Environ. Sci. Technol. 2002, 36, 5245-5251
Characteristics of a Self-Forming Dynamic Membrane Coupled with a Bioreactor for Municipal Wastewater Treatment B I N F A N †,‡ A N D X I A H U A N G * ,† Department of Environmental Science and Engineering, Environment Simulation and Pollution Control State Key Laboratory, Tsinghua University, Beijing 100084, P.R. China, and SKLEAC, Research Center for Eco-environmental Sciences, China Academy of Sciences, Beijing 100085, P.R. China
A self-forming dynamic membrane (SFDM) method that used the biomass layer formed on a coarse mesh to effect solid-liquid separation was proposed. A 100-µm Dacron mesh material was used to make the SFDM modules. A SFDM coupled bioreactor (SFDMBR) was tested to treat actual municipal wastewater, and the performance and mechanisms of the SFDM were investigated. The SFDMBR worked by gravity filtration, and the water head drop was generally 100 000, but the rejection capability declined remarkably with a decrease in MW (14). Compared to the results, the change of the SFDM removal efficiencies of DOC with various MW ranges was not as remarkable as in conventional MBR. Although only 56.81% of dissolved organic substances with MW > 100 000 in terms of DOC was removed from the supernatant, this value being lower than that of the conventional MBR, the removal of those with MW < 3000 could also reached to 33.75%. The DOC rejection mechanisms of the SFDM were probably related to two events. First, the SFDM might reject some large molecules, e.g., MW > 100 000. However, such rejection capability was limited and lower than that of the conventional MF or UF membranes. Second, microorganisms in the cake layer might take some available organic carbon in the effluent while it permeated through the cake layer. The latter was considered the main mechanism to be responsible for the removal of small molecules, especially those with MW < 3000. It was found that MLSS concentration increased by 1000-2000 mg L-1 immediately after the bottom aeration. Therefore, the biomass in the cake layer could not be ignored. However, the SFDM method is different from the biofilm methods, which have been widely used in, for example, biofilter, rotating disks, etc. The biofilm method is to effect biological treatment through attachment of microorganisms on the supporting materials whereas the SFDM method’s main function is to enhance the solid-liquid separation effect through SFDM formation on the porous media although its biological reaction could not be ignored. In the SFDMBR, most biomass was suspended in the bioreactor.
FIGURE 7. Water head drops and duration of the stable stage of the SFDM at different fluxes. (3) Permeation and Stability of the SFDM. Permeation and stability are the most important essentials of membrane filtration. As described before, only the stable stage was the qualified work period of the SFDM in a running cycle. In a certain flux, the water head drop and the duration of the stable stage stands for the permeability and the stability of the SFDM. Figure 7 shows the head drops at 1 h after the start of one running cycle and the duration of the stable stage at different permeation fluxes of the SFDM. It can be seen that the water head drops were lower than 5 cm. For example, when the flux was 14.8 L m-2 h-1, the water head drop was only 5.8 mm. The results proved that the SFDM had an excellent permeability. As shown in Figure 7, the duration of the stable stage was highly dependent on the permeation flux; i.e., the higher the flux, the shorter the duration. If the SFDM modules were clogged too frequently, the management of the SFDMBR would be still onerous although the clogged modules could be cleaned in situ. It is necessary to optimize the permeation flux and the stable duration based on a comprehensive consideration. To set the flux between 15 and 20 L m-2 h-1 and accordingly to let the duration be as long as 4-7 d might be acceptable when the SFDMBR is being used in practical wastewater treatment. (4) Cleaning of the Clogged SFDM Modules. Bottom aeration was found adequate for cleaning out all clogs of the SFDM. It was not necessary for the modules to be disassembled for any physical or chemical cleaning. Additionally, as described before, the SFDM was readily and quickly reformed after the module was physically cleaned through the bottom aeration. This is one of the main advantages of the SFDMBR. To clean and restore the clogged membrane is still a problem for conventional MBR, and it is often costly and onerous work. Mechanisms of the SFDM. (1) Microscopic Structures of the SFDM. The biomass layer of a qualified SFDM could be further divided into two sublayers, a cake layer and a gel layer. The cake layer attached loosely. Once the modules left the bulk liquor, most of the cake layer would slough off even if the modules were picked up quite carefully. The cake layer was mainly composed of visible sludge flocs. Figure 8 is a typical SEM picture of the cake layer, which seems the same as suspended sludge flocs and comprises an abundance of microbes, such as Coccus, Bacillus, chainlike Cocci, filamentous bacteria, epiphytic organisms, etc. The gel layer was seen after the cake layer was sloughed off. It appeared to be composed mainly of gellike matter instead of sludge flocs and to be much thinner than the cake layer. The gel layer adhered tightly to the supporting surface and could hardly be flushed. Figure 9a is a typical picture of the mesh together with the gel layer. It should be noted that the bottom aeration and the sample preparing procedures before SEM analysis might break away some of the gel layer, so only part of the mesh
FIGURE 8. SEM picture of the cake layer. eyes appeared to be covered. Through SEM, the gel layer was seen to be composed mainly of abiotic matter with bodies of microbes scattering on it. Mainly two types of structures were identified, as described below. Figure 9b was a weblike structure of the gel layer. Most of its pore sizes was estimated as 0.1-1 µm being in the range of MF membranes. Some Bacilli and Cocci were seen retained on the weblike structure. By comparing panel b of Figure 9 with panel d, which was an 8000× picture of a 0.4µm hollow fiber membrane (Mitsubishi Rayon Co.) having been used for 10 days, the openings of the weblike structure of the gel layer were much greater than those of the hollow fiber membrane. Additionally, the gel layer was considered mainly composed of the extracellular polymer (ECP). The ECP substances excreted from the cells represent a hydrogel in nature, a layer of which can be considered as permeable (15). These might be the reason the permeability of the SFDM was much better than the conventional membranes. Figure 9c is a hairlike structure of the gel layer. The “hairs” were composed of filamentous bacteria, and some Cocci, Bacilli, and Spirilla scattering among them. Though the pores of this structure were difficult to identify because of the resolving power limitation of the SEM instrument, many of them were estimated to be much less than 0.1 µm, with a few ranging from 0.1 to 0.5 µm. Generally, the weblike structure was prone to appear where the gel layer was thinner, while the hairlike structure appeared where the gel layer was thicker. The hairlike structure might have some rejection capability of large soluble molecules. The gel layer is considered to act the key role in the SFDM rejection capability of the fine particles for the following reasons: (1) the gel layer had structures that can reject fine particles, e.g., those of >1.0 µm; and (2) the cake layer did not have adequate capability to reject fine particles. The difference of SS rejection between the formation and the recovery processes (see in Formation of the SS Rejection Capability of the SFDM) was thought to be caused by whether the gel layer had been formed. On one hand, although the cake layer was quickly formed in the initial period of the formation process, the formation of the gel layer required a much longer time. Therefore, the effluent would remain at a concentration as high as 30-50 mg L-1 for 3-5 days until a gel layer was finally formed. On the other hand, the gel layer adhered so strongly on the supporting surface that most of it could avoid being removed while the cake layer was easily removed by the bottom aerations. The gel layer needed only some repair in the next recovery stage. Therefore, effluent SS concentration could rapidly decline to near zero in the recovery process. VOL. 36, NO. 23, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5249
FIGURE 9. (a) Picture (100×) of the gel layer together with the mesh, (b) 8000× picture of the weblike structure gel layer, (c) 10000× picture of the hairlike structure gel layer, and (d) 8000× picture of a 0.4-µm hollow fiber membrane (Mitsubish Rayon Co.). The rejection effect of the gel layer formed in the conventional membranes has been reported. When studying a cross-flow microfiltration of a suspension of Saccharomyces cerevisiae, Kawakatsu et al. (16) found a dense layer, which was formed at the interface of the membrane surface and the cake layer, able to reject bovine serum albumin and Dextran T70. The supporting membrane surface was observed with the SEM, and it was confirmed that a dense layer with compressed cells and biopolymers surrounding the cell was formed on the surface. However, the T70 rejection was not observed with the cake layer of 0.8-µm poly(methyl methacrylate) particles because no dense layer was formed with the particles. The reported dense layer is considered similar to the gel layer in this research. The gel layer formed on the conventional membranes was beneficial to enhance virus and COD removal of MBR (17). As shown in Figure 9b and c, the pore size distribution of the gel layer was uneven, and some leaks larger than 1 µm still existed. This might be the reason the filtration precision of the SFDM was limited. The gel layer is also one of the chief mechanism of membrane fouling in conventional MBRs. It was considered to be mainly composed of colloidal matter in the mixed liquor (18, 19), including extracellular polymers (ECP) and high-MW matter of soluble microbial products (SMP). This matter tends to adhere on membranes to form a gel layer. The cake layer plays a two-edged role in the SFDM filtration. As will be discussed later, the cake layer contributed most of the filtration resistance to the SFDM but it was 5250
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 23, 2002
indispensable. Besides rejecting most of the coarse flocs and enhancing the effluent quality, the cake layer protected the gel layer against direct collision by the large particles. As inferred from Figure 9b, the thickness of the weblike structure was estimated to be smaller than 0.1 µm, so the strength of the gel layer was doubtful without protection of the cake layer. Actually, if the water head drop was raised too rapidly in the stable filtration stage, e.g., 5 cm in 0.5 h, the effluent would become very turbid and a sudden increase of the permeate flux would happen. The reason is that a sudden increased pressure would break the SFDM. This phenomenon illustrated that the SFDM was weak to some extent. (2) Permeation Resistance and Its Composition of the SFDM. The permeability of the SFDM has been partly discussed in Permeation and Stability of the SFDM. Here a further discussion will be carried out and a method to calculate the permeation resistance of the SFDM defined by eq 1 will be given. In the operation period, the permeation resistance of the SFDM was attributed to the cake layer, the gel layer, and the mesh. The cake layer attached loosely on the module, but the gel layer tightly adhered on the mesh; thus, they can be thought as an integer. Thus, the total permeation resistance of a SFDM that was working in the stable stage can be expresses as follows:
Rtot ) Rcake + Rmesh+gel
(2)
where the subscripts tot, cake, and mesh+gel were respec-
FIGURE 10. Profiles of the experimental and simulated head drops versus permeation flux for the unused and used modules, respectively. (T ) 16 °C). tively the whole module, the cake layer, and the integer of the mesh with the gel layer. Assuming that the viscosity of the SFDMBR effluent was approximately equal to that of pure water, Rmesh+gel can be obtained from the CWP test. Similarly, Rtot can be obtained based on the J and H in the stable stage of the SFDMBR (see Figure 7). However, it is necessary to note that the TMP (see in eq 1) does not equal The flow in the outlet pipe of the SFDM module was proved to be laminar. According to the related hydraulic theorems, the following equation can be deduced:
H)
µR 8A2κ2 J + 2 4 J2 Fg πdg
that formed the gel layer is generally hydrophilic. Therefore, when a gel layer was formed, the surfaces of the used modules were actually those of the gel layer, and the initial head loss disappeared. Another interesting phenomenon was that permeation resistance Rmesh (1.94 × 109 m-1) of the new naked module was little higher than that of the used module Rmesh+gel (1.84 × 109 m-1). This was also caused by the surface hydrophobicity. There were dead water regions around the threads of the new naked mesh because of the remarkable surface hydrophobicity of the Dacron material. Therefore, the effective permeation areas were less than the opening areas of the mesh. When a gel layer was formed, the filtration surfaces became more hydrophilic, and the effective permeation areas were enlarged although the gel matter had a little thickness. Therefore, a used module with the gel layer may show a smaller permeation resistance than an unused module. The total permeation resistance of the SFDM in the stable filtration stage was estimated as Rtot ) 1.56 × 1010 m-1 when MLSS ≈ 7500 mg L-1, J ) 14.8 L m-2 h-1, and H ) 5.8 mm. In conventional submerged MBRs, the total filtration resistances are usually of 12-14 orders of magnitude for keeping a similar stable flux.
Acknowledgments We thank the National Natural Science Foundation of China for the support under Grant 50178039, also to Ms. Yan Yu for her assistance in the experiments.
Literature Cited (3)
where H is the water head drop, F is the density of water, g is the gravitation acceleration, A is the filtration area, κ is a dimensionless coefficient of the total flow head loss, and d is the hydraulic diameter of the outlet pipe. The first term in the righhand side of eq 3 stands for the TMP (expressed using water head), and the second term stands for the sum of the kinetic water head and all the hydrodynamic head losses. Equation 3 suggests that the relationship of H-J follows a quadratic function. Figure 10 shows the profiles of H-J of a module when it was unused and used, respectively, and also the simulated curves using the quadratic function. It can be seen that the experiment data fit very well to the simulated curves. According to the second power term coefficient of the simulated equations, the coefficients of the total flow head loss, κ, were calculated as 2.36 and 2.37, respectively, when the module was unused and used. The two values were approximately equal and fitted well to the actual hydraulic properties of the experimental settings. According to the first power term coefficient of eq 3, the permeation resistance of the new naked module was calculated as Rmesh ) 1.94 × 109 m-1, and the permeation resistance of the used module as Rmesh+gel ) 1.84 × 1010 m-1. There were two interesting phenomena. One was that the H coordinate intercept of the simulated curve of the unused module was not zero as expected in eq 3 whereas the intercept of the used one was nearing zero. Such an intercept meant there was an initial head loss. It is considered that the surface hydrophobicity of the filtration media resulted in the initial head loss since the fresh surface of the Dacron mesh had a remarkable hydrophobicity. However, the colloidal matter
(1) Chiemchaisri, C.; Yamamoto, K.; Vigneswaran, S. Water Sci. Technol. 1993, 27 (1), 171-178. (2) Brindle, K.; Stephenson, T. Biotech Bioeng. 1996, 49, 601-610. (3) Krauth, K. H.; Staab, K. F. Water Res. 1993, 27, 405-411. (4) Marcinkoweky, A. E.; Philips, K. A.; Johnson, H. O.; Shon, J. S. J. Am. Chem. Soc. 1966, 88, 5744-5746. (5) Kuberkar, V. T.; Davis, R. H. J. Membr. Sci. 2000, 168, 243-158. (6) Spencer, G.; Thomas, R. Food Technol. 1991, 45, 98-99. (7) Ohtani, T.; Watanabe, C.; Hoahino, C.; Kimura, S. Int. Chem. Eng. 1987, 27, 195-305. (8) Kishihara, H.; Tamaki, S.; Fuji, S.; Komoto, M. Desalin 1988, 70, 121-126. (9) Jiraratananon, R.; Uttapap, D.; Tangamornsuksun, C. J. Membr. Sci. 1997, 129, 135-143. (10) Chinese NEPA. Water and Wastewater Monitoring Methods, 3rd ed.; Chinese Environmental Science Publishing House: Beijing, China, 1997. (11) Huang, X.; Gui, P.; Qian, Y. Proc. Biochem. 2000, 36, 1001-1006. (12) Zhang, B.; Yamamoto, K.; Ohgaki S.; Kamico, N. Water Sci. Technol. 1997, 35(6), 37-44. (13) Belfort, G.; Davis, R. H.; Zydney, A. L. J. Membr. Sci. 1994, 96, 1-58. (14) Rui, L. Ph.D. Dissertation, Tsinghua University, Beijing, China, 2000. (15) McDonogh, R.; Schaule, G.; Flemming, H.-C. J. Membr Sci. 1994, 87, 199-217. (16) Kawakatsu, T.; Nakao, S.; Kimura, S. J. Chem. Eng. Jpn. 1993, 26, 656-661. (17) Chiemchaisri, C.; Wong, Y. K.; Urase, T.; Yamamoto, K. Water Sci. Technol. 1992, 25, 231-240. (18) Czekaj, P.; Lopez, F.; Guell, C. J. Membr. Sci. 2000, 166, 199212. (19) Lee, J.; Ahn, W.-Y.; Lee, C.-H. Water Res. 2001, 35, 2435-2445.
Received for review May 15, 2002. Revised manuscript received September 20, 2002. Accepted September 25, 2002. ES025789N
VOL. 36, NO. 23, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5251
