions to MBR fouling. For SRT increase from 8 days (1) to 40 days (2); F/M ratio of 0.5, results based on modied fouling index (3); based on ux reduction after 600 min of each fraction ltration (4); for SRT increase from 20 days (5) to 60 days (6).

seems to decrease fouling at low MLSS concentration (<6 g/l), more fouling is expected as the MLSS concentration increases above 15 g/l. The level of MLSS does not appear to have significant effect on membrane fouling between 8 and 12 g/l. Another study [153] reviewed the signicant effect of MLSS for concentrations lower than 5 g/l, and indicated that hydrodynamics (more than MLSS concentration) control the critical ux (Jc ) for greater MLSS levels [153]. This is only partially veried by the data reported in Table 3. More subtle studies showed apparent contradictory trends from data obtained in the same study. For example, the cake resistance (Rc ) was observed to increase and the specic cake resistance (c ) to decrease as MLSS increased. Although having similar meaning conceptually Rc and c seemed to behave inversely [146]. This can be reconciled by noting: Rc = c mc (1)

were discussed in Section 2.3. The protocol illustrated in Fig. 7 is limited because it relies on dead-end ltration tests with a specic membrane. However, studies on biomass fractionation have also been reported for crossow and submerged congurations. An attempt to compare results obtained from different studies is reported in Fig. 8 where relative contributions have been calculated. The relative contribution of the biomass supernatant (soluble and colloids, generally dened as soluble microbial products or SMP) to overall membrane fouling ranges from 17% [143] to 81% [145]. These wide discrepancies may surprise and are probably explained by the different operating conditions and biological states of the suspension used in the reported studies. They also conrmed the relatively low fouling role played by the suspended solids (biooc and the attached EPS) compared to those of the SMP (Section 3.2.6). In terms of fouling mechanisms, soluble and colloidal materials are assumed to be responsible for the pore blockage of the membrane, while suspended solids account mainly for the cake layer resistance [145] (Section 3.4.2). However, because MBRs are typically operated at modest ux, the formation of a biomass cake tends not to occur. The smaller species (like SMP) are much more likely to deposit. 3.2.3. Biomass (bulk) parameters 3.2.3.1. MLSS concentration. Often considered at rst sight as the main foulant parameter, MLSS concentration has indeed a complex interaction with MBR fouling, and controversial ndings about the effect of this parameter on membrane ltration have been reported. If the other biomass characteristics are not accounted for, the increase in MLSS concentration seems to have a mostly negative impact (higher TMP or lower ux) on the MBR hydraulic performances [146,147]. However, some authors have reported positive impact [26,148], and some observed insignificant impact [84,149,150]. The existence of a threshold above which the MLSS concentration has a negative inuence was also reported (at 30 g/l [151]). A more detailed fouling trend has been described by Rosenberger et al. [152]. While a rise in MLSS

where mc is the cake load/area of membrane. The cake load mc would tend to rise with MLSS concentration. Bin et al. observed the permeate ux to decrease (but at a lower fouling rate) when MLSS increased [154]. This was explained by the creation of a rapid fouling cake layer (potentially protecting the membrane) at high concentration, while progressive pore blocking created by colloids and particles was thought to take place at lower MLSS concentration. Since the value of Jc is often determined during short-term experiments, it is expected that Jc indicates the deposition of suspended solids rather than colloidal and soluble materials. As a result, the ux value at which the experiment is carried out, has a signicant impact on the determination of the effect of the MLSS concentration. Similarly, the test duration can be a factor. While MBR performances are expected to decrease for higher MLSS (at applied ux superior to Jc ), the MLSS concentration may not
Table 3 Inuence of shift in MLSS concentration (g/l) on MBR fouling MLSS shift Fouling parameters References [146] [96] [155] [154] [92]a [156]a [157]a [26]a [158] [84] [148]

Fouling increase 0.093.7 Rc : 21 to 54 1011 m1 and c : 18.5 to 0.7 108 m/kg 2.49.6 Rp : 9 to 22 1011 m1 718 Jc : 4736 l/m2 h (for SRT: 30100 days) 2.19.6 Jc : 138 l/m2 h 110 Jc : 7535 l/m2 h 215 Limiting ux: 10550 l/m2 h 1.622 Stabilized ux: 6525 l/m2 h Fouling decrease 3.510 Jc : <60 to >80 l/m2 h No (or little) effect 914 No impact on fouling rate 4.411.6 No impact between 4 and 8 g/l, slightly less fouling for 12 g/l 618 Similar fouling rates for J < 10 l/m2 h, and slightly lower fouling rates for higher J 415.1 Jc decreased from 25 to 22 l/m2 h 3.68.4
a

[24] [149]

Sidestream MBR.

play a signicant role in fouling propensity when the MBR is operated at low uxes. In that later case, EPS components and concentrations have more effect on the MBR fouling than the MLSS concentration (Sections 3.2.5 and 3.2.6). Contradictory results may also arise from the mode of ltration, i.e. constant ux versus constant TMP (Section 3.4). Empirically derived equations predicting ux performance have been proposed in numerous papers [96,159161]. However, these equations have limited use as they are generally obtained under very specic conditions and take into account some specic operating parameters and disregard some others. A mathematical expression linking MLSS concentration, EPS and TMP with cake specic resistance has been proposed by Cho et al. [162]. In this study, specic resistance did not change signicantly for MLSS ranging from 4 to 10 g/l and when the EPS and TMP were kept constant. The experimental method used for changing MLSS concentration can also signicantly impact upon biomass characteristics since biomass acclimatization periods are not always respected [147]. Although the removal performances are generally high for MBR processes, MLSS concentration also plays a signicant role in this regard. For example, an optimal MLSS concentration at 6 g/l was obtained based on the highest COD removal [163] and on the highest virus removal [164]. The lack of a clear correlation between MLSS concentration and any other foulant characteristics indicates that the MLSS concentration (alone) is a poor indicator of biomass fouling propensity [13,165]. These authors recommended the use of fundamental operating parameters like HRT and SRT for prediction of foulant production. This has been supported by the relatively stable foulant characteristics obtained once true steady-state was established in the bioreactor. Current studies tend to consider the non-settleable organic substances (rather than the MLSS concentration) as the main players in the fouling propensity in MBRs (see Sections 3.2.5 and 3.2.6). 3.2.3.2. Viscosity. In the MBR, like in conventional activated sludge processes, biomass viscosity is closely related to its concentration, and has been cited as a foulant parameter [166]. A critical MLSS concentration exists under which the viscosity remains low and rises only slowly with the concentration. Above this critical value, suspension viscosity tends to increase exponentially with the solids concentration [145]. This critical value was observed to change from 10 to 17 g MLSS/l for different operating conditions (conventional and hybrid (precoagulation/sedimentation) MBRs, respectively). Similar observations were reported for the behavior of the capillary suction time (CST), another parameter closely related to viscosity [30]. The importance of MLSS viscosity is that it modies bubble size and can dampen the movement of hollow bers in submerged bundles [121]. The net result of this phenomenon would be a greater rate of fouling. Increased viscosity also reduces the efciency of mass transfer of oxygen and can therefore effect dissolved oxygen (DO) [167]; fouling tends to be worse at low DO (see below). The effect of MLSS concentration on viscosity at different shear rates obtained from a submerged MBR is shown in Fig. 9. These results also indicate the pseudo-

Fig. 9. Viscosity obtained at different MLSS concentrations and shear rates [105].

plastic (or shear-thinning) property of the sludge obtained in MBR. 3.2.3.3. Temperature. Temperature impacts on membrane ltration through its inuence on the permeate uid viscosity [168]. The common approach to comparing hydraulic performance obtained at different temperatures is to normalize the operating ux at a reference temperature (generally 25 C). This could be done by applying a temperature correction factor [169]. To avoid the interference of the temperature effects on MBR fouling, non-linear regression between critical ux and temperature was obtained [29]: Jc,t = Jc,20 1.025t 20 (2)

Interestingly, experiments carried out at two sets of temperatures (1718 and 1314 C) featured different hydraulic resistances even after the ux had been normalized [170]. The greater resistances observed at low temperature were explained by four phenomena occurring in the system: (1) within that temperature range, the sludge viscosity (rather than permeate viscosity) was calculated to increase by 10%, reducing the shear stress generated by coarse bubbles, (2) intensied deoculation tend to occur at low temperature, reducing biomass oc size and releasing EPS to the solution, (3) particle back transport velocity, calculated with the Brownian diffusion coefcient (linearly related to temperature), is less at low temperature, and (4) biodegradation of COD was also reduced at decreased temperature, resulting in a higher concentration of solute and particle COD in the reactor [170]. This last phenomenon was also observed by Fawehinmi et al. [171] with higher SMP levels measured in an anaerobic MBR operated at 20 C rather than at 30 C. All of these factors are directly linked to membrane fouling, so it is expected to observe greater deposition of materials on the membrane surface at lower temperatures [158]. 3.2.3.4. Dissolved oxygen (DO). The average level of DO in the bioreactor is controlled by the aeration rate, which not only provides oxygen to the biomass but also tend to limit fouling formation on the membrane surface. The effects of DO on MBR fouling are therefore multiple and may include changes in

biolm structure, SMP levels, and oc size distribution [89]. As a general trend, higher DO tends to lead to better lterability, and lower fouling rate. This was explained by the lower specic cake resistance of the fouling layer which featured larger particle sizes and greater porosity [172,173]. As expected, signicant differences were observed in microbial communities and resulting biofouling when the MBR was operated under various DO levels (from 6 to less than 0.1 mg/l in [172]). Surprisingly, the COD in biomass suspension (i.e. an indicator for SMP level) decreased from 37 to 27 mg/l for DO of 3.4 and 0.9 mgO2 /l, respectively [174] and therefore cannot explained the hydraulic performances obtained in MBR operated at higher DO. Moreover, the contribution of SMP to membrane lterability was found to be a minimum compared to those of the physico-chemical properties of cake layer (i.e. particle size and porosity) [173]. In a study obtained with anoxic and aerobic sludges [175], oc deterioration was observed and used as a possible explanation for the higher fouling rates obtained for the denitrication assay. The effect of oxygen limitation causing a lowering of the cell surface hydrophobicity, and consecutive oc deterioration, was concluded to be the main reason for the worsen MBR fouling for anoxic conditions. In the MBR, fouling may also be due to the creation of a biolm layer on the membrane surface. As a general denition, bacterial biolm characterizes the population of microorganisms concentrating, depositing and/or growing at the solid/liquid interface [176]. As described later in the review (Section 3.2.5), the formation of biolms is possible through the active role of EPS which surround the microorganisms. Biolm properties such as adhesion strength (interaction between microorganisms and membrane) and cohesion strength (interaction between microorganisms themselves) can be determined and are directly dependant of the nature of the EPS [176]. As the thickness of the biological fouling layer increases with extended MBR ltration time, some biolm regions have been observed to become anaerobic [86]. Because of the poor oxygen transfer within the biolm structure, the fouling sub-layers (on the membrane surface) may become anaerobic, and therefore affect membrane fouling differently. Endogenous decay, similar to that expected within the fouling layer, was simulated and revealed the level of carbohydrate in the extracted EPS (eEPSc) to signicantly increase. Since the transition between aerobic to anaerobic conditions seems to produce a large amount of EPS, this phenomenon could also be responsible for MBR fouling [86]. More details about fouling in anaerobic MBR can be found in [1]. The direct impact of air bubbles (as a foulant parameter) on MBR ltration was even investigated by Jang et al. However, it was concluded that the effect of air blocking on the surface can be ignored in MBR processes with high MLSS concentration (as it accounted for less than 1% of total resistance Rt ) [177]. However, under some circumstances, air bubbles my be present or be formed in the lumens of hollow bers and this can be detrimental [178]. Finally, it is important to keep in mind that the aeration rate (discussed in Section 3.3.1) controls biological requirements and parameters such as DO, ammonium/nitrate ratio [179].

3.2.4. Floc characteristics 3.2.4.1. Floc size. In MBR systems, aggregation of the microorganisms, and the formation of large oc is a signicant element in the effective separation of suspended biomass from the treated water, although it is more critical in CASP. In terms of oc size, biomass suspensions in MBRs feature a wide distribution, which ranges signicantly from one study to another. Comparison of the aggregate size distribution of CASP and MBR sludges was carried out [180] and revealed a distinct difference in terms of mean particle sizes (160 and 240 m, respectively). A bimodal distribution was even observed for MBR sludge (520 and 240 m); the high concentration of small colloids, particles and free bacteria was caused by their complete retention by the membrane. In this study, sludges were collected from small-scale experimental rigs and measured by particle size analyzer (MasterSizer 2000). In another study, partial characterization of the oc (up to 100 m) reported the oc size to range from 10 to 40 m, with a mean size of 25 m [143]. These authors also claimed that the oc size distribution obtained with the MBR sludge are lower than the results generally obtained from CASP. In comparison, the particles present in the supernatant, obtained after 4 h of gravitational sedimentation, have a mean size of around 9 m. The oc size distributions obtained with three MBR operated at different SRTs were similar, although the mean oc size increased slightly from 5.2 to 6.6 m for SRT increasing from 20 to 60 days [181]. Given the large size of the oc particles, compared to the pore size of the membrane generally used in MBR, it is expected that oc cannot directly block pore entrances. Nor would the oc deposit on the membrane surface due to drag forces resulting from the low/modest uxes and the shear induced back transport phenomenon experienced by large particles. However, independent of their size, biological oc play a major role in the formation of the fouling cake on the membrane surface. The effect of the EPS level on oc size will be discussed in Section 3.2.5. The addition of aerobic granules, activated carbon or polymer can signicantly increase the oc size. Their effect on MBR fouling is reported in Section 4.2.2. 3.2.4.2. Hydrophobicity/surface charge. In the MBR process, like in CASPs, hydrophobic ocs lead to high occulation propensity and low interaction with the (generally) hydrophilic membrane. However, reports of highly hydrophobic ocs fouling MBR membranes can be found in the literature. Relative hydrophobicity of oc can be directly measured by bacterial adhesion to hydrocarbons (hexane) [182], or estimated by contact angle determination [128]. Although the direct effect of oc hydrophobicity on MBR fouling is difcult to assess, hydrophobicity measurement of sludge and EPS solutions revealed that the decrease of EPS relative hydrophobicity may cause oc deterioration (and consequent increase of Rc ) [175,182]. EPS level and lamentous index (parameter related to the relative presence of lamentous bacteria in sludge) have a direct inuence on the relative hydrophobicity and zeta potential measured in the biomass oc. The excess growth of lamentous bacteria, known to be responsible for severe MBR fouling, also resulted in higher EPS levels, lower zeta potential, more irregular oc shape

Fig. 10. Simplied representation of EPS, eEPS and SMP.

and higher hydrophobicity [183]. In another example, foaming sludge showed greater ux decline (more than 100 times) than the non-foaming sludge. The increase was attributed to the hydrophobic and waxy nature of the foaming sludge surface [184]. Due to the negative charges from ionization of the anionic functional groups, ocs (and EPS) of most activated sludge feature zeta potential and surface charges ranging from 0.2 to 0.6 mequiv./g VSS and from 20 to 30 mV, respectively [185]. The surface charge of MBR microbial oc (obtained by the titration method) conrms the general trend, as it ranged from 0.7 to 0.4 mequiv./g VSS when the MBR was operated at various SRT (from 20 to 60 days) [181]. This was accompanied by an increase in contact angle (from 34 to 44 ). In this study where the fouling resistance caused by microbial oc was found to increase with the SRT, contact angle and surface charge demonstrated a strong positive correlation with the microbialoc-caused fouling propensity. For CASP, Liu and Fang also reviewed a positive effect of long SRT on hydrophobicity and occulation [185]. 3.2.5. Extracellular polymeric substances (EPS) Given the large number of recent publications dealing with the fouling of MBRs by bio-polymeric substances, a major section of this review is focused on their nature, method of determination and inuence on MBR fouling. Extracellular polymeric substances (EPS) are the construction materials for microbial aggregates such as biolms, ocs and activated sludge liquors. The term EPS is used as a general and comprehensive concept for different classes of macromolecules such as polysaccharides, proteins, nucleic acids, (phosphor-)lipids and other polymeric compounds which have been found at or outside the cell surface and in the intercellular space of microbial aggregates [186]. They consist of insoluble materials (sheaths, capsular polymers, condensed gel, loosely

bound polymers and attached organic material) produced by active secretion, shedding of cell surface material or cell lysis [175]. The functions of EPS matrix are multiple and include aggregation of bacterial cells in ocs and biolms, formation of a protective barrier around the bacteria, retention of water and adhesion to surfaces [187]. With its heterogeneous and changing nature, EPS can form a highly hydrated gel matrix in which microbial cells are embedded [188]. Therefore they can be responsible for the creation of a signicant barrier to permeate ow in membrane processes. Finally, bioocs attached to the membrane can play a major nutrient source during the biolm formation on the membrane surface [189]. Their effects on MBR ltration have been reported for more than a decade [190] and have received considerable attention in recent years with many reports indicating the EPS to be the most signicant factor affecting fouling in MBRs [83]. In this review, distinction will be made between the EPS extracted articially from the biological cell oc (eEPS) and the soluble EPS present in the activated sludge supernatant and unassociated with the cell (soluble microbial products or SMP). The term EPS is used as a general parameter to characterize bio-polymeric substances in the reactor (Fig. 10). It is important to recognize that the exact denitions of eEPS, and SMP are directly dependant of the methods used to obtain and characterize chemically these solutions. Studies on the effects of EPS in MBR fouling rely on extraction of EPS from the sludge oc. So far, no standard method of extraction exists, making comparison between research groups difcult. Methods of extraction are numerous and include cation exchange resin [175,191,192], heating methods [193], centrifugation with formaldehyde [194]. These techniques, along with others, have been compared under various conditions to assess their efcacies [185], and results have revealed that formaldehyde extraction was the most effective to extract the largest concentration of eEPS. However, because of its simplicity, the heating method is sometimes preferred (Fig. 11). Typically, the solution containing eEPS is then characterized by its relative content of protein (eEPSp) and carbohydrate (eEPSc), measured by photometric methods (Lowry et al. [195] and Dubois et al. [196] methods, respectively). Although the EPS characterization is sometimes reported in terms of polysaccharides, this review will only use the term carbohydrate, which denition comprises compounds like polysaccharides. While eEPSp has generally a hydrophobic tendency, eEPSc is more hydrophilic [185]. eEPS

Fig. 11. Proposed method for EPS and SMP extractions and measurements.

Table 4 Concentration of the eEPS components in different MBR systems (units are in mg/gSS by default) eEPSp 2530 29 120 31116 20 1146 25 3036 73 60 116101 eEPSc 78 36 40 615 14 1240 9 3328 30 17 2224 Other Humic: 1213 SUVA: 2.83.1 l/m mg TOC: 3765 TOC: 4447 TOC: 42 EPSp + EPSc = 8 TOC: 250 mg/l TOC: 2683 mg/gVSS Details R (10) S S () Four pilot-scale plants, municipal Pilot-scale plant, Industrial Three full-scale plants, municipal Full-scale plants, industrial From 20 to 60 S () R () S, MLSS: 14 g/l From 8 to 80 S (20) References [180] [199]a [200] [165]

[175] [181] [14] [198] [197] [174]

S, synthetic wastewater; R, real wastewater; SRT are given in days in bracket; , innite SRT (i.e. no wastage). a Anaerobic upow-sludge bed lter (UBF) and an aerobic MBR [199].

solution can also be characterized in terms of its TOC level [197,198] and, less frequently, its aromaticity or hydrophobicity (by the measurement of the specic ultraviolet absorbance (SUVA) [199]). Table 4 reports eEPS concentrations from various MBR set-ups and reveals a relatively narrow range of the eEPSp and eEPSc levels measured. In most cases, eEPSp (with a maximum concentration of 120 mg/gSS) was greater than eEPSc (maximum concentration of 40 mg/gSS). Sludge ocs have also been characterized in terms of protein and carbohydrate levels, with colorimetric analysis carried out directly from the washed biomass [174]. Low correlation was found between these two new indicators and MBR fouling propensity. Finally, the measurement of humic substances, generally overlooked for protein and carbohydrate, have revealed their signicant occurrence in activated liquors [185], and may require more attention in future research on MBR fouling. As mentioned before, EPS has been identied as a major fouling parameter [25,184,201203]. More recently, a functional relationship between specic resistance, MLVSS, TMP, permeate viscosity and eEPS was obtained by dimensional analysis [197]. eEPS was found to have no effect on the specic resistance below 20 and above 80 mgEPS/gMLVSS, but played a signicant role on MBR fouling between these two limits. This was conrmed by another study reporting no clear relation between bound EPS (or eEPS) and membrane fouling for concentrations lower than 10 mg/gSS [138]. In another example obtained with an anaerobic MBR, specic resistance increased linearly with eEPS rising from 20 to 130 mg/gSS [171]. In order to reach a better understanding of membrane fouling caused by EPS, further insights into eEPS identication were recently obtained for MBR sludge [175,182]. In a study based on an intermittently aerated MBR, eEPS solution featured three main MW peaks at 100, 500 and 2000 kDa (fractionation conducted by gel chromatography). In this example, eEPS with MW larger than 1000 kDa was assumed to be mainly responsible for MBR fouling [198]. For drinking water treatment, high performance size exclusion chromatography (HPSEC) has been widely used for accurate characterization (or ngerprint-

ing) of the apparent molecular weight distributions of natural organic matters (NOM) [204]. The same technology was applied to compare eEPS solutions obtained from different congurations, operating conditions or treatment plants (Fig. 12). Preliminary results revealed, for example, the similarity of the eEPS proles between different treatment plants [13,30,165]. This conrms previous ndings obtained with CASP sludge. Size exclusion chromatography combined with infrared microspectroscopy techniques was used for CASP sludge [191].

Fig. 12. HPSEC proles of eEPS (a) and SMP (b) from municipal plants operated at different MLSS concentrations [13].

Resulting chromatographs of eEPS solution exhibited seven distinct peaks. The analysis of the complex mixtures of eEPS polymers revealed the presence of 45670 kDa MW proteins, and 0.51 kDa MW carbohydrates. The presence of both proteins and carbohydrates around the biological cells was proposed as a key parameter in the oc formation, and therefore may also have a signicant role in MBR fouling. From this point of view, the work reported for characterization of eEPS obtained from CASP and their inuence of occulation, settling and dewatering may also be an interesting parallel for MBR fouling. Recent reviews [185,205] reporting these issues are valuable tools for future EPSMBR studies. Since the EPS matrix plays a major role in the hydrophobic interactions among microbial cells and thus in the oc formation [185], it was proposed that a decrease in EPS levels may cause oc deterioration. Experimental results obtained during the comparative study of nitrication/denitrication in a MBR tend to conrm this theory [175]. The repercussion of low EPS level, and therefore oc deterioration on membrane fouling may be detrimental for the MBR performances. If veried more thoroughly, this would indicate the existence of an optimum EPS level for which oc structure is maintained without featuring high fouling propensity. Many parameters including gas sparging, substrate composition [171], loading rate [206,207] affect EPS characteristics in the MBR, but SRT probably remains the most signicant of them [208]. A clear decrease of EPS levels was observed for extended SRT, but this reduction became negligible for SRT greater than 30 days [165]. However, Lee et al. [181] observed an increase in protein concentration (along with stable carbohydrate levels) when SRT was increased. Further comments on the role of SRT are given in Section 3.3.2. 3.2.6. Soluble microbial products (SMP) In an attempt to protect the membrane from direct contact with MLSS, a dual compartment MBR (bioreactor coupled with settling tank in which membrane lters biomass supernatant) has been built in Singapore [207]. In this set-up, higher ltration resistance was observed from the membranes ltering
Table 5 Concentration of the SMP components (in mg/l) SMPp 8 0.59a 0.51a 0.5a 23 1034 4.56 SMPc 25 314 26.5 n.d.10a n.d. n.d. 7 533 4.53.7 Other Humic subs: 36 TOC: up to 8 mg/l 437a (TOC) 11a 1.5a TOC: 3070 mg/l DOC: 5 mg/l TOC: 810 mg/l

supernatant rather than those ltering 4 g/l of biomass. This example clearly indicates that the composition and concentration of the organics present in the biomass supernatant (i.e. soluble microbial products (SMP)) have a large impact on membrane performance. SMP are dened as soluble cellular components that are released during cell lysis, diffuse through the cell membrane, are lost during synthesis or are excreted for some purpose [187,209]. In MBR systems, they can also be provided from the feed substrate (Fig. 10). It has been now widely accepted that the concepts of soluble EPS and SMP are identical [152,175,187]. During ltration, SMP adsorb on the membrane surface, block membrane pores and/or form a gel structure on the membrane surface where they provide a possible nutrient source for biolm formation and a hydraulic resistance to permeate ow [152]. Although the inuence of dissolved matter has been studied for a decade, the concept of SMP fouling in the MBR is relatively new as no report on SMP levels existed for MBRs prior to 2001 [83]. In order to reveal the feasibility and relevance of liquid phase analyses on MBR lterability and potentially standardize the method, Rosenberger et al. reported four MBR cases studies based on SMP analysis for membrane fouling [152]. Three methods of separating the water phase from the biomass have been investigated, and the simple ltration through lter paper was found to be the most effective technique over centrifugation and sedimentation [210]. Although these authors used a large pore lter paper (12 m), it is suggested in this review to lter the solution with, at least, a 1.2 m lter in order to remove colloids (Fig. 11). Similarly to eEPS, SMP solution is then characterized with its relative amount of protein and carbohydrate [210], with its TOC level [200] or more rarely with SUVA measurement [211]. Examples of SMPp and SMPc reported in the literature are given in Table 5. HPSEC analysis has also been conducted on SMP solutions (Fig. 12). Not surprisingly, the molecular weight distribution of the organics present in MBR supernatant was found to be signicantly different for reactors operated under various conditions [158]. However, the SMP solution ngerprint was largely unchanged for weekly measurement from the same reactor, indicating no signicant change in SMP characteristics when the

Operating conditions R (10) R (8) R (15) S ( ) Four pilot-scale plants, municipal Three full-scale plants, municipal Full-scale plant, industrial S (), MLSS 15 g/l R (not available) S (20) R (21) R (from 40 to 8) S (20)

References [180] [158] [200] [165] [212] [210] [211] [115] [213] [174]

n.d., non-detected; S, synthetic wastewater; R, real wastewater; SRT are given in days in bracket; , innite SRT (i.e. no wastage). a In mg/gSS.

biomass is acclimatized to given operating conditions. When compared to EPS molecular weight distribution, the SMP solution featured generally larger macromolecules [165]. Comparison between acclimatized sludges obtained from MBR and CASP pilot plants revealed similar levels in terms of eEPSp, eEPSc and eEPShumic [180]. The presence of the membrane in the MBR process does not seem to affect the content of eEPS within the ocs. However, SMPp, SMPc and SMPhumic levels were signicantly greater for the MBR sludge, presumably due to the retention of large macromolecules by the membrane. Critical ux tests carried out under the same conditions for both MBR and CASP sludges revealed the higher fouling propensity of MBR sludge over CASP (critical uxes were around 1015 and 3243 l/m2 h, respectively). Since the measured levels of EPS was unchanged, only the SMP components can be accounted for the higher membrane fouling observed for MBR sludge [180]. During this study, Cabassud and co-workers observed signicant biological activity in the MBR supernatant, indicating the presence of free bacteria (presumably submicron colloidal size), which could also be another cause of membrane fouling. Direct linear relationships between loss of MBR hydraulic performances and SMP concentration has been also reported for an anaerobic MBR [171] and has been mathematically modeled [214]. Different mechanisms for the interaction between the macromolecules present in the supernatant and the membrane surface have also been proposed recently [158]. The creation of the fouling layer on the membrane surface would act as a secondary membrane, increasing the retention and/or the adsorption of macromolecules. The formation of a biolm could also lead to the degradation of the macromolecules as the permeate ows through the membrane. Finally, interaction between the macromolecules and other solutes (humics, divalent cations) within the membrane pores may be responsible for the reduction of the membrane pore size over time. In an attempt to dene the biomass fraction with the highest fouling potential, Lesjean et al. [150] conducted a methodical comparison between permeate and supernatant solutions. Assuming that the materials observed in the biological supernatant and not in the permeate solution are responsible for MBR fouling, this group clearly revealed the higher concentration of carbohydrates, proteins and organic colloids in the MBR supernatant compared to those in the permeate. These ndings conrmed similar results previously reported [30,210]. Since, direct relationships between the carbohydrate level in SMP solution with fouling rate [150], ltration index and CST [213,215,216], critical ux tests [102], and specic ux [152] have been clearly described, this reveals the SMPc as the major foulant indicator in MBR systems. However, the nature and fouling propensity of SMPc were observed to change during the study of unsteady MBR operation [217]. In this specic study, it was not possible to correlate SMPc to fouling. So far, the effect of the protein fraction contained in the SMP solution on MBR fouling has been more rarely reported. Since a signicant amount of proteins is retained by the membrane (from 15% [210] to 90% [217]), it is expected that this plays a role in MBR fouling. This was recently conrmed by the value of specic resistance increasing by a factor of 10 as the SMPp increased from 30 to 100 mg/l

[208]. However, in two separate studies, analyses of the fouling layer have revealed a higher concentration of carbohydrate and lower concentration of proteins compared to their levels in the activated sludge [86,218]. This further conrms the greater distribution of SMPc in the fouling layer compared to that of SMPp. With a smaller MW, humic substances contained in the liquid phase are not retained by the membrane, and therefore may not signicantly participate to MBR fouling [217]. As expected, many operating parameters affect SMP levels in MBRs. As for eEPS, SMP levels decreased with increasing SRT [165]. For SRT ranging from 4 to 22 days, SMPp and SMPc levels were reduced by factors of 3 and 6, respectively [213]. In their long-term study, Rosenberger et al. identied temperature and stress to the microorganisms (and at a lower degree, SRT) to be the main parameters affecting SMPc [158]. In order to obtain better control of the environmental conditions, many research studies are based on the use of synthetic/analogue solutions, which attempt to model real wastewaters. These solutions are sometimes very basic (mainly composed of glucose) and therefore are very easily biodegradable. As a result, it is expected that SMP levels in such systems are lower than in real systems. Since it may be assumed that there are almost no substrate residuals from glucose in the supernatant, the less biodegradable SMP induced by cell lysis or cell release would account for most of the SMP measured in synthetically fed MBRs. This could explain the lower inuence of SMP compared to those of eEPS reported in some MBR studies. When using synthetic substrate, Cho et al. [197] concluded that the membrane fouling was affected more by the bound EPS of activated sludge oc rather than the dissolved organic matter. SUVA measurement carried out from supernatant of MBR fed with synthetic solution conrmed the presence of a portion of larger, more aromatic, more hydrophobic and double-bond-rich organics, which originated from the decayed biomass rather than the feed [211]. Another important study [219], also based on synthetic wastewater, revealed that soluble organics alone cannot predict MBR fouling. By comparing lterabilities of attached and suspended growth microorganisms, Lee and co-workers observed the rate of membrane fouling of the attached growth system (MLSS 0.1 g/l and attached biomass of 2 g/l) to be about seven times higher than that of suspended growth MBR (MLSS of 3 g/l). With similar soluble fraction characteristics in both reactors, the ltration discrepancy was explained by the formation of a protective dynamic membrane created by suspended solids. The results of Ng et al. [207] reported earlier, with more fouling from supernatant that mixed liquor, can also be explained by this mechanism. Another group of organic materials has been recently introduced by Wang et al. [220]. The biopolymer clusters (BPC) have been dened as non-lterable material issued from the afnity clustering of the free EPS and SMP present in the sludge cake deposited on the membrane surface. Although BPC is expected to accumulate on the pore of the sludge cake, this material can be readily separated from the fouling cake by simple stirring. This work highlights the recent interest in the polymeric characterization of the fouling layer [172,221]. For this purpose, confocal laser scanning microscopy can be used

to precisely locate different EPS compounds on the membrane surface [172,176,222,223]. 3.3. Operating conditions 3.3.1. Aeration, crossow velocity Controlling fouling in submerged membrane systems remains more challenging than in pumped crossow or dead-end rigs, in which the feed-liquid can be more accurately managed. Since the origin of the submerged MBR, bubbling has been dened as the strategy of choice to induce ow circulation and shear stress on the membrane surface. Aeration used in MBR systems has three major roles: providing oxygen to the biomass, maintaining the activated sludge in suspension and mitigating fouling by constant scouring of the membrane surface [224]. The use of gas bubbling to enhance membrane processes, and MBRs in particular, has been thoroughly investigated and reviewed [106]. While fundamental studies and mathematical models have been applied to well-dened membrane congurations (such as tubular [225,226] and at sheet [227229]), the effect of bubbles on submerged hollow ber modules is still being assessed. More recently, the uneven distribution of the aeration turbulent shear intensity has been taken into account in the development of a mathematical model [230]. Theoretical analysis becomes even more problematical when the complex nature of biomass mixture (non-Newtonian uid containing solutes, colloids and particulates) present in MBRs is taken into account. However, general anti-fouling phenomena due to aeration occurring in MBRs can be described. Basically, the bubbles owing near to the membrane surface induce local shear transients and liquid ow uctuations, increasing back transport phenomenon. The tangential shear at the membrane surface prevents large particle deposition on the membrane surface. However, the effect of tangential shear is a function of particle diameter, with lower shear induced diffusion and lateral migration velocity for smaller particles, leading to more severe membrane fouling by ne materials [231]. Aeration also affects MBR performance by causing ber lateral movement (or sway) in hollow ber congurations [121]. The effect of bubbling can help to overcome issues related to high packing density in hollow ber bundles. However, it is still a challenging task for MBR designers to achieve effective aeration throughout the population of bers in a bundle [232]. All these effects, described in more detail elsewhere [104,106,233235], contribute to a signicant reduction in fouling propensity. A novel explanation for the inuence of aeration on MBR fouling has been proposed by Ji and Zhou [174]. According to their results, aeration rate directly controls the quantity and composition of the polymeric compounds (EPS) in the biological ocs, and ultimately the ratio of protein/carbohydrate deposited on the membrane surface. However, this mechanism cannot explain the effects of bubble-induced fouling control observed with model feeds [121]. An optimum aeration rate, beyond which a further increase has no signicant effect on fouling suppression, was originally observed by Ueda in 1997 [236], and has been veried in many occasions since [84,237,238]. During MBR optimization stud-

ies, the limits of the aeration were demonstrated through its effect on the netux-ratio (netux over instantaneous ux calculated for operation with membrane relaxation or backwashing). In this case, aeration intensity could not further improve hydraulic performances below the critical netux-ratio of 0.85. Above this value, the aeration rate was able to limit, to a certain extend, the formation of severe fouling [239]. This study revealed that, for specic conditions, the ltration mode has more effect on MBR fouling than the changes in aeration rates. Intense aeration rate may also damage the oc structure reducing their size, and releasing EPS in the bioreactor [174,240]. These phenomena have been similarly described in the sidestream MBR conguration in which the circulation pump is responsible for the break up of bacterial ocs [241,242]. However, in a small crossow cell, fouling was found to decrease linearly with increasing CFV (up to 4.5 m/s), and no CFV optimum was observed [243]. Detailed study of the various calculated hydraulic resistances revealed that CFV values of 2 and 3 m/s were sufcient to prevent the formation of reversible fouling in UF (30 kDa) and MF (0.3 m) systems, respectively. Finally, it was shown that high CFV was more effective in reducing fouling in the MF rather than in the UFMBR system. This was conrmed by analysis of the mass, thickness and density of the fouling layer deposited in/on the membrane under different operating conditions. This may be due to the effect of CFV on critical ux of particulates. At a low CFV of 0.1 m/s, the UF membrane fouled less, being less susceptible to the particle deposition, whereas at a high CFV of 3.5 m/s, the majority of particles would not deposit (raised critical ux) on either membrane. However in current MBRs generation, the CFV is relatively low which could favor the use of tighter (UF) membranes. Determination of the CFV induced by aeration of the membrane surface can be difcult to assess and techniques such as electromagnetic ow velocity meter [244], particle image velocimetry [232], constant temperature anemometry [245,246], have be used for liquid velocity estimation in submerged MBR reactors. Based on the observed CFV in tap water in the riser section (CFVtp ) of a internal loop-airlift MBR system, Liu et al. set up a modied model for calculating the CFV of the aerated activated sludge (CFVas ) over the membrane surface [237]:
.226 CFVas = 1.406 CFV1 0.147 tp

(3)

where is the viscosity of the MLSS, and CFVtp can be estimated from equation given in Liu et al.s earlier publication [247]. 3.3.2. Solid retention time (SRT) SRT (and consequently the F/M ratio), which ultimately controls biomass characteristics, is probably the most important operating parameter impacting on fouling propensity in MBRs. Operating an MBR at higher SRT leads inevitably to increase of MLSS concentration, but this in itself may not necessary lead to greater fouling (Section 3.2.3) [248]. Extremely low SRTs (down to 2 days) have been tested to assess fouling propensity [249]. Not surprisingly, fouling rate increased nearly 10 times when SRT was lowered from

10 to 2 days (corresponding to F/M ratio from to 0.5 to 2.4 gCOD/gMLVSS/day and MLSS of 7.86.9 g/l). There is no reason (other than purely research-based work) to run MBRs at such extreme conditions, and, as a general rule, F/M ratio is recommended to be maintained below 0.5. Other characteristics of performance at SRT as low as 0.25 day are discussed in [250]. The reasons suggested for the increased fouling rate at very low SRT include the increased levels of production of EPS. It should be noted that an increased F/M ratio could occur during transients in unsteady operation (Section 3.3.3). At the other end of the spectrum, the temptation to run MBRs at extended SRT is great, considering the advantages of this process over CASP. Indeed the early MBRs were typically run at very long SRTs to minimize excess sludge. Studies reporting operation at innite SRT are generally bench scale, use synthetic feed (or very nely preltered sewage) [212,251254] and therefore do not take into consideration the accumulation of inert material in the tank. The progressive accumulation in the MBR tank of non-biodegradable materials (like hair and lint), which are not completely removed by the MBR pre-treatment processes, undeniably leads to clogging of the membrane module [255]. The increase in MLSS concentration related to extended SRT could also result in higher fouling propensity (see Section 3.2.3) even with the aeration raised signicantly. Previous experiments revealed an increase of MLSS levels from 7 to 18 g/l and a decrease of F/M ratio from 0.15 to 0.05 kgCOD/kgMLSS/day when the SRT was increased from 30 to 100 days. Even after increasing the aeration rate from 15 to 25 l/min, fouling was nearly twice as great for the longer SRT conditions [155]. In this scenario, the increased shear provided to control fouling could breakup biooc as well as causing cell lysis. Moreover, the increase in aeration intensity to keep the high MLSS levels in suspension and properly oxygenated may not be a sustainable option for the treatment process. During a 300-day operation of a pilot-scale MBR without wastage (innite SRT), the MLVSS increased steadily from 3 to 15 g/l and both removal efciency and membrane performances remained constant. At innite SRT, most of the substrate is consumed to ensure the maintenance needs and the synthesis of storage products. The very low apparent net biomass generation observed can also explain the low fouling propensity observed for high SRT operation in this study [253]. These two studies show that extended-SRT-operation does necessary offer lower fouling; other operating conditions such as ux and aeration rates would also have a major inuence on fouling propensity. The other difculty with very high SRT is the raised viscosity that could attenuate the effect of bubbling. The effects of SRT on biological parameters like MLSS, SMP, eEPS concentrations, described in Section 3.2, also reveal the major impact of this operating parameter on MBR fouling. As a result, selection of the SRT must be considered very carefully in order to optimize MBR operation (see Section 4.2.2). Some studies reported the DOC of the supernatant to be independent of SRT [252]. The lower fouling generally observed at extended SRT is partially explained by the lower organic carbon concentration in eEPS rather than in SMP. Overall, it is likely that there is an optimal SRT, between the high fouling tendency of very low SRT operation and the high viscosity suspension prevalent

for very long SRT. However, the difculty related to properly acclimatize an MBR (pilot) plant to different SRTs and conduct a fair comparison does not allow the determination of an optimum SRT value. This could also explain the discrepancies in the SRT effects reported in the literature. Criteria such as designed rate of sludge and MLSS concentration recommended by the membrane supplier is more prone to dene the working SRT. This point is further discussed in Section 4.2.2. 3.3.3. Unsteady state operation Unsteady state such as variations in operating conditions (ow input/HRT and organic load) and shifts in oxygen supply have also been dened as additional factors leading to changes in MBR fouling propensity. In real-world applications, such unsteady state conditions could occur regularly. In an experiment carried out with a large pilot-scale MBR, the effects of unstable ow input and unintentional sludge wastage have been assessed [217]. Although the level of polysaccharides in the ltrate varied in a chaotic manner, the concentration of this specic compound increased before and after each sludge withdrawal. While the increase after wastage was due to the sudden stress experienced by cells, increase before sludge withdrawal was explained by the increasingly high MLSS concentration and the resulting low DO level in the bioreactor. It was concluded that unsteady operation changed the nature and/or structure (and fouling propensity) of the polysaccharide rather than the overall EPS formation, and therefore could worsen the fouling propensity. These ndings conrmed results previously reported on the effects of transient conditions in feeding patterns. The addition of a spike of acetate in the feed water signicantly decreased the lterability of the biomass in an MBR; this was due to the rise in SMP levels resulting from the feed spike [256]. The effects of starvation conditions on the biological suspension have been assessed by incorporating different substrate impulses in batch tests [257]. Exogenous phases were followed by starvation periods, both characterized by the So/Xo ratio (Eq. (4)). For high So/Xo, multiplication of bacteria cells was observed, while compound storage, characterized by decrease of MLVSS and the absence of SMPp production and bacteria lysis, were obtained at low So/Xo ratio. The low F/M conditions generally used in MBRs are theoretically close to what would be considered as starvation conditions. Although the inuence of these operating conditions on MBR fouling have not been reported, the lower amount of SMPp produced may lead to less severe fouling propensity: So = HRT Xo F M (4)

with So, the feed substrate concentrate and Xo, the MLVSS concentration. The start-up phase can also be considered as unsteady operation and data collected before biomass stabilization (including the period necessary to reach acclimatization) may become relevant in the design of MBRs. Cho et al. [197] reported temporal changes of the bound EPS levels when the MBR was acclimated to three different SRTs (8, 20 and 80 days). As expected (considering the general trends described in

Section 3.2.5), the concentration of eEPS was lower for the longer SRT (8326 mgTOC/gSS for SRT of 8 up to 80 days, respectively). More interestingly, an initial latent phase was observed for which eEPS concentration did not vary signicantly. However, eEPS levels increased exponentially after 40 days of operation at 8-day-SRT, and after 70 days when the MBR was operated with an SRT of 20 days. No change in eEPS was observed during the 80 days of operation at 80-day-SRT. In another MBR set-up operated at innite SRT, no signicant changes in SMP concentration during 100 days of operation were observed, although the MLSS increased from 1.8 to 4.5 g/l in the meantime [258]. After a latent phase of 30 days, MLSS and SMP levels started to signicantly increase and stabilized by reaching a plateau after 140 days of operation at innite SRT (operated for 210 days). In this example, eEPS increased continuously from day 1 to also reach steady-state on day 140 [251]. Nagaoka and Nemoto [198] observed a rise of MLSS concentration from 4 to 14 g/l over 100 days (SRT value not given) along with a relatively constant increase in eEPS (from 50 to 250 mg TOC/l). From these examples and the relationship between eEPS-SMP and fouling propensity dened earlier, further research on the exact impact of unsteady states on MBR fouling is required to implement more sustainable MBR operation. At this point in time, the evidence points to increased membrane fouling during unsteady state (particularly when F/M is increasing). 3.4. Fouling mechanisms in MBRs 3.4.1. Constant TMP operation The current trend in MBR design is to operate at constant ux, and as a result, very few recent studies report the operation of MBR at constant TMP. In the MBR, like other membrane ltration processes at constant TMP, a rapid ux decline is expected to occur during the initial stages of the ltration. The rate of fouling then decreases before reaching a plateau. Bae and Tak recently summarized the hypothetical three-phase-process-mechanisms for initial cake layer formation occurring in MBR [143]. They ltered MBR mixed liquor samples at 100 kPa over periods of up to 5 h with a range of UF membranes. According to these authors, the main parameter affecting the initial fouling (phase 1) would be the irreversible deposition of the soluble fraction of the biomass suspension (presumably SMP). During this phase, the sludge particles and the colloids would not take part in fouling since they are supposed to be, respectively, removed by crossow (size effect) and to be in too low a concentration to have a signicant effect on fouling. Deposition of sludge particles on the membrane surface and in the previously deposited layers is the main phenomenon occurring during phase 2 when the ux declines more slowly. Phase 3 is then dened when ux appears to stabilize, indicating that permeation drag and back transport have reached equilibrium. Although reduced permeation drag limits further severe fouling, compaction of the cake layer would play a signicant role in the slight increase in ltration resistance observed during this last phase. As little fouling still occurs during phase 3, this operation can be maintained during a certain ltration period, before cleaning of the membrane is required. However, in this study, uxes typically dropped to

about 10 l/m2 h which is less than normal constant ux MBR operation. It is important to note that in phase 3, two factors are at work. Firstly, the lowered ow slows down the convection of foulant; it becomes self-limiting. The other factor is that once the ux is low, the fouling resistance is high, relative to the membrane resistance. To see signicant further decline (say by a further 50%) requires the fouling resistance to double. It becomes increasingly difcult to detect fouling trends by simply inspecting the ux decline prole. 3.4.2. Constant ux operation With the constant ux approach, the convection of foulant does not diminish and fouling phenomena self-accelerates and can eventually create a sharp increase of TMP. With fouling rate, and therefore cleaning frequency, increasing with ux, operation conditions favors the MBR to be run at modest uxes to limit fouling severity (Section 4.2.2). As a result, numerous studies have reported the fouling behavior for long-term MBR ltration carried out at sub-critical ux. However, these long-term experiments have revealed noticeable fouling for MBRs operated at sub-critical ux. Since its rst reference to MBRs in 2001 [28], fouling behavior over time is generally characterized by a twostep pattern. During the rst period, a very small TMP rise was observed. For trials carried out over extended time periods, a noticeable change in the rate of TMP increase then arises after some critical time period (Fig. 13). Pollice et al. [32] reviewed the phenomena and two parameters were introduced as indicators for operation under sub-critical conditions: the critical time over which the prolonged rst step is maintained (tcrit ) and the fouling rate (dTMP/dt) during that step. Table 6 reports tcrit and dTMP/dt for recent trials and reveals the long periods of ltration (up to 1200 h) for which fouling rate can be maintained at very low values (down to 2 104 kPa/h). The fouling rates for the high TMP-rise-period have also been reported previously [148]. Prior to these two ltration-steps generally described in the literature, a conditioning period has also been observed [86]. This conditioning period (from now called stage 1) has not been observed or described as such until recently, but may be a key aspect of fouling creation in MBRs. The dynamics of the biomass detachment from the membrane in relation to the

Fig. 13. Long-term ltration for constant ux operation (adapted from [14]).

Table 6 Sub-critical long-term parameters (completed from [32]) Flux (l/m2 h) 17 22 25 30 n.a. 20 8 30 10 8 7 9 18 12 4 6 8 dTMP/dt (kPa/h) 0.005 0.011 0.024 0.072 0.023 0.036 0.036 0.03 0.006 0.004 0.104 0.0002 0.013 0.031 0.6 tcrit (h) >600 1200 300 250 350 600 350 360 550 72 96 240 48 300 192 137 74 References [259] [259] [259] [259] [260] [261] [262] [25] [263] [30] [14] [14] [14] [264] [148] [148] [148]

ltration and aeration turbulence has also been rarely reported, but was recently considered in the formation of a mathematical model [230]. Based on the recent work reported by Zhang et al. [86], a detailed analysis of the mechanisms and factors involved in these three fouling stages follows and is summarized in Fig. 14. 3.4.2.1. Stage 1conditioning fouling. As in constant TMP operation, strong interactions between the membrane surface and the EPS present in the mixed liquor are probably responsi-

ble for the initial stage of fouling during constant ux operation. Ognier et al. [95] described the rapid fouling phenomena inducing irreversible resistance and taking place in the early stage of MBR ltration (in frontal mode, i.e. dead-end operation). Passive adsorption of colloids and organics has been observed even for zero-ux operation, and before any deposition mechanism initiates [86]. Another detailed study based on passive adsorption revealed that the hydraulic resistance due to this process was almost independent of tangential shear. In terms of relative hydraulic resistance contribution, the initial adsorption has been reported to account for 202000% of the clean membrane resistance (mainly depending on the pore size) [263]. In a more recent study, its contribution to the overall resistance was found to become negligible once ltration was conducted [88]. The adsorption propensity (determined with the modied Freundlich isothermal adsorption equation) was also studied in relation to the ltration modes employed in submerged MBRs [265]. As a result, colloid adsorption and initial pore blocking [170] of new or cleaned membranes by organics substances is expected in MBRs. The intensity of this effect depends on membrane pore size distribution and surface chemistry (and especially hydrophobicity) [95] (Sections 3.2.5 and 3.2.6). In a test cell equipped with direct observation through the membrane (DOTM) technology, and with crossow but zero ux, oc was visually observed to temporarily land on the membrane [86]. This was dened as a random interaction process rather than proper cake formation phenomenon. While some ocs were seen to roll and slide across the membrane, biological aggregates typically detached and left a residual footprint of smaller ocs or EPS material. Biomass approaching the membrane surface

Fig. 14. Fouling mechanisms for MBR operated at constant ux (adapted from [86]).

was then able to attach more easily to the membrane, colonize the separation surface and contribute to stage 2. 3.4.2.2. Stage 2slow (steady) fouling. Even though MBRs are operated below the critical ux for the biomass, biooc may randomly land (see above) and contribute to the second fouling stage. After stage 1, the membrane surface is expected to be mostly covered by SMP, leading to the higher attachment propensity of biomass particles and colloids. Because of the low critical ux determined for SMP species, further adsorption and deposition of organics on the membrane surface may also occur during stage 2. Since adsorption may take place not only at the membrane pores but also on the whole surface, biological ocs may initiate cake formation without directly affecting the permeability in this stage. Over time, this phenomenon would worsen. The rate of EPS deposition, and resulting TMP rise, is expected to increase when the operating ux is higher, leading to a shorter stage 2 operation (Table 6). The fouling mechanisms described above would prevail even with a good hydrodynamic environment that provides adequate surface shear over the membrane surface. However as maldistributions of ow, shear or ux are generally expected in MBRs, irregular fouling patterns can be anticipated. 3.4.2.3. Stage 3TMP jump. With regions or pores of the membrane more fouled than others, ux is expected to signicantly decrease in those specic locations. As a result, the overall permeate productivity redistributes to the less fouled membrane areas or pores, for which local ux increases (see Section 2), exceeding a critical ux (dened as sustainable ux in Section 4.2.2). These phenomena have a self-accelerating nature and severe fouling, characterized by an exponential TMP increase, is generally obtained if the ltration is maintained. The sudden rise in TMP or jump is a consequence of constant ux operation and several mechanisms can be postulated for the rapid increase in TMP at a given condition [86]: (i) The inhomogeneous fouling (area loss) model. This model was proposed to explain the observed TMP proles in nominally sub-critical ltration of upow anaerobic sludge [25]. The TMP jump appeared to coincide with a measured loss of local permeability at different positions along the membrane, due to slow fouling by EPS. It was argued that the ux redistribution (to maintain the constant average ux) resulted in regions of supra-critical ux and consequently in rapid fouling and TMP rise. (ii) The inhomogeneous fouling (pore loss) model. Similar TMP transients have been observed for the crossow MF of a model biopolymer (alginate) [17]. These trends revealed that the TMP transient can occur with relatively simple feeds. The data obtained have been explained by a model that involves ux redistribution among open pores, allowing for the pore size distribution. Local pore velocities eventually exceed the critical ux of alginate aggregates that rapidly block the pores. This idea was also the basis of the model proposed by Ognier et al. [31]. While the area loss model

considers macroscopic redistribution of ux, the pore loss model focuses on microscopic scale. In MBR systems, it is expected that both mechanisms occur simultaneously. (iii) The critical suction pressure model. Using a ne colloid, ltered in dead-end mode, onto an immersed hollow ber, gradual TMP rise followed by a rapid increase in TMP was observed. Both autopsy and modeling suggested a critical suction pressure at which coagulation occurs at the base of the cake [266]. The very thin dense layer observed next to the membrane conrmed the rapid increase in resistance leading to the TMP jump. Although this model was obtained with dead-end rather than crossow operation, there is no reason why this mechanism could not apply to sidestream or submerged MBRs. A requirement for that model is that fouling continues to occur over time until the critical suction pressure is reached, and that the deposit compound(s) have the potential to coalesce or collapse. Biolms and deposit layers in MBRs are likely to have this tendency. (iv) Percolation theory. According to percolation theory, the porosity of the fouling layer gradually reduces due to the continuous ltration and material deposition within the deposit layer. At a critical condition, the fouling cake loses connectivity and resistance, and TMP, increase rapidly. This model has been proposed for MBRs [34], but the model indicates a very rapid change (within minutes), which has not been observed in practice. However, it is plausible that the percolation theory approach, combined with the inhomogeneous fouling (area loss) model, could satisfy the more gradual kinetics of the typical TMP transient. Similarly, fractal theory was successfully applied to describe cake microstructure and properties and to explain the cake compression observed during MBR operation [267]. (v) The inhomogeneous ber bundle model. Another manifestation of the TMP transient has been observed for model ber bundles where the ow from individual bers was monitored [118]. The bundle was operated under suction at constant permeate ow, giving constant average ux, and initially this was evenly distributed amongst the bers. However over time, the ows became less evenly distributed so that the standard deviation of the uxes of individual bers started to increase from the initial range of 0.10.15 up to 0.4 l/m2 h. Consequently, the TMP rose to maintain the average ux across the ber bundle, mirroring the increase in the uxes standard deviation. At some point both TMP and standard deviation showed a rapid rise. This is believed to be due to ow maldistribution within the bundle leading to local blockages between bers and membrane fouling. It was possible to obtain more steadily TMP and standard deviation proles when the ow regime around the bers was more vigorous (higher liquid and/or air intensity). Although this trend was observed for a small model bundle, the phenomena are likely to occur in larger bundles. The mechanisms (i)(v) listed above are all self-accelerating and this is a feature of stage 3 fouling. It is probable that more than one of these mechanisms apply simultaneously when an MBR reaches the TMP jump condition.

4. Mitigation of MBR fouling 4.1. Removal of fouling 4.1.1. Physical cleaning Physical cleaning techniques for MBRs include mainly membrane relaxation (where ltration is paused) and membrane backwashing (where permeate is pumped in the reverse direction through the membrane). These techniques have been incorporated in most MBR designs as standard operating strategies to limit fouling; although vigorous backwashing is not an option for at plate submerged membranes. Backwashing (also called backushing) has been found to successfully remove most of the reversible fouling due to pore blocking, transport it back into the bioreactor, and partially dislodge loosely attached sludge cake from the membrane surface. In some cases, clogging near the membrane surface may also be partially loosened or removed by backwashing. The efciency of backwashing has been studied in detail [268270]. Key parameters in the design of backwashing are its frequency, duration, the ratio between those two parameters and its intensity. For example, less frequent, but longer backwashing (600 s ltration/45 s backwashing) was found to be more efcient than more frequent backwashing (200 s ltration/15 s backwashing) [170]. In another study based on factorial design, suction time (between 8 and 16 min) was found to have more effect on fouling removal than both the aeration intensity (0.30.9 m3 /m2 h) and the backwash time (2545 s) [239]. Although more fouling is expected to be removed when backwashing duration and frequency are increased, optimization of backwashing is required in regard to energy and permeate consumptions. This was achieved by the design of a generic control system which automatically optimized the duration of the backwash according to the monitored value of TMP [271]. This anti-fouling operation obviously affects operating costs as energy is required to achieve a pressure suitable for ow reversion. Moreover, between 5 and 30% of the produced permeate is used in the process. Comparison between submerged hollow ber and at sheet MBR revealed the slightly higher overall ux obtained when operating the membrane constantly at low ux [108]. In this example, at sheet membranes, which cannot be backwashed, were operated constantly with ux ranging between 20 and 27 l/m2 h. The hollow ber MBR was operated at higher ux (2333 l/m2 h) but with 25% of the permeate product being recycled for backwashing (45 s of backwashing after every 600 s of operation). Air can also be used as the backushing medium [272]. Although improving the ux by nearly 400% (compared to continuous operation), 15 min of air backwash was required every 15 min of ltration to obtain this result [273]. However, air backwashing is an efcient method for ux recovery, it may also present potential issues of membrane breakage and rewetting. Membrane relaxation (or non-continuous operation of the membrane) signicantly improves membrane productivity. Under relaxation, back transport of foulants is naturally enhanced as non-irreversibly attached foulants can diffuse away from the membrane surface through the concentration gradi-

ent. The fouling removal efciency of this method can be further increased when air scouring is applied during relaxation [149,274]. Detailed studies of the TMP behavior during this type of operation revealed that although the fouling rate is generally higher than for continuous ltration, membrane relaxation allows ltration to be maintained for longer period of times before the need for cleaning [207]. Although some have reported that this type of operation may not be economically feasible for large-scale MBRs [149], further cost and productivity analysis are probably required to compare this method against backwashing. Recent studies assessing alternative strategies for fouling mitigation tend to combine intermittent operation with frequent backwashing for optimum results [137,275]. 4.1.2. Chemical cleaning It is expected that membrane relaxation and backwashing effectiveness tend to decrease with operation time as more irreversible fouling accumulates on the membrane surface. Therefore, in addition to the physical cleaning strategies, different types/intensities of chemical cleaning may also be recommended. They include: Chemically enhanced backwash (on a daily basis), Maintenance cleaning with higher chemical concentration (weekly), and Intensive (or recovery) chemical cleaning (once or twice a year). Maintenance cleaning is used to maintain design permeability and helps to reduce the frequency of intense cleaning. Intensive cleaning is generally carried out when further ltration is no longer sustainable because of an elevated TMP. Each of the four main MBR suppliers (Kubota, Memcor, Mitsubishi and Zenon) proposes their own chemical cleaning recipes, which differ mainly in terms of concentration and methods (Table 7). Under normal conditions, the prevalent cleaning agents remain sodium hypochlorite (for organic foulants) and citric acid (for inorganics). Sodium hypochloride hydrolyzes the organic molecules, and therefore loosen the particles and biolm attached to the membrane. The effects of cleaning chemical agents like NaOCl on microbial community have been also recently studied for modeled MBR processes [276]. It is also common for MBR suppliers to adapt specic protocols for chemical cleanings (i.e. chemical concentrations and cleaning frequencies) for individual facilities [115,179,255]. It also has been mentioned that the level of pollutants (measured as TOC) in the permeate rises just after the chemical cleaning episodes [115]. This is important for MBRs used in reclamation process trains (i.e. upstream of RO for example). So far, no systematic studies on cleaning agents or procedures have been published [277]. This is probably due to the site-specic nature of the MBR fouling. Maintenance cleaning, taking up to 30 min for a complete cycle, is normally carried out every 37 days at a moderate reagent concentration of 0.01 wt.% NaOCl. Recovery cleaning employs rather higher reagent concentrations of 0.20.5 wt.% NaOCl coupled with 0.20.3 wt.% citric acid or 0.51 wt.% oxalic acid (Table 7).

Table 7 Intensive chemical cleaning protocols for four MBR suppliersa (from [255]) Type Mitsubishi Zenon Memcor Kubota CIL CIP CIP CIL Chemicals NaOCl Citric acid NaOCl Citric acid NaOCl Citric acid NaOCl Oxalic acid Concentration (%) 0.3 0.2 0.2 0.20.3 0.01 0.2 0.5 1 Protocols Backow through membrane (2 h) + soaking (2 h) Backpulse and recirculate Recirculate through lumens, mixed liquors and in-tank air manifolds Backow and soaking (2 h)

CIL: cleaning in line where chemical solutions are generally backow (under gravity) inside the membrane. CIP: cleaning in place where membrane tank is isolated and drained; the module is rinsed before being soaked in the cleaning solution and rinsed to remove excess of chlorine. a The exact protocol for chemical cleaning can vary from a plant to another.

Research on the efciency of sonication for removing cake layers in MBRs has also been carried out [96]. The sonication cleaning process is based on the breakdown of the fouling cake into smaller fragments. Although sonication can successfully remove the cake from the membrane surface, this cleaning method was not effective on all types of fouling due to pore blocking and may even worsen this type of fouling. A combination of sonication with backwashing and chemical cleaning appeared to achieve almost complete ux recovery [278]. More details of the cleaning mechanisms are available in [278]. However, sonication would be difcult to apply at a large-scale due to the focused nature of the sonic energy. 4.2. Limitation of fouling It may also be possible to prevent fouling before its occurrence by (1) improving the anti-fouling properties of the membrane, (2) operating the MBR under specic non-or-littlefouling conditions and/or (3) pre-treating the biomass suspension to limit its fouling propensity. 4.2.1. Optimization of membrane characteristics In MBRs, chemical modications of the membrane surface have been shown to efciently improve anti-fouling properties. As mentioned above (Section 3.1.2), more severe fouling is expected when hydrophobic membranes are used in the MBR, and efforts have been focused on increasing membrane hydrophilicity through membrane modication. Recent examples for MBRs comprise NH3 and CO2 plasma treatments of polypropylene hollow bers [128,129]. In both cases, X-ray photoelectron spectroscopy (XPS) and SEM were used to characterize the structural and morphological nature of the modied membrane surface. With the introduction of polar groups (from oxygen and nitrogen) on the membrane surface, membrane hydrophilicity signicantly increased and new membranes presented better ltration performances and ux recovery than those of unmodied membranes. In another study, the addition of TiO2 nanoparticles to the casting solution and a precoat of TiO2 allowed the preparation of two types of TiO2 -immobilized UF membrane (entrapped and deposited, respectively), which were also used in MBR systems [143,279]. As a result, lower

ux decline was obtained with the TiO2 -membranes compared to that of unmodied membranes. As it was possible to add a larger amount of TiO2 particles on as a precoat to a membrane, this lter showed greater fouling mitigation compared to that of the TiO2 -entrappedmembrane. Similarly, when MBR membranes were precoated with ferric hydroxide ocs and compared to an unmodied MBR, both efuent quality and productivity were found to increase [280]. This phenomenon was explained by the adsorption of soluble organics on ferric hydroxide ocs, limiting the direct contact between the organics and the membrane. Finally, fouling phenomena have been used to investigate the creation of self-forming dynamic membrane coupled bioreactors [281]. By using coarse pore-sized substrates and allowing cake and gel layers to deposit on the surface, a self-forming membrane developed with a high ux and good removal efciencies. However, because of the nature of the ltration barrier, the efuent quality cannot be guaranteed, and this is of concern in many applications. 4.2.2. Optimization of operating conditions 4.2.2.1. Aeration. Since the energy involved in providing aeration to the membrane remains a signicant cost factor in MBR design, efforts have been focused on optimization of air owrate. The specic design of airow patterns and location of aerators have also been dened as crucial parameters in fouling mitigation. Recent developments in aeration design carried out by MBR suppliers are often reported in patent format, and involve cyclic aeration systems [282], and improved aerator systems [283,284] for example. A recent study reported a detailed comparison of various aeration devices used in tubular membranes. The results indicated that complex aeration systems with multiple orices injecting air homogeneously in the feed ow featured the highest performances [285]. As mentioned before, the effect of aeration varies from hollow ber to at plate membrane congurations. The presence of a bi-chamber (riser and down-comer) in a Kubota MBR plays a signicant role in inducing high CFV [244]. In the same study, lower uplift resistance and higher CFV were induced by uniformly distributed ne air bubbles (issued from a porous media with 0.5 mm holes) compared to performances obtained with large bubbles (from

2 mm hole diffuser) at similar aeration rates. The air/permeate ratio (m3 /m3 ) can also be a useful parameter to characterize the intensity of aeration required to obtain a given amount of treated water. Values given by MBR suppliers may vary between 24 and 50, depending of the membrane conguration (at sheet versus hollow ber) and the MBR tank design (membrane and aerobic zone combined into one tank or not) [115]. Preliminary work carried out in Singapore on large-scale MBRs revealed these original ratios to be quite conservative, since it was possible to decrease them (down to 56% of their original value) without signicant fouling increase [115]. Attempts to increase the critical ux in submerged MBR by varying the aeration rates has been reported [286]. In order to minimize fouling during high throughput operation, aeration was increased and returned to lower values for the low throughput period. Based on these short-term experiments, it was possible to use this technique to minimize energy consumption. However, a recent study from Choi et al. [88] carried out with a crossow MBR device, indicates the tangential shear to have no effect on ux decline when pseudo steady-state is reached. In other words, increasing CFV does not decrease fouling intensity when the deposition layer starts to govern the permeate ux behavior. In the absence of CFV, ux decline was predominantly caused by reversible fouling, while slightly higher irreversible fouling was detected when CFV was applied [88]. The intermittent operation of aeration has also been reported for (de)nitrication MBR systems [198,287]. In this uncommon scenario of single tank MBR used for both anoxic and aerobic biological degradation, ltration is carried out during the aerobic phase to take advantage of the anti-fouling properties of the air scouring. While some authors testing intermittent aeration do not recommend this type of operation as severe fouling was observed as soon air sparging ceases [170,238], others have reported that efcient fouling control was achieved by intermittent bubbling [288,289]. Pulsing air at a frequency of 1 s on/1 s off allowed an improvement in operating ux ranging from 20 to 100% and was found more efcient than lower frequencies (510 s on/510 s off) more conventionally applied in the industry [288]. However, such system may require the operation of a robust activators and valves at these high frequencies and may not be economically practical. 4.2.2.2. Other operating conditions. As mentioned before (Section 3.3.2), SRT remains probably the main operating parameter dening the characteristics of the biomass suspension and its fouling propensity. With the numerous reports dening the relation between SRT and concentrations of both eEPS and SMP, it appears that the overall performance of the MBR is closely related to the choice of SRT value. Further optimizations of operating conditions through reactor design have been studied and include the addition of a spiral occulator [290], vibrating membranes [291], helical bafes [292], suction mode [97] and high performance compact reactor [293], novel types of air lift [104], porous and exible suspended membrane carriers [294] and the sequencing batch MBR [295] for example. Finally, the membrane module design remains another important parameter in the optimization of the MBR operation, and

more precisely, the use of air sparging. In a specially designed module in which air bubbles were conned in close proximity to the hollow ber (rather than diffusing in the reactor), higher permeability was obtained [296]. 4.2.2.3. Sustainable ux. The energy demand for operation is a potential weakness for the future development of the MBR process. It is recognized that the energy usage of MBRs is still higher than conventional activated sludge systems due to the need to control membrane fouling by different strategies. At the end of the day, MBRs can be economically viable only if it delivers a reasonable ux rate without signicant fouling. Since permeation rate and fouling decrease simultaneously, most MBR systems operate at low uxes to limit rapid and severe membrane fouling. The concept of sustainable ux in MBRs can be dened as the ux for which the TMP increases gradually at an acceptable rate, such that chemical cleaning is not necessary [207]. The rate of TMP increase and the period of ltration before chemical cleaning is required are left to the operators discretion, and therefore a more detailed denition of sustainable ux cannot be possible. While critical ux was mainly determined during short-term experiments, sustainable ux can only be assessed through longer ltration periods. However, sustainable ux can also be dened as sub-critical ux by default. In such a system, not only the ux value is of importance but also the strategies used to maintain this given ux. 4.2.3. Modication of biomass characteristics 4.2.3.1. Coagulant/occulent. Ferric chloride and aluminum sulfate (alum) are two types of coagulant commonly used for water and wastewater treatments. Both have been added to reduce signicantly membrane fouling in MBRs. Once dissolved in water, alum forms hydroxide precipitates which adsorb materials such as suspended particles, colloids and soluble organics. In MBR-based trials, the addition of alum led to a signicant decrease of the SMPc concentration, along with an improvement in membrane hydraulic performances [297]. Because of back transport and shear induced fouling control mechanisms, large microbial ocs are expected to have a lower impact on membrane fouling. The permeability enhancement observed for hybrid coagulant/MBR systems are therefore due to the largest ocs formed. A recent MBR-based example reported that small biological colloids (from 0.1 to 2 m) coagulated and formed larger aggregate when alum was added to MBR activated sludge [298]. Although more expensive, ferric chloride was found to have higher efciency than that of alum. Zeolite has also been used in MBRs and allowed the creation of rigid ocs that have lower specic fouling resistance. Further details about the mechanisms of performance enhancement due to zeolite and alum can be obtained from [298]. The addition of ferric iron has also been tested on an MBR for enhancing the production of ironoxidizing bacteria, responsible for the degradation of gaseous H2 S. In this study, specic ferric precipitate like ferric phosphate and K-jarosite (K-Fe3 (SO4 )2 (OH)6 ) have been observed to foul the membrane [299]. Pre-treatment of the efuent is also possible and studies based on the pre-coagulation/sedimentation of efuent before its introduction in the bioreactor revealed the

fouling limitation offered by this technique. Obviously, pretreatment of the feed is a crucial step in the MBR process. Fundamental of sieving, current state-of-the-art mechanical pretreatment and results from the comparison between different types of sieves are given in [300]. In a recent example, the addition of iron based coagulant controlled both irreversible fouling and suspension viscosity [145]. Ferric hydroxide ocs have also been used in the MBR process as a membrane pre-coating agent. Not only the specic ux of this set up was higher, but the efuent quality was also improved compared to the non-coated MBR system [280]. In this study, additional ferric chloride was added to successfully remove the non-biodegradable organics which accumulated in the bioreactor. This operation also led to a rapid increase in membrane specic ux. 4.2.3.2. Adsorbent agents. Addition of adsorbents into biological treatment systems decreases the level of pollutants, and more particularly organic compounds. When PAC is mixed with the MBR biological suspension, biologically activated carbon forms and is responsible for signicant uptake of soluble organics. During long-term runs, PAC gradually incorporates to the biooc to form some biologically activated carbon [301]. Adsorption of EPS on PAC has been studied during the comparison of sidestream and submerged hybrid PAC-MBRs [302]. For this reason, lower fouling propensity is expected in MBR processes when biomass is mixed with adsorbents. Results conducted with only MBR supernatant also clearly revealed lower fouling when PAC was added (up to 1 g/l) [303]. An optimum PAC concentration of 1.2 g/l was obtained for ltration of activated sludge [262]. In this study, oc size distribution and apparent viscosity of the biomass were the main factors responsible for the lower cake resistance observed when PAC was added to the bioreactor. However, no signicant improvement of ltration was obtained with the addition of 5 g/l of PAC and no sludge wastage [207]. It was postulated that the originally introduced-PAC was quickly saturated with organic pollutants. Only the regular addition of PAC into the bioreactor showed good fouling limitation, as the system was operated at lower SRT. Results reported by Fang et al. [304] conrmed this hypothesis as virgin PAC was responsible for 22% reduction of the ltration resistance, while pre-sorbed PAC only reduced the resistance by 14%. Finally, a detailed mathematical model considering subprocesses like biological reactions in the bulk liquid solution, lm transfer from bulk liquid phase to the biolm, diffusion with biological reaction inside biolm, adsorption equilibrium at the biolmadsorbent interface, and diffusion within the PAC particles has been proposed for predicting performances for hybrid PAC/MBR systems [305]. Numerous other studies reported the use of PAC for MBR fouling limitation, but generally failed to assess key issues such as extra operating cost and disposal of the elevated amount of sludge to be wasted. However, Ng et al. recently assessed more clearly the long-term performances of such hybrid systems [301]. In order to obtain higher biological aggregates in the bioreactor, aerobic granular sludge has also been used in MBR systems [144]. With an average size around 1 mm, granular sludge increased the membrane permeability by 50%, but lower clean-

ing recoveries were observed (88% of those obtained with a conventional MBR). A novel membrane performance enhancer (MPE 50) has been recently developed by Nalco and applied to MBRs. When 1 g/l of cationic polymer-based compound was added directly to the bioreactor, SMPc was found to decrease from 41 to 21 mg/l [306]. The interaction between the polymer and the soluble organics in general, and SMPc in particular, was named as the main mechanism responsible for the performance enhancement when Nalcos polymer was used. In another example, an MBR operated with MLSS as high as 45 g/l featured a lower fouling propensity when 2.2 g/l of polymer was mixed to the bioreactor. Experiments conducted with different system congurations of submerged hollow ber membranes allowed direct comparison of hydraulic performances for pre-occulation and PAC addition. Under the operating conditions used in this study, preocculation presented higher fouling mitigation than that of PAC addition [290]. However, when both strategies were used simultaneously, membrane performances were maximum [290,307]. 5. Conclusions After more than 10 years of intensive research, consensus on the exact fouling phenomena in MBRs has not been reached yet. Originally, it was suspected that aeration rate and MLSS concentration had the main impact on MBR fouling. Notwithstanding their signicant effects, new areas of research have been since developed around the more detailed characterization of these parameters. Efforts now concentrate on optimizing air distribution along the membrane modules and on more precise identication of the biological parameters, which have the most inuence on the membrane performances. With the signicant changes in biomass characteristics from one plant to another, it is not surprising to observe different biomass parameters affecting MBR fouling with various propensities. These disparities are also partly due to the different analytical methods and instruments used in the reported studies. In other words, the quest for a single fouling parameter in MBR seems in vain. A large number of recent publications indicate the biomass supernatant (SMP) and its carbohydrate fraction to be one of the main parameters affecting MBR fouling. However, the more detailed characterization of the supernatant and the fouling layer currently carried out also reveals the signicant role played by the protein fraction. The effect of pore size on membrane fouling is also crucial for MBR design, but the assessment of an optimized membrane pore size is time-dependant. MF-based MBR systems seem to rely on initial fouling and the resulting creation of a dynamic membrane to produce high product quality, while UF-based MBRs feature good rejection from the early stage of ltration. However, this review revealed no clear advantage of using tight membranes over more open pores (within a given ux range). Finally, the ltration time (short-term versus long-term), the mode of operation (constant ux versus constant TMP), the initial stage of the membrane (new versus cleaned), the operating conditions and the cleaning protocol are also crucial elements when the fouling experiments are designed and should be carefully selected, reported and analyzed in view of the results. The critical ux

concept and its determination with the ux-stepping experiment remains an interesting tool to assess fouling propensity for a given operating condition, but cannot be used for long-term ltration predictions. Instead, the concept of sustainable ux, for which ltration can be maintained over an extended period of time, is more appropriate for real MBR plants. Effectiveness and strategies for physical and chemical cleanings are underreported in the open literature, and there are still opportunities to match cleaning protocols with the foulant species present. At this stage in time, it is difcult to propose a short-listing of all the parameters which could predict and/or model MBR fouling. The large number of studies published on the subject and reviewed in Section 3 reveals the complex interactions existing between the different fouling parameters. Further understanding of the nature of MBR foulants and their interactions with the membrane material may provide new directions for cleaning agents and protocols, and fouling mitigation strategies for MBRs. In that effort, previous studies reported for occulation, settling and dewatering of activated sludge can be used as interesting parallels. Acknowledgments The authors gratefully thank the Australian Research Council and the NewSouth Global Postdoctoral Fellowship Program for the nancial support of this study, and Prof Simon Judd, Dr. Yun Ye and Ms. Yulita Marselina for their contribution to this review.

Nomenclature BPC biopolymer clusters BSA bovin serum albumin CASP conventional activated sludge process CFV crossow velocity (m/s) CFVas crossow velocity of activated sludge (m/s) CFVtp crossow velocity of tap water (m/s) COD chemical oxygen demand (mg/l) CST capillarity suction time (s) dTMP/dt rate of TMP increase (or fouling rate) (kPa/h) DO dissolved oxygen (mg/l) DOC dissolved organic carbon (mg/l) DOTM direct observation through membrane eEPSc fraction of carbohydrate contained in extracted solution from sludge (mg/gSS) eEPSp fraction of protein contained in extracted solution from sludge (mg/gSS) EPS extracellular polymeric substances (mg/gSS) F/M food to microorganisms ratio HPSEC high performance size exclusion chromatography HRT hydraulic retention time (h) J ux (l/m2 h) Jc critical ux (l/m2 h) Jc,t critical ux at t (l/m2 h) Jc,20 critical ux at 20 C (l/m2 h) J0 initial ux (l/m2 h)

mc cake load/membrane area (kg/m2 ) MALDI-MS matrix-assisted laser desorption ionization mass spectrometry MBR membrane bioreactor MF microltration MLSS mixed liquor suspended solids (g/l) MLVSS mixed liquor volatile suspended solids (g/l) MW molecular weight MWCO molecular weight cut-off (kDa) NOM natural organic matter PAC powdered activated carbon Rc hydraulic resistance attributed to the cake layer (m1 ) Rcol hydraulic resistance attributed to colloid species (m1 ) Rm hydraulic resistance of the membrane (m1 ) Rp hydraulic resistance attributed to pore blocking (m1 ) Rsol hydraulic resistance attributed to soluble species (m1 ) Rss hydraulic resistance attributed to the suspended solids (m1 ) Rsup hydraulic resistance attributed to the biomass supernatant (m1 ) Rt total hydraulic resistance (m1 ) SMP soluble microbial products (mg/l) SMPc fraction of carbohydrate contained in the sludge solution (mg/gSS) SMPp fraction of protein contained in the sludge solution (mg/gSS) So substrate concentration (g/l) SRT solid retention time (day) SUVA specic ultra violet absorbance (m l mg/C) t temperature ( C) tcrit critical time over which step one is maintained (h) TMP transmembrane pressure (mbar) TOC total organic carbon (mg/l) UF ultraltration UG gas supercial velocity (m/s) UL liquid supercial velocity (m/s) V cumulative volume of permeate Xo MLSS concentration (g/l) Greek symbols c cake specic resistance (m/kg) dynamic viscosity of MLSS (mPa s) References
[1] S. Judd, The MBR Book: Principles and Applications of Membrane Bioreactors in Water and Wastewater Treatment, Elsevier, Oxford, 2006. [2] S. Atkinson, Research studies predict strong growth for MBR markets, Membr. Technol. 2006 (2006) 810. [3] W. Yang, N. Cicek, J. Ilg, State-of-the-art of membrane bioreactors: worldwide research and commercial applications in North America, J. Membr. Sci. 207 (2006) 201211.

[4] C.V. Smith, D. DiGregorio, R.M. Talcott, The use of ultraltration membranes for activated sludge separation, in: Proceedings of the 24th Annual Purdue Industrial Waste Conference, 1969. [5] D. Enegess, A.P. Togna, P.M. Sutton, Membrane separation applications to biosystems for wastewater treatment, Filt. Sep. 40 (2003) 1417. [6] K. Yamamoto, M. Hiasa, T. Mahmood, T. Matsuo, Direct solidliquid separation using hollow ber membrane in an activated-sludge aeration tank, Water Sci. Technol. 21 (1989) 4354. [7] T. Stephenson, S. Judd, B. Jefferson, K. Brindle, Membrane Bioreactors for Wastewater Treatment, IWA Publishing, London, 2000. [8] G. Belfort, R.H. Davis, A.L. Zydney, The behavior of suspensions and macromolecular solutions in crossow microltration, J. Membr. Sci. 96 (1994) 158. [9] R. Chan, V. Chen, Characterization of protein fouling on membranes: opportunities and challenges, J. Membr. Sci. 242 (2004) 169188. [10] A.D. Marshall, P.A. Munro, G. Tragardh, The effect of protein fouling in microltration and ultraltration on permeate ux, protein retention and selectivity: a literature review, Desalination 91 (1993) 65108. [11] A.G. Fane, C.J.D. Fell, A review of fouling and fouling control in ultraltration, Desalination 62 (1987) 117136. [12] R.W. Field, D. Wu, J.A. Howell, B.B. Gupta, Critical ux concept for microltration fouling, J. Membr. Sci. 100 (1995) 259272. [13] B. Jefferson, A. Brookes, P. Le-Clech, S.J. Judd, Methods for understanding organic fouling in MBRs, Water Sci. Technol. 49 (2004) 237244. [14] P. Le-Clech, B. Jefferson, I.S. Chang, S.J. Judd, Critical ux determination by the ux-step method in a submerged membrane bioreactor, J. Membr. Sci. 227 (2003) 8193. [15] H. Li, A.G. Fane, H.G.L. Coster, S. Vigneswaran, Direct observation of particle deposition on the membrane surface during crossow microltration, J. Membr. Sci. 149 (1998) 8397. [16] V. Chen, A.G. Fane, S. Madaeni, I.G. Wenten, Particle deposition during membrane ltration of colloids: transition between concentration polarization and cake formation, J. Membr. Sci. 125 (1997) 109122. [17] Y. Ye, P. Le-Clech, V. Chen, A.G. Fane, Evolution of fouling during crossow ltration of model EPS solutions, J. Membr. Sci. 264 (2005) 190199. [18] B. Espinasse, P. Bacchin, P. Aimar, On an experimental method to measure critical ux in ultraltration, Desalination 146 (2002) 9196. [19] P. Bacchin, P. Aimar, V. Sanchez, Model for colloidal fouling of membranes, AIChE J. 41 (1995) 368376. [20] P. Harmant, P. Aimar, Coagulation of colloids in a boundary layer during cross-ow ltration, Colloid Surf. A: Physiochem. Eng. Asp. 138 (1998) 217230. [21] P. Harmant, P. Aimar, Coagulation of colloids retained by porous wall, AIChE J. 42 (1996) 35233532. [22] A.C.M. Franken, J.T.M. Sluys, V. Chen, A.G. Fane, C.J.D. Fell, Role of protein conformation on membrane characteristics, in: Proceedings of Fifth World Filtration Congress, Nice, France, 1990. [23] S.T. Kelly, W. Senyo Opong, A.L. Zydney, The inuence of protein aggregates on the fouling of microltration membranes during stirred cell ltration, J. Membr. Sci. 80 (1993) 175187. [24] E.H. Bouhabila, R. Ben Aim, H. Buisson, Microltration of activated sludge using submerged membrane with air bubbling (application to wastewater treatment), Desalination 118 (1998) 315322. [25] B.D. Cho, A.G. Fane, Fouling transients in nominally sub-critical ux operation of a membrane bioreactor, J. Membr. Sci. 209 (2002) 391403. [26] L. Defrance, M.Y. Jaffrin, Reversibility of fouling formed in activated sludge ltration, J. Membr. Sci. 157 (1999) 7384. [27] L. Defrance, M.Y. Jaffrin, Comparison between ltrations at xed transmembrane pressure and xed permeate ux: application to a membrane bioreactor used for wastewater treatment, J. Membr. Sci. 152 (1999) 203210. [28] S. Ognier, C. Wisnieswski, A. Grasmick, Biofouling in membrane bioreactors: phenomenon analysis and modelling, in: Proceedings of the MBR 3, Craneld University, UK, 2001. [29] F. Fan, H. Zhou, H. Husain, Identication of wastewater sludge characteristics to predict critical ux for membrane bioreactor processes, Water Res. 40 (2006) 205212.

[30] A. Brookes, B. Jefferson, P. Le-Clech, S. Judd, Fouling of membrane bioreactors during treatment of produced water, in: Proceedings of the IMSTEC, Sydney, Australia, 2003. [31] S. Ognier, C. Wisniewski, A. Grasmick, Membrane bioreactor fouling in sub-critical ltration conditions: a local critical ux concept, J. Membr. Sci. 229 (2004) 171177. [32] A. Pollice, A. Brookes, B. Jefferson, S. Judd, Sub-critical ux fouling in membrane bioreactorsa review of recent literature, Desalination 174 (2005) 221230. [33] Y. Ye, V. Chen, A.G. Fane, Modeling long term sub-critical ltration of model EPS solutions, Desalination 191 (2006) 318327. [34] S.W. Hermanowicz, Membrane ltration of biological solids: a unied framework and its applications to MBR, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. [35] H.P. Grace, Structure and performance of lter media, AIChE J. (1956) 307315. [36] S. Chellam, J.G. Jacangelo, Existence of critical recovery and impacts of operational mode on potable water microltration, J. Environ. Eng. ASCE 124 (1998) 12111219. [37] H. Carrere, F. Blaszkowa, H. Roux de Balmann, Modelling the microltration of lactic acid fermentation broths and comparison of operating modes, Desalination 145 (2002) 201206. [38] H.K. Vyas, R.J. Bennett, A.D. Marshall, Performance of crossow microltration during constant transmembrane pressure and constant ux operations, Intern. Dairy J. 12 (2002) 473479. [39] D.N. Petsev, V.M. Starov, I.B. Ivanov, Concentrated dispersions of charged colloidal particles: sedimentation, ultraltration and diffusion, Colloid Surf. A: Physiochem. Eng. Asp. 81 (1993) 6581. [40] R.M. McDonogh, T. Gruber, N. Stroh, H. Bauser, E. Walitza, H. Chmiel, H. Strathmann, Criteria for fouling layer disengagement during ltration of feeds containing a wide range of solutes, J. Membr. Sci. 73 (1992) 181189. [41] A.G. Fane, C.J.D. Fell, A.G. Waters, Ultraltration of protein solutions through partially permeable membranesthe effect of adsorption and solution environment, J. Membr. Sci. 16 (1983) 211224. [42] M. Hamachi, M. Mietton-Peuchot, Experimental investigations of cake characteristics in crossow microltration, Chem. Eng. Sci. 54 (1999) 40234030. [43] Y. Lee, M.M. Clark, Modeling of ux decline during crossow ultraltration of colloidal suspensions, J. Membr. Sci. 149 (1998) 181202. [44] B. Keskinler, E. Yildiz, E. Erhan, M. Dogru, Y.K. Bayhan, G. Akay, Crossow microltration of low concentration-nonliving yeast suspensions, J. Membr. Sci. 233 (2004) 5969. [45] T. Tanaka, K.-I. Abe, H. Asakawa, H. Yoshida, K. Nakanishi, Filtration characteristics and structure of cake in crossow ltration of bacterial suspension, J. Ferment. Bioeng. 78 (1994) 455461. [46] T. Tanaka, R. Kamimura, R. Fujiwara, K. Nakanishi, Cross-ow ltration of yeast broth cultivated in molasses, Biotechnol. Bioeng. 43 (1994) 10941101. [47] T. Tanaka, K. Usui, K. Nakanishi, Formation of the gel layer of polymers and its effect on the permeation ux in crossow ltration of Corynebacterium glutamicum broth, Sep. Sci. Technol. 33 (1998) 707722. [48] A.A. McCarthy, P.K. Walsh, G. Foley, Characterising the packing and dead-end lter cake compressibility of the polymorphic yeast Kluyveromyces marxianus var. marxianus NRRLy2415, J. Membr. Sci. 198 (2002) 8794. [49] Y.Z. Xujiang, J. Dodds, D. Leclerc, Cake characteristics in cross-ow and dead-end microltration, Filt. Sep. 32 (1995) 795798. [50] M. Mota, J.A. Teixeira, A. Yelshin, Inuence of cell-shape on the cake resistance in dead-end and cross-ow ltrations, Sep. Purif. Technol. 27 (2002) 137144. [51] D. Hughes, R.W. Field, Crossow ltration of washed and unwashed yeast suspensions at constant shear under nominally sub-critical conditions, J. Membr. Sci. 280 (2006) 8998. [52] J.S. Knutsen, R.H. Davis, Deposition of foulant particles during tangential ow ltration, J. Membr. Sci. 271 (2006) 101113.

[53] G. Foley, A review of factors affecting lter cake properties in deadend microltration of microbial suspensions, J. Membr. Sci. 274 (2006) 3846. [54] N. Arora, R.H. Davis, Yeast cake layers as secondary membranes in deadend microltration of bovine serum albumin, J. Membr. Sci. 92 (1994) 247256. [55] C. Guell, P. Czekaj, R.H. Davis, Microltration of protein mixtures and the effects of yeast on membrane fouling, J. Membr. Sci. 155 (1999) 113122. [56] V.T. Kuberkar, R.H. Davis, Effects of added yeast on protein transmission and ux in cross-ow membrane microltration, Biotechnol. Prog. 15 (1999) 472479. [57] Y. Ye, V. Chen, Reversibility of heterogeneous deposits formed from yeast and proteins during microltration, J. Membr. Sci. 265 (2005) 2028. [58] J.M.K. Timmer, H.C. van der Horst, J.P. Labbe, Cross-ow microltration of [beta]-lactoglobulin solutions and the inuence of silicates on the ow resistance, J. Membr. Sci. 136 (1997) 4156. [59] C. Causserand, Y. Kara, P. Aimar, Protein fractionation using selective adsorption on clay surface before ltration, J. Membr. Sci. 186 (2001) 165181. [60] C. Causserand, K. Jover, P. Aimar, M. Meireles, Modication of clay cake permeability by adsorption of protein, J. Membr. Sci. 137 (1997) 3144. [61] C.W. van Oers, M.A.G. Vorstman, P.J.A.M. Kerkhof, Solute rejection in the presence of a deposited layer during ultraltration, J. Membr. Sci. 107 (1995) 173192. [62] S. Panpanit, C. Visvanathan, The role of bentonite addition in UF ux enhancement mechanisms for oil/water emulsion, J. Membr. Sci. 184 (2001) 5968. [63] R.M. McDonogh, K. Welsch, A.G. Fane, C.J.D. Fell, Incorporation of the cake pressure proles in the calculation of the effect of particle charge on the permeability of lter cakes obtained in the ltration of colloids and particulates, J. Membr. Sci. 72 (1992) 197204. [64] G. Foley, A.A. McCarthy, P.K. Walsh, Evidence for shape-dependent deposition in crossow microltration of microbial cells, J. Membr. Sci. 250 (2005) 311313. [65] A.A. McCarthy, D.G. OShea, N.T. Murray, P.K. Walsh, G. Foley, Effect of cell morphology on dead-end ltration of the dimorphic yeast Kluyveromyces marxianus var. marxianus NRRLy2415, Biotechnol. Prog. 14 (1998) 279285. [66] A.A. McCarthy, P. Gilboy, P.K. Walsh, G. Foley, Characterisation of cake compressibility in dead-end microltration of microbial suspensions, Chem. Eng. Commun. 173 (1999) 7990. [67] A.A. McCarthy, P.K. Walsh, G. Foley, Experimental techniques for quantifying the cake mass, the cake and membrane resistances and the specic cake resistance during crossow ltration of microbial suspensions, J. Membr. Sci. 201 (2002) 3145. [68] P.H. Hodgson, G.L. Leslie, A.G. Fane, R.P. Schneider, C.J.D. Fell, K.C. Marshall, Cake resistance and solute rejection in bacterial microltration: the role of the extracellular matrix, J. Membr. Sci. 79 (1993) 3553. [69] K. Ohmori, C.E. Glatz, Effect of carbon source on microltration of Corynebacterium glutamicum, J. Membr. Sci. 171 (2000) 263271. [70] E. Matthiasson, The role of macromolecular adsorption in fouling of ultraltration membranes, J. Membr. Sci. 16 (1983) 2336. [71] J.L. Nilsson, Protein fouling of uf membranes: causes and consequences, J. Membr. Sci. 52 (1990) 121142. [72] S. Metsamuuronen, J. Howell, M. Nystrom, Critical ux in ultraltration of myoglobin and bakers yeast, J. Membr. Sci. 196 (2002) 1325. [73] R. Chan, V. Chen, M.P. Bucknall, Quantitative analysis of membrane fouling by protein mixtures using MALDI-MS, Biotechnol. Bioeng. 85 (2004) 190201. [74] R. Chan, V. Chen, M.P. Bucknall, Ultraltration of protein mixtures: measurement of apparent critical ux, rejection performance, and identication of protein deposition, Desalination 146 (2002) 8390. [75] V. Chen, Performance of partially permeable microltration membranes under low fouling conditions, J. Membr. Sci. 147 (1998) 265278. [76] D. Wu, J.A. Howell, R.W. Field, Critical ux measurement for model colloids, J. Membr. Sci. 152 (1999) 8998.

[77] V. Chen, K.J. Kim, A.G. Fane, Effect of membrane morphology and operation on protein deposition in ultraltration membranes, Biotechnol. Bioeng. 47 (1995) 174180. [78] K.J. Kim, V. Chen, A.G. Fane, Some factors determining protein aggregation during ultraltration, Biotechnol. Bioeng. 42 (1993) 260265. [79] C.-C. Ho, A.L. Zydney, Effect of membrane morphology on the initial rate of protein fouling during microltration, J. Membr. Sci. 155 (1999) 261275. [80] M. Hlavacek, F. Bouchet, Constant owrate blocking laws and an example of their application to dead-end microltration of protein solutions, J. Membr. Sci. 82 (1993) 285295. [81] C.C. Ho, A.L. Zydney, A combined pore blockage and cake ltration model for protein fouling during microltration, J. Colloid Interf. Sci. 232 (2000) 389399. [82] W. Yuan, A. Kocic, A.L. Zydney, Analysis of humic acid fouling during microltration using a pore blockage-cake ltration model, J. Membr. Sci. 198 (2002) 5162. [83] I.S. Chang, P. Le-Clech, B. Jefferson, S. Judd, Membrane fouling in membrane bioreactors for wastewater treatment, J. Environ. Eng. ASCE 128 (2002) 10181029. [84] P. Le-Clech, B. Jefferson, S.J. Judd, Impact of aeration, solids concentration and membrane characteristics on the hydraulic performance of a membrane bioreactor, J. Membr. Sci. 218 (2003) 117129. [85] H. Choi, K. Zhang, D.D. Dionysiou, D.B. Oerther, G.A. Sorial, Effect of activated sludge properties and membrane operation conditions on fouling characteristics in membrane bioreactors, Chemosphere 63 (2006) 16991708. [86] J. Zhang, H.C. Chua, J. Zhou, A.G. Fane, Factors affecting the membrane performance in submerged MBR, J. Membr. Sci., in press. [87] Y. He, P. Xu, C. Li, B. Zhang, High-concentration food wastewater treatment by an anaerobic membrane bioreactor, Water Res. 39 (2005) 41104118. [88] H. Choi, K. Zhang, D.D. Dionysiou, D.B. Oerther, G.A. Sorial, Effect of permeate ux and tangential ow on membrane fouling for wastewater treatment, Sep. Purif. Technol. 45 (2005) 6878. [89] W. Lee, J.-H. Jeon, Y. Cho, K.Y. Chung, B.-R. Min, Behavior of TMP according to membrane pore size, in: Proceedings of the International Congress on Membranes and Membrane Processes (ICOM), Seoul, Korea, 2005. [90] I.S. Chang, M. Gander, B. Jefferson, S.J. Judd, Low-cost membranes for use in a submerged MBR, Proc. Saf. Environ. Protect. 79 (2001) 183188. [91] M.A. Gander, B. Jefferson, S.J. Judd, Membrane bioreactors for use in small wastewater treatment plants: membrane materials and efuent quality, Water Sci. Technol. 41 (2000) 205211. [92] S.S. Madaeni, A.G. Fane, D.E. Wiley, Factors inuencing critical ux in membrane ltration of activated sludge, J. Chem. Technol. Biotechnol. 74 (1999) 539543. [93] K.H. Choo, C.H. Lee, Effect of anaerobic digestion broth composition on membrane permeability, Water Sci. Technol. 34 (1996) 173179. [94] I.-S. Chang, K.-H. Choo, C.-H. Lee, U.-H. Pek, U.-C. Koh, S.-W. Kim, J.H. Koh, Application of ceramic membrane as a pretreatment in anaerobic digestion of alcohol-distillery wastes, J. Membr. Sci. 90 (1994) 131139. [95] S. Ognier, C. Wisniewski, A. Grasmick, Inuence of macromolecule adsorption during ltration of a membrane bioreactor mixed liquor suspension, J. Membr. Sci. 209 (2002) 2737. [96] H.H.P. Fang, X. Shi, Pore fouling of microltration membranes by activated sludge, J. Membr. Sci. 264 (2005) 161166. [97] J. Kim, M. Jang, H. Chio, S. Kim, Characteristics of membrane and module affecting membrane fouling, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. [98] S. Kang, E.M.V. Hoek, H. Choi, H. Shin, Effect of membrane surface properties during the fast evaluation of cell attachment, Sep. Sci. Technol. 41 (2006) 14751487. [99] C.C. Ho, A.L. Zydney, Overview of fouling phenomena and modeling approaches for membrane bioreactors, Sep. Sci. Technol. 41 (2006) 12311251. [100] S. Judd, Fouling control in submerged membrane bioreactors, Water Sci. Technol. 51 (2005) 2734.

[101] A.G. Fane, S. Chang, E. Chardon, Submerged hollow bre membrane moduledesign options and operational considerations, Desalination 146 (2002) 231236. [102] P. Le-Clech, B. Jefferson, S.J. Judd, A comparison of submerged and sidestream tubular membrane bioreactor congurations, Desalination 173 (2005) 113122. [103] B. Gunder, K. Krauth, Replacement of secondary clarication by membrane separationresults with tubular, plate and hollow bre modules, Water Sci. Technol. 40 (1999) 311320. [104] I.S. Chang, S.J. Judd, Air sparging of a submerged MBR for municipal wastewater treatment, Process Biochem. 37 (2002) 915920. [105] P. Le-Clech, H. Alvarez-Vazquez, B. Jefferson, S. Judd, Fluid hydrodynamics in submerged and sidestream membrane bioreactors, Water Sci. Technol. 48 (2003) 113119. [106] Z.F. Cui, S. Chang, A.G. Fane, The use of gas bubbling to enhance membrane processes, J. Membr. Sci. 221 (2003) 135. [107] B. Gunder, K. Krauth, Replacement of secondary clarication by membrane separationresults with plate and hollow bre modules, Water Sci. Technol. 38 (1998) 383393. [108] S. Judd, Submerged membrane bioreactors: at plate or hollow bre? Filt. Sep. 39 (2002) 3031. [109] F.I. Hai, K. Yamamoto, K. Fukushi, Different fouling modes of submerged hollow-ber and at-sheet membranes induced by high strength wastewater with concurrent biofouling, Desalination 180 (2005) 8997. [110] S. Adham, P. Gagliardo, Membrane bioreactors for water repurication phase I, Desalination Research and Development Program Report No. 34, Project No. 1425-97-FC-81-3 0006 J, Bureau of Reclamation, 1998. [111] S. Adham, R. Mirlo, P. Gagliardo, Membrane bioreactors for water reclamationphase II, Desalination Research and Development Program Report No. 60, Project No. 98-FC-81-0031, Bureau of Reclamation, 2000. [112] S. Adham, J.F. DeCarolis, W. Pearce, Optimization of various MBR systems for water reclamationphase III, Desalination and Water Purication Research and Development Program, Final Report No. 103, Agreement No. 01-FC-81-0736, 2004. [113] H.F. van der Roest, A.G.N. van Bentem, D.P. Lawrence, MBR-technology in municipal wastewater treatment: challenging the traditional treatment technologies, Water Sci. Technol. 46 (2002) 273280. [114] MBR-Varsseveld, http://www.mbrvarsseveld.nl/UK/index.htm, accessed: August 2006. [115] G. Tao, K. Kekre, Z. Wei, T.C. Lee, B. Viswanath, H. Seah, Membrane bioreactors for water reclamation, Water Sci. Technol. 51 (2005) 431440. [116] S.Y. Zhang, R. van Houten, D.H. Eikelboom, H. Doddema, Z.C. Jiang, Y. Fan, J.S. Wang, Sewage treatment by a low energy membrane bioreactor, Biores. Technol. 90 (2003) 185192. [117] T. Ueda, K. Hata, Domestic wastewater treatment by a submerged membrane bioreactor with gravitational ltration, Water Res. 33 (1999) 28882892. [118] A.P.S. Yeo, A.W.K. Law, A.G. Fane, Factors affecting the performance of a submerged hollow ber bundle, J. Membr. Sci. 280 (2006) 969982. [119] A. Yeo, A.G. Fane, Performance of individual bers in a submerged hollow ber bundle, Water Sci. Technol. 51 (2005) 165172. [120] S. Vigneswaran, W.G. Shim, D.S. Chaudhary, R. Ben Aim, Mathematical modelling of submerged MBR system used in wastewater treatment, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. [121] F. Wicaksana, A.G. Fane, V. Chen, Fibre movement induced by bubbling using submerged hollow bre membranes, J. Membr. Sci. 271 (2006) 186195. [122] J. Kim, F.A. DiGiano, Dening critical ux in submerged membranes: inuence of length-distributed ux, J. Membr. Sci. 280 (2006) 752761. [123] S. Chang, A.G. Fane, Filtration of biomass with laboratory-scale submerged hollow bre moduleseffect of operating conditions and module conguration, J. Chem. Technol. Biotechnol. 77 (2002) 10301038. [124] S. Chang, A.G. Fane, S. Vigneswaran, Modeling and optimizing submerged hollow ber membrane modules, AIChE J. 48 (2002) 22032212.

[125] F. Lipnizki, R.W. Field, Pervaporation-based hybrid processes in treating phenolic wastewater: technical aspects and cost engineering, Sep. Sci. Technol. 36 (2001) 33113335. [126] X. Zheng, Y.B. Fan, Y.S. Wei, A pilot scale anoxic/oxic membrane bioreactor (A/O MBR) for woolen mill dyeing wastewater treatment, J. Environ. Sci. China 15 (2003) 449455. [127] D. Zhongwei, L. Liying, M. Runyu, Study on the effect of ow maldistribution on the performance of the hollow ber modules used in membrane distillation, J. Membr. Sci. 215 (2003) 1123. [128] H.Y. Yu, M.X. Hu, Z.K. Xu, J.L. Wang, S.Y. Wang, Surface modication of polypropylene microporous membranes to improve their antifouling property in MBR: NH3 plasma treatment, Sep. Purif. Technol. 45 (2005) 815. [129] H.-Y. Yu, Y.-J. Xie, M.-X. Hu, J.-L. Wang, S.-Y. Wang, Z.-K. Xu, Surface modication of polypropylene microporous membrane to improve its antifouling property in MBR: CO2 plasma treatment, J. Membr. Sci. 254 (2005) 219227. [130] I.S. Chang, C.H. Lee, K.H. Ahn, Membrane ltration characteristics in membrane-coupled activated sludge system: the effect of oc structure on membrane fouling, Sep. Sci. Technol. 34 (1999) 17431758. [131] I.-S. Chang, S.-O. Bag, C.-H. Lee, Effects of membrane fouling on solute rejection during membrane ltration of activated sludge, Process Biochem. 36 (2001) 855860. [132] J.A. Scott, D.J. Neilson, W. Liu, P.N. Boon, A dual function membrane bioreactor system for enhanced aerobic remediation of high-strength industrial waste, Water Sci. Technol. 38 (1998) 413420. [133] A. Luonsi, N. Laitinen, K. Beyer, E. Levanen, Y. Poussade, M. Nystrom, Separation of CTMP mill-activated sludge with ceramic membranes, Desalination 146 (2002) 399404. [134] X.-J. Fan, V. Urbain, Y. Qian, J. Manem, Nitrication and mass balance with a membrane bioreactor for municipal wastewater treatment, Water Sci. Technol. 34 (1996) 129136. [135] N. Xu, W.H. Xing, N.P. Xu, J. Shi, Study on ceramic membrane bioreactor with turbulence promoter, Sep. Purif. Technol. 32 (2003) 403 410. [136] S.J. Judd, T. Robinson, J. Holdner, H. Alvarez-Vazquez, B. Jefferson, Impact of membrane material on membrane bioreactor permeability, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. [137] S.T. Zhang, Y.B. Qu, Y.H. Liu, F.L. Yang, X.W. Zhang, K. Furukawa, Y. Yamada, Experimental study of domestic sewage treatment with a metal membrane bioreactor, Desalination 177 (2005) 8393. [138] N. Yamato, K. Kimura, T. Miyoshi, Y. Watanabe, Difference in membrane fouling in membrane bioreactors (MBRs) caused by membrane polymer materials, J. Membr. Sci. 280 (2006) 911919. [139] G.A. Schrader, A. Zwijnenburg, M. Wessling, The effect of WWTP efuent zeta-potential on direct nanoltration performance, J. Membr. Sci. 266 (2005) 8093. [140] W. Fuchs, R. Braun, M. Theiss, Inuence of various wastewater parameters on the fouling capacity during membrane ltration, in: Proceedings of the International Congress on Membranes and Membrane Processes (ICOM), Seoul, Korea, 2005. [141] S. Judd, B. Jefferson, Membranes for Industrial Wastewater Recovery and Re-use, Elsevier, Oxford, 2003. [142] L.S. Tam, T.W. Tang, W.Y. Leung, G.H. Chen, K.R. Sharma, A pilot study on performance of a membrane bio-reactor in treating fresh water sewage and saline sewage in Hong Kong, Sep. Sci. Technol. 41 (2006) 12531264. [143] T.-H. Bae, T.-M. Tak, Interpretation of fouling characteristics of ultraltration membranes during the ltration of membrane bioreactor mixed liquor, J. Membr. Sci. 264 (2005) 151160. [144] X. Li, F. Gao, Z. Hua, G. Du, J. Chen, Treatment of synthetic wastewater by a novel MBR with granular sludge developed for controlling membrane fouling, Sep. Purif. Technol. 46 (2005) 1925. [145] T. Itonaga, K. Kimura, Y. Watanabe, Inuence of suspension viscosity and colloidal particles on permeability of membrane used in membrane bioreactor (MBR), Water Sci. Technol. 50 (2004) 301309.

[146] I.-S. Chang, S.-N. Kim, Wastewater treatment using membrane ltrationeffect of biosolids concentration on cake resistance, Process Biochem. 40 (2005) 13071314. [147] N. Cicek, J.P. Franco, M.T. Suidan, V. Urbain, J. Manem, Characterization and comparison of a membrane bioreactor and a conventional activatedsludge system in the treatment of wastewater containing high-molecularweight compounds, Water Environ. Res. 71 (1999) 6470. [148] A. Brookes, B. Jefferson, G. Guglielmi, S.J. Judd, Sustainable ux fouling in a membrane bioreactor: impact of ux and MLSS, Sep. Sci. Technol. 41 (2006) 12791291. [149] S.P. Hong, T.H. Bae, T.M. Tak, S. Hong, A. Randall, Fouling control in activated sludge submerged hollow ber membrane bioreactors, Desalination 143 (2002) 219228. [150] B. Lesjean, S. Rosenberger, C. Laabs, M. Jekel, R. Gnirss, G. Amy, Correlation between membrane fouling and soluble/colloidal organic substances in membrane bioreactors for municipal wastewater treatment, Water Sci. Technol. 51 (2005) 18. [151] S. Lubbecke, A. Vogelpohl, W. Dewjanin, Wastewater treatment in a biological high-performance system with high biomass concentration, Water Res. 29 (1995) 793802. [152] S. Rosenberger, H. Evenblij, S. te Poele, T. Wintgens, C. Laabs, The importance of liquid phase analyses to understand fouling in membrane assisted activated sludge processes-six case studies of different European research groups, J. Membr. Sci. 263 (2005) 113126. [153] WERF, http://www.werf.org/products/MembraneTool/home/default.asp, accessed: August 2006. [154] C. Bin, W. Xiaochang, W. Enrang, Effects of TMP, MLSS concentration and intermittent membrane permeation on a hybrid submerged MBR fouling, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. [155] S.S. Han, T.H. Bae, G.G. Jang, T.M. Tak, Inuence of sludge retention time on membrane fouling and bioactivities in membrane bioreactor system, Process Biochem. 40 (2005) 23932400. [156] N. Cicek, J.P. Franco, M.T. Suidan, V. Urbain, Using a membrane bioreactor to reclaim wastewater, J. Am. Wat. Works Ass. 90 (1998) 105113. [157] A. Beaubien, M. Baty, F. Jeannot, E. Francoeur, J. Manem, Design and operation of anaerobic membrane bioreactors: development of a ltration testing strategy, J. Membr. Sci. 109 (1996) 173184. [158] S. Rosenberger, C. Laabs, B. Lesjean, R. Gnirss, G. Amy, M. Jekel, J.C. Schrotter, Impact of colloidal and soluble organic material on membrane performance in membrane bioreactors for municipal wastewater treatment, Water Res. 40 (2006) 710720. [159] T. Sato, Y. Ishii, Effects of activated-sludge properties on water ux of ultraltration membrane used for human excrement treatment, Water Sci. Technol. 23 (1991) 16011608. [160] K. Krauth, K.F. Staab, Pressurized bioreactor with membrane ltration for waste-water treatment, Water Res. 27 (1993) 405411. [161] Y. Shimizu, Y.-I. Okuno, K. Uryu, S. Ohtsubo, A. Watanabe, Filtration characteristics of hollow ber microltration membranes used in membrane bioreactor for domestic wastewater treatment, Water Res. 30 (1996) 23852392. [162] J. Cho, K.-G. Song, K.-H. Ahn, The activated sludge and microbial substances inuences on membrane fouling in submerged membrane bioreactor: unstirred batch cell test, Desalination 183 (2005) 425429. [163] N. Ren, Z. Chen, A. Wang, D. Hu, Removal of organic pollutants and analysis of MLSS-COD removal relationship at different HRTs in a submerged membrane bioreactor, Int. Biodeterior. Biodegrad. 55 (2005) 279284. [164] H.M. Wong, C. Shang, G. Chen, Factors affecting virus removal in a membrane bioreactor, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. [165] A. Brookes, S. Judd, E. Reid, E. Germain, S. Smith, H. Alvarez-Vazquez, P. Le-Clech, T. Stephenson, E. Turra, B. Jefferson, Biomass characterisation in membrane bioreactors, in: Proceedings of the IMSTEC, Sydney, Australia, 2003. [166] I.T. Yeom, K.R. Lee, Y.G. Choi, H.S. Kim, Y. Lee, Evaluation of a membrane bioreactor system coupled with sludge pretreatment for aerobic

[167]

[168] [169] [170]

[171]

[172]

[173]

[174]

[175]

[176]

[177]

[178]

[179]

[180]

[181]

[182]

[183]

[184]

[185]

[186]

sludge digestion, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. E. Germain, T. Stephenson, Biomass characteristics, aeration and oxygen transfer in membrane bioreactors: their interrelations explained by a review of aerobic biological processes, Rev. Environ. Sci. Bio/Technol. 4 (2005) 223233. M. Mulder, Basic Principles of Membrane Technology, Kluwer Academic Publishers, Dordrecht, 2000. R. Rautenbach, R. Albrecht, Membrane Processes, John Wiley and Sons, New York, 1989. T. Jiang, M.D. Kennedy, B.F. Guinzbourg, P.A. Vanrolleghem, J.C. Schippers, Optimising the operation of a MBR pilot plant by quantitative analysis of the membrane fouling mechanism, Water Sci. Technol. 51 (2005) 1925. F. Fawehinmi, P. Lens, T. Stephenson, F. Rogalla, B. Jefferson, The inuence of operating conditions on EPS, SMP and bio-fouling in anaerobic MBR, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. H.Y. Kim, K.M. Yeon, C.H. Lee, S. Lee, T. Swaminathan, Biolm structure and extracellular polymeric substances in low and high dissolved oxygen membrane bioreactors, Sep. Sci. Technol. 41 (2006) 12131230. I.-J. Kang, C.-H. Lee, K.-J. Kim, Characteristics of microltration membranes in a membrane coupled sequencing batch reactor system, Water Res. 37 (2003) 11921197. L. Ji, J. Zhou, Inuence of aeration on microbial polymers and membrane fouling in submerged membrane bioreactors, J. Membr. Sci. 276 (2006) 168177. N. Jang, X. Ren, K. Choi, I.S. Kim, Comparison of membrane biofouling in nitrication and denitrication for the membrane bio-reactor (MBR), in: Proceedings of the IWA on Aspire, Singapore, 2005. M.-A. Yun, K.-M. Yeon, J.-S. Park, C.-H. Lee, J. Chun, D.J. Lim, Characterization of biolm structure and its effect on membrane permeability in MBR for dye wastewater treatment, Water Res. 40 (2006) 4552. N.J. Jang, Y.H. Yeo, M.H. Hwang, S. Vigneswaran, J. Cho, I.S. Kim, The effect of dissolved air on the ltration resistance in the hollow ber MBR, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. S. Chang, A. Yeo, A. Fane, M. Cholewa, Y. Ping, H. Moser, Observation of ow characteristics in a hollow ber lumen using non-invasive X-ray microimaging (XMI), J. Membr. Sci., submitted for publication. L.S.D.M. Kox, Membrane bioreactor in Varsseveld, the Dutch approach, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. C. Cabassud, A. Mass e, M. Espinosa-Bouchot, M. Sp erandio, Submerged membrane bioreactors: interactions between membrane ltration and biological activity, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. W. Lee, S. Kang, H. Shin, Sludge characteristics and their contribution to microltration in submerged membrane bioreactors, J. Membr. Sci. 216 (2003) 217227. N.J. Jang, R.S. Trussell, R.P. Merlo, D. Jenkins, S.W. Hermanowicz, I.S. Kim, Exocellular polymeric substances molecular weight distribution and ltration resistance as a function of food to microorganism ratio in the submerged membrane bioreactor, in: Proceedings of the International Congress on Membranes and Membrane Processes (ICOM), Seoul, Korea, 2005. F. Meng, H. Zhang, F. Yang, Y. Li, J. Xiao, X. Zhang, Effect of lamentous bacteria on membrane fouling in submerged membrane bioreactor, J. Membr. Sci. 272 (2006) 161168. I.S. Chang, C.H. Lee, Membrane ltration characteristics in membranecoupled activated sludge systemthe effect of physiological states of activated sludge on membrane fouling, Desalination 120 (1998) 221233. Y. Liu, H.H.P. Fang, Inuences of extracellular polymeric substances (EPS) on occulation, settling, and dewatering of activated sludge, Crit. Rev. Environ. Sci. Technol. 33 (2003) 237273. H.C. Flemming, J. Wingender, Relevance of microbial extracellular polymeric substances (EPSs). Part I. Structural and ecological aspects, Water Sci. Technol. 43 (2001) 18.

[187] C.S. Laspidou, B.E. Rittmann, A unied theory for extracellular polymeric substances, soluble microbial products, and active and inert biomass, Water Res. 36 (2002) 27112720. [188] P.H. Nielson, A. Jahn, Extraction of EPS, in: J. Wingender, T.R. Neu, H.C.E. Flemming (Eds.), Microbial Extracellular Polymeric Substances, Springer-Verlag, Berlin, 1999. [189] H.C. Flemming, G. Schaule, T. Griebe, J. Schmitt, A. Tamachkiarowa, Biofoulingthe Achilles heel of membrane processes, Desalination 113 (1997) 215225. [190] K. Ishiguro, K. Imai, S. Sawada, Effects of biological treatment conditions on permeate ux of UF membrane in a membrane/activated-sludge wastewater treatment system, Desalination 98 (1994) 119126. [191] T. Gorner, P. de Donato, M.-H. Ameil, E. Montarges-Pelletier, B.S. Lartiges, Activated sludge exopolymers: separation and identication using size exclusion chromatography and infrared micro-spectroscopy, Water Res. 37 (2003) 23882393. [192] B. Frolund, R. Palmgren, K. Keiding, P.H. Nielsen, Extraction of extracellular polymers from activated sludge using a cation exchange resin, Water Res. 30 (1996) 17491758. [193] J.W. Morgan, C.F. Forster, L. Evison, A comparative study of the nature of biopolymers extracted from anaerobic and activated sludges, Water Res. 24 (1990) 743750. [194] X.Q. Zhang, P.L. Bishop, B.K. Kinkle, Comparison of extraction methods for quantifying extracellular polymers in biolms, Water Sci. Technol. 39 (1999) 211218. [195] O.H. Lowry, N.J. Rosebourgh, A.R. Farr, R.J. Randall, Protein measurement with the folin phenol reagent, J. Biol. Chem. 193 (1951) 265275. [196] M. Dubois, K.A. Gilles, J.K. Hamilton, P.A. Rebers, P. Smith, Colorimetric method for determination of sugars and related substances, Anal. Chem. 28 (1956) 350356. [197] J.W. Cho, K.G. Song, S.H. Lee, K.H. Ahn, Sequencing anoxic/anaerobic membrane bioreactor (SAM) pilot plant for advanced wastewater treatment, Desalination 178 (2005) 219225. [198] H. Nagaoka, H. Nemoto, Inuence of extracellular polymeric substances on nitrogen removal in an intermittently-aerated membrane bioreactor, Water Sci. Technol. 51 (2005) 151158. [199] Y.T. Ahn, S.T. Kang, S.R. Chae, J.L. Lim, S.H. Lee, H.S. Shin, Effect of internal recycle rate on the high-strength nitrogen wastewater treatment in the combined UBF/MBR system, Water Sci. Technol. 51 (2005) 241247. [200] M. Gao, M. Yang, H. Li, Q. Yang, Y. Zhang, Comparison between a submerged membrane bioreactor and a conventional activated sludge system on treating ammonia-bearing inorganic wastewater, J. Biotechnol. 108 (2004) 265269. [201] H. Nagaoka, S. Ueda, A. Miya, Inuence of bacterial extracellular polymers on the membrane separation activated sludge process, Water Sci. Technol. 34 (1996) 165172. [202] H. Nagaoka, S. Yamanishi, A. Miya, Modeling of biofouling by extracellular polymers in a membrane separation activated sludge system, Water Sci. Technol. 38 (1998) 497504. [203] S. Rosenberger, M. Kraume, Filterability of activated sludge in membrane bioreactors, Desalination 146 (2002) 373379. [204] T.K. Nissinen, I.T. Miettinen, P.J. Martikainen, T. Vartiainen, Molecular size distribution of natural organic matter in raw and drinking waters, Chemosphere 45 (2001) 865873. [205] X. Yin, P.F. Han, X.P. Lu, Y.R. Wang, A review on the dewaterability of bio-sludge and ultrasound pretreatment, Ultrason. Sonochem. 11 (2004) 337348. [206] G.-C. Cha, T.-Y. Jeong, I.-K. Yoo, D.-J. Kim, Kinetics characteristics of SMP and ECP in relation to loading rate in a MBR process, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. [207] C.A. Ng, D. Sun, J. Zhang, H.C. Chua, W. Bing, S. Tay, A. Fane, Strategies to improve the sustainable operation of membrane bioreactors, in: Proceedings of the International Desalination Association Conference, Singapore, 2005. [208] M.E. Hernandez Rojas, R. Van Kaam, S. Schetrite, C. Albasi, Role and variations of supernatant compounds in submerged membrane bioreactor fouling, Desalination 179 (2005) 95107.

[209] H. Li, M. Yang, Y. Zhang, X. Liu, M. Gao, Y. Kamagata, Comparison of nitrication performance and microbial community between submerged membrane bioreactor and conventional activated sludge system, Water Sci. Technol. 51 (2005) 193200. [210] H. Evenblij, J. van der Graaf, Occurrence of EPS in activated sludge from a membrane bioreactor treating municipal wastewater, Water Sci. Technol. 50 (2004) 293300. [211] H.-S. Shin, S.-T. Kang, Characteristics and fates of soluble microbial products in ceramic membrane bioreactor at various sludge retention times, Water Res. 37 (2003) 121127. [212] R. Liu, X. Huang, L. Chen, X. Wen, Y. Qian, Operational performance of a submerged membrane bioreactor for reclamation of bath wastewater, Process Biochem. 40 (2005) 125130. [213] P. Grelier, S. Rosenberger, A. Tazi-Pain, Inuence of sludge retention time on membrane bioreactor hydraulic performance, in: Proceedings of the International Congress on Membranes and Membrane Processes (ICOM), Seoul, Korea, 2005. [214] Y. Lee, J. Cho, Y. Seo, J.W. Lee, K.-H. Ahn, Modeling of submerged membrane bioreactor process for wastewater treatment, Desalination 146 (2002) 451457. [215] H. Evenblij, S. Geilvoet, J. van der Graaf, H.F. van der Roest, Filtration characterisation for assessing MBR performance: three cases compared, Desalination 178 (2005) 115124. [216] K. Tarnacki, S. Lyko, T. Wintgens, T. Melin, F. Natau, Impact of extracellular polymeric substances on the lterability of activated sludge in membrane bioreactors for landll leachate treatment, Desalination 179 (2005) 181190. [217] A. Drews, M. Vocks, V. Iversen, B. Lesjean, M. Kraume, Inuence of unsteady membrane bioreactor operation on EPS formation and ltration resistance, in: Proceedings of the International Congress on Membranes and Membrane Processes (ICOM), Seoul, Korea, 2005. [218] H.P. Chu, X.Y. Li, Membrane fouling in a membrane bioreactor (MBR): sludge cake formation and fouling characteristics, Biotechnol. Bioeng. 90 (2005) 323331. [219] J. Lee, W.-Y. Ahn, C.-H. Lee, Comparison of the ltration characteristics between attached and suspended growth microorganisms in submerged membrane bioreactor, Water Res. 35 (2001) 24352445. [220] X.-M. Wang, X.-Y. Li, X. Huang, Membrane fouling in a submerged membrane bioreactor (SMBR): characterisation of the sludge cake and its high ltration resistance, Sep. Purif. Technol., in press. [221] P. Le-Clech, U. Metzger, R. Stuetz, V. Chen, Modes of ltration in membrane bioreactors and its effects on polymeric fouling, in: Proceedings of the Eurombra Workshop on Biofouling in membrane systems, Trondheim, Norway, 2006. [222] P. Le-Clech, Y. Marselina, R. Stuetz, V. Chen, Non-destructive observation of carbohydrate fouling on microporous membranes, J. Membr. Sci., submitted for publication. [223] M.Y. Chen, D.J. Lee, J.H. Tay, Extracellular polymeric substances in fouling layer, Sep. Sci. Technol. 41 (2006) 14671474. [224] R. Dufresne, R.E. Lebrun, H.C. Lavallee, Comparative study on uxes and performances during papermill wastewater treatment with membrane bioreactor, Can. J. Chem. Eng. 75 (1997) 95103. [225] S.R. Smith, T. Taha, Z.F. Cui, Using an improved 1D boundary layer model with CFD for ux prediction in gas-sparged tubular membrane ultraltration, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. [226] S.J. Judd, P. Le-Clech, T. Taha, Z.F. Cui, Theoretical and experimental representation of a submerged MBR system, in: Proceedings of the MBR 3, Craneld University, UK, 2001. [227] G. Ducom, F.P. Puech, C. Cabassud, Air sparging with at sheet nanoltration: a link between wall shear stresses and ux enhancement, Desalination 145 (2002) 97102. [228] N.V. Ndinisa, A.G. Fane, D.E. Wiley, Fouling control in a submerged at sheet membrane system. Part I. Bubbling and hydrodynamic effects, Sep. Sci. Technol. 41 (2006) 13831409. [229] N.V. Ndinisa, A.G. Fane, D.E. Wiley, D.F. Fletcher, Fouling control in a submerged at sheet membrane system. Part II. Two-Phase ow characterization and CFD simulations, Sep. Sci. Technol. 41 (2006) 14111445.

[230] X.-Y. Li, X.-M. Wang, Modelling of membrane fouling in a submerged membrane bioreactor, J. Membr. Sci. 278 (2006) 151161. [231] K.-H. Choo, C.-H. Lee, Hydrodynamic behavior of anaerobic biosolids during crossow ltration in the membrane anaerobic bioreactor, Water Res. 32 (1998) 33873397. [232] A. Yeo, A.G. Fane, Performance of individual bers in a submerged hollow ber bundle, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. [233] C. Cabassud, S. Laborie, L. Durand-Bourlier, J.M. Laine, Air sparging in ultraltration hollow bers: relationship between ux enhancement, cake characteristics and hydrodynamic parameters, J. Membr. Sci. 181 (2001) 5769. [234] M. Mercier-Bonin, C. Maranges, C. Lafforgue, C. Fonade, A. Line, Hydrodynamics of slug ow applied to cross-ow ltration in narrow tubes, AIChE J. 46 (2000) 476488. [235] P.R. Berube, E. Lei, The effect of hydrodynamic conditions and system congurations on the permeate ux in a submerged hollow ber membrane system, J. Membr. Sci. 271 (2006) 2937. [236] T. Ueda, K. Hata, Y. Kikuoka, O. Seino, Effects of aeration on suction pressure in a submerged membrane bioreactor, Water Res. 31 (1997) 489494. [237] R. Liu, X. Huang, Y.F. Sun, Y. Qian, Hydrodynamic effect on sludge accumulation over membrane surfaces in a submerged membrane bioreactor, Process Biochem. 39 (2003) 157163. [238] C. Psoch, S. Schiewer, Long-term study of an intermittent air sparged MBR for synthetic wastewater treatment, J. Membr. Sci. 260 (2005) 5665. [239] P. Schoeberl, M. Brik, M. Bertoni, R. Braun, W. Fuchs, Optimization of operational parameters for a submerged membrane bioreactor treating dyehouse wastewater, Sep. Purif. Technol. 44 (2005) 61 68. [240] J.S. Park, K.M. Yeon, C.H. Lee, Hydrodynamics and microbial physiology affecting performance of a new MBR, membrane-coupled highperformance compact reactor, Desalination 172 (2005) 181188. [241] E. Tardieu, A. Grasmick, V. Geaugey, J. Manem, Inuence of hydrodynamics on fouling velocity in a recirculated MBR for wastewater treatment, J. Membr. Sci. 156 (1999) 131140. [242] C. Wisniewski, A. Grasmick, Floc size distribution in a membrane bioreactor and consequences for membrane fouling, Colloid Surf. A: Physiochem. Eng. Asp. 138 (1998) 403411. [243] H. Choi, K. Zhang, D.D. Dionysiou, D.B. Oerther, G.A. Sorial, Inuence of cross-ow velocity on membrane performance during ltration of biological suspension, J. Membr. Sci. 248 (2005) 189 199. [244] A. Soa, W.J. Ng, S.L. Ong, Engineering design approaches for minimum fouling in submerged MBR, Desalination 160 (2004) 6774. [245] P. Le-Clech, Z. Cao, P.Y. Wan, D.E. Wiley, A.G. Fane, The application of constant temperature anemometry to membrane processes, J. Membr. Sci., in press. coulement diphasique sur les performances [246] A. Madec, Inuence dun e ` membranes immerg de ltration dun proc ed ea ees, PhD Thesis, Institut National des Sciences Appliqu ees, Toulouse, France, 2000. [247] R. Liu, X. Huang, C.W. Wang, L.J. Chen, Y. Qian, Study on hydraulic characteristics in a submerged membrane bioreactor process, Process Biochem. 36 (2000) 249254. [248] J.S. Zhang, C.H. Chuan, J.T. Zhou, A.G. Fane, Effect of sludge retention time on membrane bio-fouling intensity in a submerged membrane bioreactor, Sep. Sci. Technol. 41 (2006) 13131329. [249] R.S. Trussell, R.P. Merlo, S.W. Hermanowicz, D. Jenkins, The effect of organic loading on process performance and membrane fouling in a submerged membrane bioreactor treating municipal wastewater, Water Res. 40 (2006) 26752683. [250] H.Y. Ng, S.W. Hermanowicz, Membrane bioreactor operation at short solids retention times: performance and biomass characteristics, Water Res. 39 (2005) 981992. [251] M. Gao, M. Yang, H. Li, Y. Wang, F. Pan, Nitrication and sludge characteristics in a submerged membrane bioreactor on synthetic inorganic wastewater, Desalination 170 (2004) 177185.

[252] C. Nuengjamnong, J.H. Kweon, J. Cho, C. Polprasert, K.H. Ahn, Membrane fouling caused by extracellular polymeric substances during microltration processes, Desalination 179 (2005) 117124. [253] J.C. Orantes, C. Wisniewski, M. Heran, A. Grasmick, Inuence of total sludge retension on the performance of a submerge membrane bioreactor, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. [254] Y. Ito, K. Otake, H. Takabatake, K. Dan, M. Henmi, T. Uemura, Challenge to zero excess sludge production in wastewater treatment process using membrane technology, in: Proceedings of the International Congress on Membranes and Membrane Processes (ICOM), Seoul, Korea, 2005. [255] P. Le-Clech, A. Fane, G. Leslie, A. Childress, The operators perspective, Filt. Sep. 42 (2005) 2023. [256] H. Evenblij, B. Verrecht, J. van der Graaf, B. Van der Bruggen, Manipulating lterability of MBR activated sludge by pulsed substrate addition, Desalination 178 (2005) 193201. [257] J. Lobos, C. Wisniewski, M. Heran, A. Grasmick, Effects of starvation conditions on biomass behaviour for minimization of sludge production in membrane bioreactors, Water Sci. Technol. 51 (2005) 3544. [258] P. Jinhua, K. Fukushi, K. Yamamoto, Structure of microbial communities on membrane surface in a submerged MBR, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. [259] X. Wen, Q. Bu, X. Huang, Study on fouling characteristic of a axial hollow bers cross-ow microltration under different ux operations, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. [260] K. Frederickson, N. Cicek, Performance comparison of a pilot-scale MBR and a full-scale sequencing batch reactor with sand ltration: treatment of low strength wastewater from a Northern Canadian Aboriginal community, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. [261] B.-H. Lee, Y.-J. Choi, Effect of HRT and COD load on the nitrogen removal in a MBR with intermittent feeding and aeration, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. [262] Y.Z. Li, Y.L. He, Y.H. Liu, S.C. Yang, G.J. Zhang, Comparison of the ltration characteristics between biological powdered activated carbon sludge and activated sludge in submerged membrane bioreactors, Desalination 174 (2005) 305314. [263] S. Ognier, C. Wisniewski, A. Grasmick, Membrane fouling during constant ux ltration in membrane bioreactors, Membr. Technol. 2002 (2002) 610. [264] S. Rosenberger, U. Kruger, R. Witzig, W. Manz, U. Szewzyk, M. Kraume, Performance of a bioreactor with submerged membranes for aerobic treatment of municipal waste water, Water Res. 36 (2002) 413420. [265] L. Ma, X. Li, G. Du, J. Chen, Z. Shen, Inuence of the ltration modes on colloid adsorption on the membrane in submerged membrane bioreactor, Colloid Surf. A: Physiochem. Eng. Asp. 264 (2005) 120125. [266] S. Chang, T.G. Fane, T.D. Waite, Effect of coagulation within the cakelayer on fouling transitions with dead-end hollow ber membranes, in: Proceedings of the International Congress on Membranes and Membrane Processes (ICOM), Seoul, Korea, 2005. [267] F. Meng, H. Zhang, Y. Li, X. Zhang, F. Yang, Application of fractal permeation model to investigate membrane fouling in membrane bioreactor, J. Membr. Sci. 262 (2005) 107116. [268] E.H. Bouhabila, R. Ben Aim, H. Buisson, Fouling characterisation in membrane bioreactors, Sep. Purif. Technol. 2223 (2001) 123132. [269] C. Psoch, S. Schiewer, Critical ux aspect of air sparging and backushing on membrane bioreactors, Desalination 175 (2005) 6171. [270] C. Psoch, S. Schiewer, Resistance analysis for enhanced wastewater membrane ltration, J. Membr. Sci. 280 (2006) 284297. [271] P.J. Smith, S. Vigneswaran, H.H. Ngo, R. Ben-Aim, H. Nguyen, Design of a generic control system for optimising back ush durations in a submerged membrane hybrid reactor, J. Membr. Sci. 255 (2005) 99106. [272] Y. Sun, X. Huang, F. Chen, X. Wen, A dual functional ltration/aeration membrane bioreactor for domestic wastewater treatment, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004.

[273] C. Visvanathan, B.S. Yang, S. Muttamara, R. Maythanukhraw, Application of air backushing technique in membrane bioreactor, Water Sci. Technol. 36 (1997) 259266. [274] H.C. Chua, T.C. Arnot, J.A. Howell, Controlling fouling in membrane bioreactors operated with a variable throughput, Desalination 149 (2002) 225229. [275] M.V.G. Vallero, G. Lettinga, P.N.L. Lens, High rate sulfate reduction in a submerged anaerobic membrane bioreactor (SAMBaR) at high salinity, J. Membr. Sci. 253 (2005) 217232. [276] B.-R. Lim, K.-H. Ahn, K.-G. Song, C. Jin-Woo, Microbial community in biolm on membrane surface of submerged MBR: effect of in-line cleaning chemical agent, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. [277] B.Q. Liao, D.M. Bagley, H.E. Kraemer, G.G. Leppard, S.N. Liss, A review of biofouling and its control in membrane separation bioreactors, Water Environ. Res. 76 (2004) 425436. [278] A.L. Lim, R. Bai, Membrane fouling and cleaning in microltration of activated sludge wastewater, J. Membr. Sci. 216 (2003) 279290. [279] T.-H. Bae, I.-C. Kim, T.-M. Tak, Preparation and characterization of fouling-resistant TiO2 self-assembled nanocomposite membranes, J. Membr. Sci. 275 (2006) 15. [280] Y. Zhang, D. Bu, C.-G. Liu, X. Luo, P. Gu, Study on retarding membrane fouling by ferric salts dosing in membrane bioreactors, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. [281] Y. Wu, X. Huang, X. Wen, F. Chen, Function of dynamic membrane in self-forming dynamic membrane coupled bioreactor, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. [282] H.R. Rabie, P. Cote, M. Singh, A. Janson, Cyclic aeration system for submerged membrane modules, United States Patent 684,406 (2003). [283] S. Miyashita, K. Honjyo, O. Kato, K. Watari, T. Takashima, M. Itakura, H. Okazaki, I. Kinoshita, N. Inoue, Gas diffuser for aeration vessel of membrane assembly, United States Patent 6,328,886 (2000). [284] P. Cote, Inverted air box aerator and aeration method for immersed membrane, United States Patent 6,863,823 (2002). [285] M. Mayer, R. Braun, W. Fuchs, Comparison of various aeration devices for air sparging in crossow membrane ltration, J. Membr. Sci. 277 (2006) 258269. [286] J.A. Howell, H.C. Chua, T.C. Arnot, In situ manipulation of critical ux in a submerged membrane bioreactor using variable aeration rates, and effects of membrane history, J. Membr. Sci. 242 (2004) 1319. [287] I.T. Yeom, Y.M. Nah, K.H. Ahn, Treatment of household wastewater using an intermittently aerated membrane bioreactor, Desalination 124 (1999) 193203. [288] S. Judd, H. Alvarez-Vazquez, B. Jefferson, The impact of intermittent aeration on the operation of air-lift tubular membrane bioreactors under sub-critical conditions, Sep. Sci. Technol. 41 (2006) 12931302. [289] A.G. Fane, Towards sustainability in membrane processes for water and wastewater processing, in: Proceedings of the International Congress on Membranes and Membrane Processes (ICOM), Seoul, Korea, 2005. [290] W.S. Guo, S. Vigneswaran, H.H. Ngo, A rational approach in controlling membrane fouling problems: pretreatments to a submerged hollow ber membrane system, in: Proceedings of the Water Environment-Membrane Technology Conference, Seoul, Korea, 2004. [291] G. Genkin, T.D. Waite, T.G. Fane, S. Chang, The effet of axial vibratins on the ltration performance of submerged hollow bre membranes, in: Proceedings of the International Congress on Membranes and Membrane Processes (ICOM), Seoul, Korea, 2005. [292] N. Ghaffour, R. Jassim, T. Khir, Flux enhancement by using helical bafes in ultraltration of suspended solids, Desalination 167 (2004) 201 207. [293] K.M. Yeon, J.S. Parka, C.H. Lee, S.M. Kim, Membrane coupled highperformance compact reactor: a new MBR system for advanced wastewater treatment, Water Res. 39 (2005) 19541961. [294] Q. Yang, J. Chen, F. Zhang, Membrane fouling control in a submerged membrane bioreactor with porous, exible suspended carriers, Desalination 189 (2006) 292302.

[295] H.-M. Zhang, J.-N. Xiao, Y.-J. Cheng, L.-F. Liu, X.-W. Zhang, F.-L. Yang, Comparison between a sequencing batch membrane bioreactor and a conventional membrane bioreactor, Process Biochem. (2006). [296] R. Ghosh, Enhancement of membrane permeability by gas-sparging in submerged hollow bre ultraltration of macromolecular solutions: role of module design, J. Membr. Sci. 274 (2006) 7382. [297] R.D. Holbrook, M.J. Higgins, S.N. Murthy, A.D. Fonseca, E.J. Fleischer, G.T. Daigger, T.J. Grizzard, N.G. Love, J.T. Novak, Effect of alum addition on the performance of submerged membranes for wastewater treatment, Water Environ. Res. 76 (2004) 26992702. [298] J.C. Lee, J.S. Kim, I.J. Kang, M.H. Cho, P.K. Park, C.H. Lee, Potential and limitations of alum or zeolite addition to improve the performance of a submerged membrane bioreactor, Water Sci. Technol. 43 (2001) 5966. [299] D. Park, D.S. Lee, J.M. Park, Continuous biological ferrous iron oxidation in a submerged membrane bioreactor, Water Sci. Technol. 51 (2005) 5968. [300] F.-B. Frechen, W. Schier, Mechanical pre-treatment stages of municipal MBR applications in Germany, in: Proceedings of the Membrane Science and Technology, Singapore, 2006. [301] C.A. Ng, D. Sun, A.G. Fane, Operation of membrane bioreactor with powdered activated carbon addition, Sep. Sci. Technol. 41 (2006) 14471466. [302] J.S. Kim, C.H. Lee, Effect of powdered activated carbon on the performance of an aerobic membrane bioreactor: comparison between crossow and submerged membrane systems, Water Environ. Res. 75 (2003) 300307. [303] N. Lesage, M. Sperandio, C. Cabassud, Performances of a hybrid adsorption/submerged membrane biological process for toxic waste removal, Water Sci. Technol. 51 (2005) 173180. [304] H.H.P. Fang, X. Shi, T. Zhang, Effect of activated carbon on fouling of activated sludge ltration, Desalination 189 (2006) 193199. [305] H.H. Tsai, V. Ravindran, M.D. Williams, M. Pirbazari, Forecasting the performance of membrane bioreactor process for groundwater denitrication, J. Environ. Eng. Sci. 3 (2004) 507521. [306] S.H. Yoon, J.H. Collins, D. Musale, S. Sundararajan, S.P. Tsai, G.A. Hallsby, J.F. Kong, J. Koppes, P. Cachia, Effects of ux enhancing polymer on the characteristics of sludge in membrane bioreactor process, Water Sci. Technol. 51 (2005) 151157. [307] J.-H. Cao, B.-K. Zhu, H. Lu, Y.-Y. Xu, Study on polypropylene hollow ber based recirculated membrane bioreactor for treatment of municipal wastewater, Desalination 183 (2005) 431438.

Glossary
Adhesion: Molecular attraction between materials. In membrane separation processes, mechanisms include mechanical (interlocking), chemical (ionic, covalent or hydrogen bonding) and dispersive (adsorption, van der Waals force) adhesions. Biolm: Complex aggregation of microorganisms on a solid surface. This phenomenon is possible thanks to the excretion of the protective and adhesive EPS. Biofouling: Combination of biolm formation and bacterial adhesion and deposition on the membrane surface. Biological oc: Aggregation of bacteria thanks to the EPS action. The relatively large oc size (few hundred micrometers) allows the biomass settlement in CASP. Biomass: Microorganisms and other biological solid materials growing in the bioreactor. Biomass concentration can be characterized by the mixed liquor volatile suspended solids (MLVSS) concentration. Biomass supernatant: Solution generally obtained after centrifugation (and/or ltration) of the activated sludge biomass. The supernatant is composed by colloidal and soluble species. Bioreactor: Tank designed to biologically treat wastewater by the use of biomass. Carbohydrate: Chemical compounds made of hydrogen, carbon and oxygen, such as sugars, starches, and cellulose. Polysaccharides are one type of carbohydrate prevalent in MBRs.

Colloids: Very small, nely divided solids (particles that do not dissolve) that remain dispersed in the aqueous phase due to their small size (from 1 nm to 1 m) and electrical charge. Conventional Activated Sludge Process (CASP): Process in which a clarier follows the aeration tank and is used for solids separation. Deposition: Settling of particles from a solution or suspension mixture on a pre-existing surface. Extracellular polymeric substances (EPS): Construction materials for microbial aggregates such as biolms, ocs and activated sludge liquors (see Section 3.2.5). Extracted extracellular polymeric substances (eEPS): Solution obtained after the physical and/or chemical extraction of the EPS from the biological walls of the microorganisms. Fouling: Undesirable accumulation of (particulate, colloidal, molecular) materials on the internal or external structure of the membrane. If the fouling involves living things such as microorganisms, the term biofouling may be used.

Hydraulic retention time (HRT): The HRT is equivalent to the theoretical detention time for an ideal plug ow or completely mixed reactor and is calculated as the volume of bioreactor divided by the inuent owrate. Membrane bioreactor (MBR): Technology combining biological degradation process by activated sludge with a direct solidliquid separation by ltration. Proteins: Chemical substances based on the polymerisation of amino acids. During biological degradation, proteins are hydrolysed to polypeptides, amino acids, and then ammonia and simple organic compounds. Solids residence time (SRT): The SRT (also called the mean cell residence time or sludge age) is equivalent to the average time that microorganisms spend in the MBR and is calculated as the mass of organisms in the reactor divided by the mass of organisms generated/wasted from the reactor each day. Soluble microbial products (SMP): SMP (also called soluble EPS or biomass supernatant) are dened as soluble cellular components that are released during cell lysis, diffuse through the cell membrane, are lost during synthesis or are excreted. In MBR systems, they can also be provided from the feed substrate (see Section 3.2.6).

Vous aimerez peut-être aussi