Characteristics and Formation Mechanism of Membrane Fouling in a Full- Scale RO Wastewater Reclamation Process: Membrane Autopsy and Fouling T Characterization

Journal of Membrane Science 563 (2018) 843–856

Contents lists available at ScienceDirect

Journal of Membrane Science

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

Characteristics and formation mechanism of membrane fouling in a full- scale RO wastewater reclamation process: Membrane autopsy and fouling T characterization

Libing Zhenga,b,c, Dawei Yua,c, Gang Wangd, Zenggang Yued, Chun Zhangc,e, Yawei Wanga,c, ⁎ Junya Zhanga,c, Jun Wanga,b, Guoliang Liangf, Yuansong Weia,b,c, a State Key Joint Laboratory of Environmental Simulation and Pollution Control, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing 100085, China b University of Chinese Academy of Sciences, Beijing 100049, China c Department of Water Pollution Control Technology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing 100085, China d Huaneng Jiaxiang Power Generation Co.Ltd, Shangdong 272400, China e College of Chemical Engineering, Beijing University of Chemical Technology, Beijing 100029, China f Huaneng Jining Power Generation Co.Ltd, Shangdong 272000, China

ARTICLE INFO ABSTRACT

Keywords: Membrane fouling is the bottleneck of stable operations for reverse osmosis (RO), which is the key technology Autopsy for reclaimed water reuse in thermal power plants. The foulant composition, formation mechanism, and key Membrane fouling contributors were analyzed in this study. The primary scaling substances are Ca, Mg, Al, Fe, and Si, and Ca RO accounts for 1.49 wt%. Humic substances, proteins, and polysaccharides are the primary organic constitutes, and Bio-fouling among them bio-derivatives were main composition revealed by FTIR, EEM, and SEC. Thus, bio-fouling was Vertical distribution believed to be the key contributor together with the results of the SEM and the microbial community analysis. The three leading bacteria were α-, β-, γ-proteobacteria at class levels, 38.11%, 14.19%, and 34.31%, respectively, while starkeya, acidovorax, luteimonas, and pseudoxanthomonas were the leading bacteria at the genus level. Among these bacteria, the one with metabolic processes related to nitrogen fixation and proteolysis showed higher abundance for the high concentrations of nitrogen and protein. Foulants on the endcap and membrane entrance also indicated severe bio-fouling, where Acidovorax presented significant abundance. A Vertical distribution of the microbial community was found in the cross-section of the foulant, in particular, a significant decrease of starkeya (from 48.4% to 4.65%) and acidovorax (from 17.64% to 5.61%) and an increase of pseudoxanthomonas (from 1.04% to 12.89%) from the top layer to the bottom layer were observed. The γ- proteobacteria was recognized as the pioneering bacteria for its significantly higher abundance in the deeper layer. This study helps to elucidate the RO membrane fouling composition and its key contributors, and improves our understanding of the membrane fouling mechanism and a controlling strategy.

1. Introduction remain [4–6], although numerous wastewater treatment plants have updated to tertiary treatment to meet the needs of reclaimed water Thermal power plants (TPPs) are heavy water consumers in which reuse. reclaimed water reuse had found a great opportunity for application Advanced treatment is therefore needed for reclaimed water appli- [1–3]. Approximately 90% of the TPP water consumption was used as cations in TPPs, and reverse osmosis (RO) has become the first option recirculated cooling water and boiler feedwater; the application of re- with an increasing demand of high quality reclaimed water [7–9]. claimed water is feasible in TPPs [1]. Thus, reclaimed water was However, membrane fouling is the most notable obstruction, the key strongly recommended in TPPs all over the world, and it was manda- concerns of the pretreatment processes were the removal of suspended tory to use reclaimed water for new TPPs in certain parts of China, such particulates, colloids, and dissolved matters, preventing bacterial as Shangdong Province. However, scaling and corrosion were the two growth, and inhibiting scaling formation [5,10]. Coagulation, adsorp- biggest problems for reclaimed water reuse, as certain contaminants tion, and microfiltration (MF)/ultrafiltration (UF) were widely used to

⁎ Correspondence to: 18 Shuangqing Road, Haidian District, Beijing 100085, PR China. E-mail address: [email protected] (Y. Wei). https://doi.org/10.1016/j.memsci.2018.06.043 Received 9 April 2018; Received in revised form 20 June 2018; Accepted 21 June 2018 Available online 26 June 2018 0376-7388/ © 2018 Elsevier B.V. All rights reserved. L. Zheng et al. Journal of Membrane Science 563 (2018) 843–856 alleviate the membrane fouling, as well as the dosage of antiscalants [28,33–35]. Tan et al. [29] reported that the use of non-oxidizing and bactericides and the application of online and offline chemical biocides can significantly increasing the abundance of pseudox- cleaning [5,11]. However, RO membrane fouling is inevitable because anthomonas, which belonged to γ-proteobacteria and was believed to be the pretreatment cannot completely remove all components, especially more resistant to non-oxidizing biocides. bacteria. Colloidal precipitation, inorganic scaling, organic fouling, and However, membrane fouling always appeared in a combined pat- bio-fouling were four typical fouling types in municipal wastewater tern, where the four types of membrane fouling cannot be clearly de- reclamation (MWR) by RO [12]. marcated [14]. Therefore, understanding their composition, structure Surface and bulk crystallization were two main processes for in- and contribution is important for fouling control. Membrane autopsy organic scaling because of the salt concentration effect, and especially technology, including investigations on morphology, structure, com- in the tail membrane elements, the concentration polarization effect position, and the microbial community structure of fouled membranes deteriorates the scaling, as the concentration was determined to be and foulants, as well as the interaction between membrane surface and approximately 4–10 times in the polarization layer [5,12]. Natural or- foulants was the key technique for the membrane fouling study in terms ganic matter (NOM) and bacteria were demonstrate to accelerate of the fouling mechanism analysis, selection of pretreatment technol- scaling even when the concentration was lower than the critical sa- ogies, and generating effective but targeted scale inhibiting, steriliza- turation value [13–15]. Organic fouling was another important con- tion, and membrane cleaning strategies [36,37]. The RO membrane tributor, and effluent organic matter (EfOM) and extracellular poly- autopsies were widely reported; however most of them focused on the meric substance (EPS)/soluble microbial polymers (SMP) were the lab-scale or pilot-scale RO process [6,33,38–40]. For the full-scale RO main constituents [16]. The EPS and SMP can significantly promote process, seawater and brackish water desalination processes have trans-membrane pressure (TMP) and increase up to 1000-fold of the gained more attention in the recent five years [20,40], the study of the resistance to bactericides [17]. That finding means that the organic full-scale RO membrane for the MWR is still lacking. Hu and his col- fouling and bio-fouling were mutually related and reinforced [18]. leagues have conducted a series of researches, and they investigated the Meanwhile, the EfOM acts as a nutrient for microbial growth—espe- membrane fouling composition [41], the effect of chemical cleaning cially for the pioneering bacteria— and it can easily adhere to the [42], and the fouling characteristics at different positions in a full-scale membrane surface due to the phenolic and carboxylic functional groups RO reclamation system [43]. However, to the best of our knowledge, of NOM like humic substances (HA), fulvic acids, and humin. As a re- severe membrane fouling in the summer has never been reported. sult, they act as the bridge for inorganic matter and bacteria and de- Meanwhile, most membrane autopsy studies only focus on the com- teriorate the membrane fouling [13,14,19–21]. Colloidal fouling also position of the foulants, key contributor to the membrane fouling and plays a key role in RO membrane fouling, and biopolymers, such as their synergistic effects have not been thoroughly elucidated to date. polysaccharides and protein, and inorganic colloids such as silica, iron The objective of this work is thus to understand the foulant com- hydroxide, and aluminum hydroxide, have been the main contributors position and the key factor of membrane fouling in a full-scale RO [22,23]. process. The performance of the wastewater reclamation process, fou- Bio-fouling was thought to be the most stubborn fouling, as it lant morphology, inorganic and organic matters (OMs), as well as the cannot be effectively removed by pretreatment processes and is less microbial community were studied to analyze the membrane fouling thoroughly understood [12,16,24]. The bacteria irreversibly adhere on formation mechanism. The foulants on the endcap of the membrane the membrane surface, and fast multiplication later occurs in the pre- module and foulants in the entrance of the leading element were also sence of feed water nutrients, afterwards, the stubborn biofilm forms studied, as they are also important contributors to the increase of the together with the EPS and SMP [16,25,26]. However, the microbial TMP. This work offers a systematic investigation of membrane fouling community of membrane foulants showed a significant difference, in the full-scale RO plant for water reclamation, and it will provide which makes the RO membrane fouling complicated. Khambhaty et al. useful information for understanding the membrane fouling and its [27] found γ-proteobacteria showed the highest abundance in a full- control. scale RO plant for brackish water desalination, composed mainly of pseudomonas and xanthomonas, and sphingomonas, which affiliated with β-proteobacteria, showed the third abundance. Several studies support 2. Materials and methods the predominance of γ-proteobacteria in RO fouling [28,29], while other studies presented different findings that other bacteria, such as α-pro- 2.1. Full-scale RO process in TPP teobacteria or β-proteobacteria were the leading bacteria [30–32]. These findings indicated that the microbial community on the RO membrane A lime coagulation-UF-RO combined process was applied in a TPP surface was susceptible to disturbances to such factors as the feed water for municipal water reuse as the circulating cool water and boiler quality, operation and ambient conditions, as well as the type and feedwater as shown in Fig. 1. Lime was added to remove calcium and frequency of antiscalants, bactericides, and chemical cleaning regents magnesium for softening, which caused high concentrations in the secondary effluent (SE) and were two important contributors to the RO

Fig. 1. Schematic diagram of the lime coagulation-UF-RO process for water reclamation in a thermal power plant.

844 L. Zheng et al. Journal of Membrane Science 563 (2018) 843–856

Fig. 2. Two-stage RO membrane system in a TPP together with foulants sampling’ positions. membrane scaling [5]. After coagulation, the mechanical acceleration and fouled membrane were kept in a − 4 °C refrigerator. clarification was applied to remove the suspended solids and flocs. Next, the UF membrane was used as a pretreatment process to further 2.3. Analysis methods remove particles, colloids, and macromolecular organic matter. Four two-stage RO membrane stacks were built for the advanced treatment 2.3.1. Water samples (as shown in Fig. 2). The capacity of the full-scale plant was The water samples’ pH and conductivity were tested by the elec- – 3 20,000 30,000 m /d, and the recovery rate was approximately 75%. trode method. The COD was tested by a HACH COD kit (2–150 ppm, The details can be found in the Supporting Information. + - - HACH, USA). The total nitrogen, NH4 -N, NO3 -N, NO2 -N, total The problem for the reclamation process was the severe membrane 3- phosphorous, and PO4 -P was tested by UV–Vis spectrophotometry fouling in summer starting from June, and the TMP increased sig- (Evolution 201, Thermo Scientific, USA) according to the standard fi ni cantly in the 1st-stage of the RO process. As a result, frequent che- method [44]. The metals and anions content were detected by an in- mical cleaning was needed for approximately once every two weeks, ductively coupled plasma optical emission spectrometry (ICP-OES, which was once every one to two months in other seasons. Thus, the Optima 8300, PerkinElmer, USA), inductively coupled plasma mass fi rst element of the 1st-stage RO membrane was autopsied in June to spectrometry (ICP-MS, NexION 300 ×, PerkinElmer, USA), and ion study the fouling composition and mechanism. The foulants in the chromatography (IC, ICS-1000, DIONEX, USA). The dissolved organic endcap and the entrance of the leading membrane element in the 1st- carbon (DOC) was measured by a TOC analyzer (Vario TOC, Elementar, ff stage were also analyzed in this work as which also take e ect on TMP Germany). Three-dimensional excitation-emission matrix fluorescence increase. Then the key contributor and formation mechanism of the spectra (3DEEM) was used to characterize the dissolved organic matter membrane fouling were analyzed. (DOM) by a fluorescence spectrophotometer (F-7000, Hitachi, Japan). The molecule distribution was analyzed by high-performance size ex- clusion chromatography (HPSEC, Waters 1525, Waters, USA). The de- 2.2. Water, membrane, and foulants sampling tails can be found in the Supporting Information (SI).

The influent of the system, the effluent of coagulation, UF, and RO, and RO concentrate were collected in winter (January), spring (March), 2.3.2. Foulants and fouled membrane and summer (June) to investigate effects of water quality and season on The morphology of the foulants and membranes was observed by fi membrane fouling. Meanwhile, the online chemical cleaning effluent eld emission scanning electron microscopy (FE-SEM, HITACHI was also collected to investigate the composition of the membrane SU8020, Hitachi, Japan) and atomic force microscope (AFM, Shimadzu foulant. All of the water samples were taken back to laboratory and SPM-9600, Shimadzu, Japan). The interface element content was con- filtered immediately by 0.45 µm filter membrane after pH and chemical ducted by a Hitachi S-3000N scanning electron microscopy (SEM) oxygen demand (COD) test, and then stored at 4 °C to minimize the equipped with energy dispersive spectroscopy (EDS), and EDS mapping change of constituents in water samples. was applied to investigate the distribution of each element. A fourier As the rising TMP significantly happened in the first stage and transform infrared spectrum (FTIR) was applied for functional group fl distinct foulants was found in the entrance of the lead element. recognition and organic fouling analysis in attenuated total re ectance – −1 Membrane fouling analysis was emphasized on the first element in 1st- (ATR) mode in a wave number range of 4000 650 cm by cumulating −1 stage RO membrane. The foulant in the entrance of the lead element 64 scans at a resolution of 2 cm using a Nicolet iS10 FTIR spectro- fi was taken in six paralleled row, which was named as E1-E6. The foulant meter (Thermo Scienti c, USA). Details can be found in the Supporting on the endcap of the membrane module in the front side was also Information (SI). collected and named as C1. To investigate the vertical distribution of the membrane fouling, 3 samples were obtained for foulant analysis on 2.3.3. Microbial community analysis membrane surface. In detail, the unabridged foulant on membrane Ten samples including E1-E6, C1, and M1-M3 were collected for surface together with the membrane was named as M1 which re- microbial community analysis. Total genomic DNA was extracted by a presents the total foulant; the foulant scraped by steel ruler from FAST DNA Spin Kit for Soil (MP Biomedicals, USA) according to stan- membrane surface was coded as M2 which represents the upper layer of dard protocol. Next, the bacteria community were analyzed by a high- the membrane fouling, foulant scoured by water was named as M3 throughput sequencing method. The microbial community composition which represents the loose top layer of membrane foulant. The sam- and similarity of the foulants were described using PCR amplification, pling position of foulants was presenting in Fig. 2, all of the foulants clone library construction, and phylogenetic analyses based on 16 S

845 .Zege al. et Zheng L. Table 1 Water quality of reclaimed water recycle process in Winter, Spring, and Summer.

Winter Spring

Influent Coagulation UF effluent RO effluent RO concentrate Influent Coagulation effluent effluent

pH 8.6 ± 0.1 8.2 ± 0.7 8.4 ± 0.3 7.5 ± 0.7 8.4 ± 0.1 8.3 ± 0.0 8.4 ± 0.2 Turbidity (NTU) 3.2 ± 1.6 0.4 ± 0.2 0.4 ± 0.2 0.1 ± 0.1 0.8 ± 0.3 2.5 ± 2.2 1.2 ± 1.1 Conductivity(μS/cm) 2340 ± 43 2240 ± 58 2310 ± 46 91 ± 12 6540 ± 53 2060 ± 51 2260 ± 53 COD (mg/L) 49 ± 10 28 ± 12 12 ± 3 n.a. 42 ± 5 27 ± 5 22 ± 3 TOC (mg/L) 14.2 ± 2.9 9.9 ± 1.8 6.9 ± 1.6 n.a. 15.4 ± 2.4 10.0 ± 1.7 8.2 ± 1.6 Al kality (mg/L) 760 ± 46 165 ± 35 148 ± 14 10 ± 3 803 ± 31 650 ± 50 202 ± 44 Na (mg/L) 254.6 ± 11.2 275.3 ± 25.1 183.5 ± 86.2 8.6 ± 2.4 1028.5 ± 32.1 289.6 ± 23.2 290.6 ± 15.9 Mg (mg/L) 39.6 ± 0.8 42.4 ± 2.4 38.0 ± 3.6 n.a. 103.1 ± 21.2 38.9 ± 1.5 31.3 ± 0.9 K (mg/L) 37.8 ± 3.5 21.8 ± 1.3 13.9 ± 0.8 1.2 ± 0.7 45.3 ± 1.3 33.2 ± 2.9 21.0 ± 1.0 Ca (mg/L) 110.4 ± 10.6 40.1 ± 1.0 71.1 ± 12.3 n.a. 240.3 ± 6.6 126.3 ± 1.6 73.2 ± 2.1 Cl (mg/L) 261.3 ± 4.7 256.8 ± 4.6 161.1 ± 3.9 6.4 ± 2.2 507.2 ± 13.7 240.0 ± 1.4 238.1 ± 15.7 SO4 (mg/L) 334.3 ± 5.8 432.5 ± 8.4 317.1 ± 1.1 2.6 ± 1.31 1020.3 ± 45.7 231.2 ± 3.6 252.7 ± 22.1 TP (mg/L) 4.7 ± 1.2 1.2 ± 0.2 0.7 ± 0.6 0.1 ± 0.2 3.1 ± 0.4 4.3 ± 0.3 1.5 ± 0.3 3- PO4 -P (mg/L) 1.6 ± 0.6 0.8 ± 0.2 0.6 ± 0.2 n.a. 1.2 ± 0.1 1.8 ± 0.3 1.3 ± 0.1 TN (mg/L) 13.2 ± 0.8 10.8 ± 1.0 11.3 ± 1.3 0.9 ± 0.3 35.1 ± 2.6 15.6 ± 0.4 14.6 ± 0.5 - NO3 -N (mg/L) 11.5 ± 1.2 9.9 ± 0.9 10.1 ± 1.4 0.2 ± 0.3 31.3 ± 1.8 12.7 ± 30.6 6.3 ± 0.3 - NO2 -N (mg/L) 0.01 ± 0.01 0.01 ± 0.00 n.a. n.a. 0.01 ± 0.01 0.01 ± 0.00 n.a. + NH4 -N (mg/L) 2.4 ± 0.2 2.7 ± 0.2 2.2 ± 0.2 n.a. 7.1 ± 0.2 2.6 ± 0.3 0.3 ± 0.2 LSI 1.28 − 0.14 0.24 n.a. 1.46 1.01 0.40 RSI 5.99 8.52 7.94 n.a. 5.46 6.27 7.63 SCaSO4 1.51 2.99 1.77 n.a. 5.60 0.54 1.24 846 Total silicon(mg/L) 17.8 9.2 9.1 n.a. 23.6 29.1 12.9

Spring Summer

RO effluent RO concentrate Influent Coagulation UF effluent RO effluent RO concentrate effluent

pH 7.8 ± 0.5 8.5 ± 0.1 8.3 ± 0.1 8.3 ± 0.2 8.3 ± 0.2 7.8 ± 0.4 8.5 ± 0.2 Turbidity (NTU) 0.1 ± 0.0 0.9 ± 0.5 2.3 ± 1.5 0.4 ± 0.6 0.2 ± 0.1 0.1 ± 0.1 0.7 ± 0.6 Conductivity(μS/cm) 50 ± 3 6720 ± 103 1994 ± 32 1716 ± 87 1719 ± 82 49 ± 39 6150 ± 54 COD (mg/L) n.a. 68 ± 12 27.9 ± 5.1 19.8 ± 4.1 18.6 ± 3.3 0.5 ± 1.2 53.0 ± 3.6 TOC (mg/L) 0.1 ± 0.0 16.5 ± 2.6 9.9 ± 1.2 8.2 ± 0.9 7.8 ± 1.9 1.1 ± 0.3 15.7 ± 2.0 Al kality (mg/L) 9 ± 4 462 ± 12 658 ± 20 131 ± 51 136 ± 51 12 ± 2 431 ± 32 Na (mg/L) 10.6 ± 1.8 1210.8 ± 14.5 237.7 ± 11.4 243.4 ± 12.7 239.9 ± 24.5 7.6 ± 2.3 932.2 ± 23.2 Mg (mg/L) n.a. 107.7 ± 2.6 42.3 ± 1.2 29.0 ± 4.0 28.7 ± 4.8 n.a. 102.7 ± 1.6 Journal ofMembraneScience563(2018)843–856 K (mg/L) 1.2 ± 0.9 71.9 ± 3.4 27.9 ± 1.3 24.2 ± 4.6 24.1 ± 5.3 0.8 ± 0.5 74.1 ± 6.0 Ca (mg/L) n.a. 247.1 ± 3.6 139.7 ± 4.7 79.3 ± 13.8 78.5 ± 15.0 5.5 ± 7.7 256.31 ± 4.72 Cl (mg/L) 2.8 ± 0.9 713.3 ± 25.9 334.8 ± 2.8 336.1 ± 3.2 335.6 ± 2.5 3.3 ± 0.3 1097.6 ± 4.8 SO4 (mg/L) n.a. 809.5 ± 25.1 249.5 ± 3.8 242.1 ± 3.2 262.6 ± 2.8 1.2 ± 0.6 932.9 ± 8.5 TP (mg/L) n.a. 5.0 ± 1.4 4.3 ± 0.4 0.6 ± 0.8 0.5 ± 0.3 0.2 ± 0.1 1.6 ± 0.7 3- PO4 -P (mg/L) n.a. 3.9 ± 0.1 1.7 ± 0.3 1.3 ± 0.4 1.3 ± 0.4 n.a. 5.0 ± 0.4 TN (mg/L) 0.9 ± 0.2 40.3. ± 6.2 19.1 ± 3.4 14.1 ± 2.4 14.2 ± 3.8 1.7 ± 1.0 45.8 ± 3.2 - NO3 -N (mg/L) 0.2 ± 0.1 38.4 ± 2.7 14.4 ± 1.1 11.1 ± 2.1 11.0 ± 2.0 1.1 ± 04. 34.2 ± 1.7 - NO2 -N (mg/L) n.a. n.a. 0.01 ± 0.01 0.01 ± 0.03 0.02 ± 0.02 n.a. 0.01 ± 0.01 + NH4 -N (mg/L) 0.6 ± 0.1 1.9 ± 0.2 2.7 ± 0.1 0.6 ± 0.1 0.6 ± 0.1 0.1 ± 0.1 2.0 ± 0.9 LSI n.a. 1.34 1.07 0.13 0.12 − 2.63 1.36 RSI n.a. 5.81 6.16 8.06 8.05 13.01 5.79 SCaSO4 n.a. 3.87 0.55 1.10 1.27 n.a. 4.75 Total silicon(mg/L) n.a. 9.1 23.7 26.4 14.7 n.a. 32.8

Note: n.a.- not detected. L. Zheng et al. Journal of Membrane Science 563 (2018) 843–856 rRNA gene based molecular techniques [45–48]. The details can be concentrate, which was derived from the RO influent and the microbial found in the Supporting Information (SI). metabolism as the concentration in RO influent was low. The fluorescence index, FI370, of all samples was higher than 2, indicating the 2.4. Membrane chemical cleaning source of fulvic acid was microorganism sources. The BIX data (approximately 1.2) indicated the strong microbial activity in the source

The fouled membrane was cleaned by nitric acid (HNO3), sodium water. The aromaticity of the water was decreased after coagulation hypochlorite (NaClO), and sodium hydroxide (NaOH) to investigate the indicating the strong selectivity of coagulation process for aromatic composition of the foulants. The fouled membrane was cut into DOM, while the UF did not change the HIX of water. The RO eﬄuent 5 × 5 cm sections and immersed to 100 and 50 ml solutions of 1 M showed the lowest HIX, indicating that the RO has a high selectivity of

HNO3, 1 M NaOH, and 5% NaClO, respectively. Then all samples were aromatic DOM; that also can explain the higher HIX of the RO con- placed into a constant temperature shaker for 24 h under 200 rpm and centrate [54,55]. 30 °C. More importantly, the cleaning solutions were carefully analyzed The molecular weight (MW) distribution of the water samples are by 3DEEM, ICP-OES, and ICP-MS to study the composition of the fou- exhibited in Fig. 4. The MW distribution curves of the influent, coa- lants processed by the different reagents. gulation effluent, UF effluent, and RO concentrate were the same. Among them, the compounds in the peak among 500–800 Da were 3. Results and discussion believed to be the building blocks include hydrolysates of humic substances. While the other three peaks among 800–1500 Da were attrib- 3.1. Performance of the advanced treatment process uted to the humic substances [29,43]. This means that the humic substance was the main organic in the SE and the influent of the RO, which The water quality of the process in January, March, and June is was also proved by the 3DEEM spectra in Fig. 3. However, the OMs shown in Table 1. The pH was approximately 8.4 for all samples, which concentration exhibited the following sequence: RO concentrate > was slightly higher and may benefit the inorganic salt precipitation on influent > coagulation effluent≈UF effluent. This means that the coa- the membrane surface, especially when concentration polarization was gulation process removes some OMs because humic substances have a considered. Table 1 shows that all parameters showed no significant strong affinity to metal hydroxides due to the phenolic and carboxylic variation in the different seasons, although COD and TOC were a little functional groups, as also exhibited in Fig. 3. The UF showed no sig- higher in winter. Meanwhile, the water quality has no significant nificant efficiency for OMs removal, while the RO presenting high re- change during the treatment process including coagulation, UF, and RO jection as the significantly increased intensity in the RO concentrate in different season. The concentrations of OMs, inorganic ions, ni- and significantly low intensity in the RO effluent. Meanwhile, the low trogen, and phosphorous in the RO concentrate were approximately UV254 intensity was detected with a molecular weight around 160 Da 3–4 times as those in the RO influent (i.e., UF effluent), which was for the RO effluent, which was believed to consist of low-molecular- consistent with the 75% recovery rate of the RO. It can be concluded weight neutrals and amphiphilics as shown in Fig. 4 [18]. that the water quality was not the main reason for the severe membrane fouling in summer. Thus, we believe that the microbial proliferation in 3.2. Morphological properties summer because of the higher microbial activity was the main reason, in other words, the microbial fouling may be a key factor in RO All foulants displayed a light brown color and unconsolidated membrane fouling. structure that gave off a fishy smell, and the feel of touching was wet The scaling tendency of the influent and the water samples during and creamy, indicating the existence of bacteria. The SEM images the treatment process was calculated in terms of the Langelier satura- (Fig. 5) showed that C1 was approximately 500 µm in particle size, and tion index (LSI), the Ryznar stability index (RSI) and the sulfate solu- it was clear that OMs was the main pollutant, as the dense surface was bility for CaSO4 (SCaSO4) to evaluate the potential inorganic fouling found. Meanwhile, rod-shaped bacteria inlaid in the foulant in the 2- [49,50]. For all samples, SCaSO4 ≫min (CCa,CSO4 ), indicated that there 5000 × magnified image (Fig. 5) and Fig. S1A. The foulants E1-E6 was no CaSO4 scaling tendency. However, it showed CaCO3 scaling showed the same structure with a bigger size, indicating the aggrega- tendency for the influent, coagulation effluent, and UF effluent (i.e. RO tion of bacteria. Rod-shaped bacteria dominated the microbial com- influent). The RO concentrate showed the highest LSI and RSI because munity, where heteromorphic cells also appeared, which can be found of its high calcium, alkalinity, and sulfate, meaning that the inorganic in Fig. S1B and Fig. S1C. And it was interesting to find the reticular scaling has significant potential in the RO membrane, especially in the structure between the microbial cells, which was believed to be the last few elements [51,52]. In consideration of the concentration po- reason for the aggregation of bacteria and the tight biofilm. The open larization, inorganic scaling should not be overlooked in the MWR by structure in the fouling layer indicate the existence of bacteria, as can the RO process. be found in the 5000 × magnified image (Fig. 5) as well as Fig. S1F. The organic composition and removal efficiency were also measured The salt crystals also appeared in the outer side of the membrane, in- by 3DEEM and HPSEC. As shown in Fig. 3 and Table 2, four main peaks dicating the inorganic scaling (shown in Fig. S1D and Fig. S1E). The were found in 230/330, 235/345, 255/425, and 280/340 (Ex(nm)/Em inner side of RO membrane was clean as most of the pollutant, in- (nm)), which indicated the main foulants were amino acid-based pro- cluding ions, was rejected by the membrane. However, certain parti- teins (typically Tyrosine), fulvic acid-like, humic acid-like, and SMP- cles, which were believed to be salt crystals, were also found as some like OMs, which were affiliated with regions Ⅱ, Ⅲ, Ⅳ, and Ⅴ [53]. For cations and anions can pass through the RO membrane as shown in the SE, the amino acid-based proteins and fulvic/humic acid-like OMs Table 1. were the main components, where SMP also existed by the microbial The fouled membrane was also detected by AFM to study the surface metabolism in the Anaerobic-Oxic process in the WWTP [43]. The morphology of both the outer and inner side, the results were exhibited coagulation process was effective for removing these pollutants, as the in Fig. S2. The membrane inner surface was smooth, indicating it was fluorescence intensity was significantly decreased, especially for SMP not severely fouled. The outer surface showed rugged structure, and it with a higher molecular weight. The UF process offered tiny removal can be clearly recognized as the organic fouling layer on membrane efficiency for OM; however, it was almost rejected by the RO mem- surface. Especially, the protrusion structure was sleek compared to the brane, although a subtle intensity can be found in region Ⅱ. Thus, more membrane outer surface after NaOH cleaning, as shown in Fig. S2, OMs can be found in the RO concentrate because RO has approximate which was significantly sharp with a higher roughness. Thus, it can be 75% recovery rate; the main part was protein, humic acid-like and concluded that the outer layer of the fouling layer was OMs and, as fulvic acid-like substances. SMP was also detected in the RO generally agreed, NaOH was effective for OMs removal on the

847 L. Zheng et al. Journal of Membrane Science 563 (2018) 843–856

Fig. 3. 3DEEM spectra of A: Influent; B: Coagulation effluent; C: UF effluent; D: RO effluent; E: RO concentrate; F: RO chemical cleaning effluent. (Note: the inserted number was the dilution ratio).

membrane surface. The HNO3 was not effective for RO membrane M1. A significant increase of Ca, Mg, and Si and a decrease of N can be cleaning as the tight organic fouling layer was still on the membrane found in the membrane foulant compared to others, and the con- surface. centration polarization on the membrane surface was attributed to be Based on the SEM and AFM results, it was clear that the OMs and the main reason. That means that the inorganic scaling was also a bacteria were the most important reason for the membrane fouling of contributor for the membrane fouling due to the salt rejection by the the full-scale RO process. The bacteria act as a coupling agent for the membrane (high tendency of surface crystallization) and the con- OMs because of the assimilation process and the release of EPS; centration polarization (high tendency of bulk crystallization). The therefore, the dense and smooth fouling layer was formed in the inner side of the membrane was clean as the content of other elements membrane surface [5,24]. Meanwhile, inorganic fouling also appeared was low, except the C, N, and O, same composition as polyester non- because of the concentration polarization and the interaction with the woven fabric supporting layer. However, some elements, such as Na OMs, microbes, and membrane surface. However, as the composition and Mg, had crossed the membrane, as detected by EDS (Table 3), and concentration of both inorganic and OMs in all corresponding which has also been proved by SEM. Due to the element composition samples of influent, coagulation effluent, UF effluent, RO effluent, and detected by EDS, it can be inferred that the inorganic scaling was

RO concentrate in different season did not show a significant difference mainly composed of SiO2 and salts of calcium and magnesium, princi- (Table 1), it was believed biofouling was the key reason for the severe pally as sulfate and phosphate. The EDS mapping of Ca, Mg, Na, Si, K, membrane fouling in summer. Fe, and Al is shown in Fig. 6. The scale on the top of membrane foulant can be recognized due to the EDS-mapping images, and Si, Ca, Mg, Fe, and Al were the main sediments. Meanwhile, these elements also ex- 3.3. Inorganic composition isted in the tight fouling layer; K and Na were evenly distributed on the membrane surface, resulting from the interaction with the membrane The element composition of the foulants and fouled membrane was surface, OMs, and the microbial assimilation. Phosphate was attributed measured by EDS and is exhibited in Table 3. The main elements in the to the distribution of bacteria on the membrane surface, and a phos- foulants are Ca, S, P, Mg, Si, K, Na, Fe, and Al. E1-E6 showed the same phate deposit, believed to be calcium phosphate, can also be found on element composition and the content of C, N, and O was approximately the membrane surface. However, the concentration of P by EDS was a 95–97%, indicating that organic matter/bacteria were the major com- little deviant due to the coating of platinum, especially when compared ponents, which was in accordance with the SEM. C1 has the highest to its concentration by ICP-OES as shown in Table S1 (Table 4). total content of C, N, and O, indicating that the inorganic component To further analyze the composition of the membrane foulant, the was not easy to deposit or be absorbed in the endcap, as the con- element concentration in the cleaning reagent was also measured and is centration in the influent was low. Sulfur presented a significantly high exhibited in Table S2. The elements K, Mg, Na, Ca, Fe, and Al showed content due to the detection of the supporting layer (polysulfone) in

848 L. Zheng et al. Journal of Membrane Science 563 (2018) 843–856

Table 2 Fluorescence intensity and symbols of the MWR process in the TPP.

Influent Coagulation effluent UF effluent RO effluent RO concentrate Cleaning Effluent

Region I Normalized Intensity (107AU nm2) 2.39 1.75 1.95 1.39 1.93 2.14 Percentage intensity (%) 8.47 18.96 19.92 30.05 13.87 19.86 Peak 1 (Ex/Em) 230/300 220/285 220/290 220/290 225/300 225/295 Max. Intensity 818.1 916 888 888 637.9 853.2 EEM symbol Tyrosine-like protein Peak 2 (Ex/Em) 230/330 230/330 230/330 225/320 230/330 225/330 Max. Intensity 3684 1338 1669 662.1 2142 1791 EEM symbol Aromatic protein Region Ⅱ Normalized Intensity (107AU nm2) 7.84 2.90 2.96 1.17 4.57 0.45 Percentage intensity (%) 27.83 31.43 30.19 25.27 32.84 4.18 Peak 1 (Ex/Em) 235/355 235/350 235/340 225/335 235/345 225/355 Max. Intensity 5085 1742 1805 793.3 2972 2289 EEM symbol Aromatic protein Peak 2 (Ex/Em) 230/370 Max. Intensity 2421 EEM symbol Aromatic protein Region Ⅲ Normalized Intensity (107AU nm2) 9.66 2.35 2.47 0.93 3.95 3.39 Percentage intensity (%) 34.27 25.47 25.26 19.95 28.40 31.38 Peak 1 (Ex/Em) 240/385 240/385 240/385 235/395 235/385 240/385 Max. Intensity 3680 1066 1116 337.8 1905 1883 EEM symbol Fulvic acids Peak 2 (Ex/Em) 250/435 240/410 Max. Intensity 3798 1050 EEM symbol Humic acid-like Region Ⅳ Normalized Intensity (107AU nm2) 5.09 1.65 1.79 0.96 2.42 3.82 Percentage intensity (%) 18.04 17.92 18.29 20.80 17.38 35.44 Peak 1 (Ex/Em) 285/340 280/340 275/345 280/340 280/345 280/355 Max. Intensity 1914 707.3 1943 269.4 1771 1900 EEM symbol SMP-like SMP-like SMP-like SMP-like SMP-like SMP-like Region Ⅴ Normalized Intensity (107AU nm2) 3.21 0.57 0.62 0.18 1.04 0.99 Percentage intensity (%) 11.37 6.23 6.35 3.93 7.51 9.14 Peak 1 (Ex/Em) 255/425 255/410 255/390 255/395 255/395 255/385 Max. Intensity 3462 648.5 671.6 209.9 1084 2057 EEM symbol Humic acid-like Peak 2 (Ex/Em) 255/420 Max. Intensity 601.3 EEM symbol Humic acid-like FI370 2.45 2.28 2.49 2.03 2.39 2.33 BIX 1.17 1.21 1.24 1.11 1.22 2.31 HIX 0.74 0.53 0.54 0.33 0.62 0.39

Table 1. The microbial assimilation or binding is the key reason for the high abundance of K and Na because they had low concentration in water sample but high content on the membrane foulant, and they have high solubility, and also show a low tendency to deposit on the membrane surface. That also indicates the microbial fouling on the membrane surface. What also need to mention is that Al and Fe showed a high content in the membrane foulant, and they are believed to be derived from the coagulant in the WWTP and the TPP. As the concentration of Al and Fe was low in the RO influent, it means they have a strong interaction with the membrane or the composition on the membrane surface [43]. Nitrogen and sulfur were also important composition for membrane fouling, which also showed high a concentration in the RO influent. And it also interesting to found that Mg, - K, Cu, As, and NO3 showed the highest content by means of the NaOH 2- cleaning, while Al, Cr, Se, Pb, and SO4 showed the highest content in NaClO effluent; Ca, Na, Mn, Fe, Cd, Ba, and Cl showed the highest

content by HNO3 cleaning, which is a little different from the common Fig. 4. Molecule distribution of all samples by HPSEC. knowledge that inorganic matters was easy to be cleaned by acids. The possible reason was the depth distribution of different elements in the fouling layer. high a content in the membrane foulant, while nitrate and sulfate were the main cations, similar to the EDS results. The Mg and Ca were de- posited on the membrane surface as the saturation was low because the 3.4. Organic matter pH of the RO influent was high. It can be inferred that Mg was easier to deposit on the membrane surface, as the concentration in the RO in- An ATR-FTIR analysis was conducted to investigate the functional fluent was lower than Ca. Another two important element were K and (reactive) groups of the organic foulants. As shown in Fig. 7, the foulant Na although the concentration were low in the RO influent, as shown in E1-E6 and C1 showed similar spectra, indicating similar a DOM composition. That means the parallel membrane elements in the RO process

849 L. Zheng et al. Journal of Membrane Science 563 (2018) 843–856

Fig. 5. SEM images of the foulants (Note: the inserted number was the magnification factor). showed the same fouling mechanism due to the same operation con- ATR, respectively. That means the porous polysulfone supporting layer ditions. The major bands of the foulants were at 3279 (υ-NH), 2972 (υ- was detected as the polyamide layer that was about 200 nm in the RO CH), 1635 (υ-C=O), 1540 (υ-C-N), 1405 (υ-CH), 1240 (υ-P=O), 1050 membrane. However, the peak shift was found in 1650, 1475, 1408, − − (υ-C-O), and 628 (υ-CH) cm 1 [6,55,56]. The peaks at 3279, 1635, and 1235, 1147, 554, and 579 cm 1, which was result from the organic − 1540 cm 1 were attributed to proteinaceous compounds, and the ab- fouling on the membrane surface. Meanwhile, the presence of proteins − − sorption bands at 3279, 2972, and 1405 cm 1 indicate the presence of (3279, 1635, and 1540 cm 1), humic substances (3279, 2972, and − − biological-derived humic and fulvic acid. The broad band in 1405 cm 1) and polysaccharides (2972 and 1050 cm 1) can be re- − 900–1100 cm 1 indicates the presence of carbohydrate-like com- cognized in the membrane foulants, which can also be confirmed by − pounds, together with the peak at 2972 cm 1. The absorption bands at HPSEC and 3DEEM. The membrane inner surface showed a 94.13% − − 1050 cm 1 and 2972 cm 1 reflect a polysaccharide related to aliphatic similarity to polyester from spectrum library Common materials, in- ether [18,31,39,42]. Thus, it can be concluded that the main DOM in dicating it was clean as the supporting layer was polyester. the foulants in the entrance of the membrane element and the endcap The DOM composition of membrane foulant was measured by were protein, humic substances, and polysaccharides. The weak ab- analyzing the chemical cleaning effluent. As shown in Fig. 3F, the − sorption peak at 1240 cm 1 originated from the vibration of the P=O concentration of DOM was significantly high compared to the RO in- bond in nucleic acid, indicating the bio-fouling on the RO mem- fluent, indicating the severe organic fouling. The intensity in 250/440 brane [42]. The FTIR spectra of the membrane's outer surface showed and 280/340 (EX(nm)/EM(nm)) was significantly increased compared 91.21% and 88.56% similarity to polysulfone from two spectral li- to that in 238/360 (Ex(nm)/Em(nm)), which means that the humic-like braries HR specta polymers and plasticizers and HR sprouse by and SMP were two major contributors to membrane fouling, while

Table 3 Element composition in the foulants and fouled membrane measured by EDS.

C1 E1 E2 E3 E4 E5 E6 Outer Inner

C (wt%) 62.142 ± 2.22 63.456 ± 1.02 62.784 ± 1.00 63.093 ± 1.89 62.205 ± 2.40 62.418 ± 1.42 62.419 ± 1.45 70.295 ± 0.19 73.865 ± 0.08 N (wt%) 11.068 ± 1.67 8.277 ± 1.32 9.109 ± 0.85 11.454 ± 1.09 8.22 ± 0.81 9.477 ± 1.46 9.640 ± 1.34 4.885 ± 0.77 3.195 ± 0.42 O (wt%) 24.006 ± 1.26 24.690 ± 2.14 23.424 ± 1.88 22.141 ± 2.64 24.882 ± 2.94 23.990 ± 1.01 23.626 ± 2.18 16.400 ± 0.62 20.540 ± 0.70 Na (wt%) 0.244 ± 0.04 0.201 ± 0.04 0.110 ± 0.08 0.161 ± 0.06 0.233 ± 0.08 0.135 ± 0.11 0.139 ± 0.12 0.125 ± 0.02 0.115 ± 0.04 Mg (wt%) 0.116 ± 0.04 0.293 ± 0.02 0.270 ± 0.07 0.230 ± 0.06 0.240 ± 0.05 0.233 ± 0.04 0.264 ± 0.07 0.395 ± 0.05 0.115 ± 0.09 Al (wt%) 0.062 ± 0.05 0.091 ± 0.07 0.091 ± 0.05 0.036 ± 0.04 0.092 ± 0.05 0.075 ± 0.05 0.084 ± 0.06 0.135 ± 0.01 0.050 ± 0.00 Si(wt%) 0.122 ± 0.03 0.123 ± 0.07 0.202 ± 0.06 0.058 ± 0.05 0.475 ± 0.24 0.203 ± 0.16 0.190 ± 0.15 0.425 ± 0.22 0.035 ± 0.05 P (wt%) 0.652 ± 0.07 0.724 ± 0.17 0.863 ± 0.22 0.814 ± 0.28 0.710 ± 0.18 0.848 ± 0.29 0.751 ± 0.21 0.695 ± 0.04 0.135 ± 0.12 S (wt%) 0.496 ± 0.09 0.873 ± 0.10 1.200 ± 0.08 0.835 ± 0.11 1.512 ± 0.41 0.835 ± 0.11 1.300 ± 0.60 3.625 ± 0.36 0.085 ± 0.12 Cl (wt%) 0.058 ± 0.05 0.073 ± 0.07 0.081 ± 0.08 0.070 ± 0.07 0.137 ± 0.07 0.128 ± 0.08 0.097 ± 0.08 0.275 ± 0.05 0.040 ± 0.06 K (wt%) 0.100 ± 0.04 0.036 ± 0.06 0.072 ± 0.05 0.037 ± 0.06 0.172 ± 0.13 0.153 ± 0.13 0.104 ± 0.14 0.160 ± 0.01 0.080 ± 0.11 Ca (wt%) 0.290 ± 0.06 0.730 ± 0.08 1.306 ± 0.39 0.618 ± 0.15 0.780 ± 0.31 1.045 ± 0.48 0.891 ± 0.38 1.490 ± 0.18 0.055 ± 0.15 Fe (wt%) 0.650 ± 0.88 0.424 ± 0.16 0.488 ± 0.12 0.456 ± 0.16 0.245 ± 0.22 0.458 ± 0.15 0.496 ± 0.13 0.650 ± 0.35 0.015 ± 0.29 Total (C,N,O) 97.216 96.423 95.317 96.688 95.307 95.885 95.685 ––

850 L. Zheng et al. Journal of Membrane Science 563 (2018) 843–856

Fig. 6. EDS-mapping of elements on membrane outer surface. amino acid-like protein has more possibility to dissolve in concentrate. in RO membrane fouling. Note that a new peak was found at 320/380 (Ex(nm)/Em(nm)), which In conclusion, the primary organics in the membrane foulant were was recognized as humic substance, in the cleaning effluent. It was proteins, humic substances and polysaccharides. Moreover, most of believed to be derived from the microbial metabolism on the membrane them were biologically derived humic and SMP-like substances. That surface as it was not found in the influent as proved by FTIR. The DOM finding indicates the importance of bio-fouling and bio-derived organic molecular weight distribution of the chemical effluent was presenting fouling in the RO process. in Fig. 4. The UV254 absorbance was significantly high which also indicates a higher DOM concentration, especially for two peaks at 1260 3.5. Microbial community structure and 1590 Da. That means that these two humic substance were biological originated [18]. Meanwhile, a broad peak among As shown in the SEM micrographs and confirmed by FTIR and 11,220–50,000 Da appeared, which was a one-to-one correspondence 3DEEM, microbial fouling is a very important contributor to membrane to the peak at 320/380 (Ex(nm)/Em(nm)) in the 3DEEM spectra, in- fouling. The microbial community composition and similarity of the dicating its microbial source [57]. The 3DEEM spectra of the cleaning foulants were described using the PCR amplification, clone library solutions are presented in Fig. S3; it is clear that the DOM concentration construction, and phylogenetic analyses based on the 16 S rRNA gene- was high in the 50 ml cleaning reagent. The NaOH was more effective at based molecular techniques. removing OMs, especially for the humic substances, as a high intensity peak appeared in region V. The NaClO can also remove most of the OMs 3.5.1. Similarity although the cleaning performance was weaker than NaOH, as shown in The similarity of the 10 samples based on the OUT analysis is pre- Fig. S3. The HNO3 can only remove amino acid-based proteins from the sented in Fig. 8A; the foulant on the membrane surface (M1) showed a fouled membrane, and showed a low performance on OMs removal as significant difference from other samples. The M3 and C1 has high si- widely reported in other researches. Thus, the alkaline chemical milarity and shared the same cluster, as shown in Fig. 8B; the reason is cleaning reagent should be considered in cleaning strategy develop- that M3, representing the uppermost layer of the membrane foulant on fl ment, and these results indicate that organic fouling has more in uence the membrane surface, was in contact with the influent, the same as C1.

851 L. Zheng et al. Journal of Membrane Science 563 (2018) 843–856

Table 4 Concentrations of inorganic matters in the membrane surface by diﬀerent chemical cleaning reagents.

The E1-E6 showed a similar microbial community and shared one membrane fouling in different positions in a single line of the entire cluster because of the similar operation and cleaning conditions. full-scale two-stage RO process has been investigated and reported However, differences still existed although all samples were obtained widely, but the difference of the paralleled membrane module has not from the totally parallel membrane process, especially for E5. The gained attention [43]. causes may relate to the subtle differences during the general operation An interesting finding is the significant difference of the microbial and cleaning process because of the hydraulic differences. The community for M1, M2, and M3, as shown in both neighbor joining

852 L. Zheng et al. Journal of Membrane Science 563 (2018) 843–856

tolerance to non-oxidizing bactericides during sterilization and the membrane chemical cleaning process. The M1, M2, and M3 showed significant differences, as proved by the microbial spatial distribution, which indicated the vertical distribution of microbes in the cross-section of the fouling layer. Rhizobiales, Burkholderisles, Xanthomonadales, and Bdellovibrionales were the major constituents accounting for approximately 80–90% of all microbes at the order level, while Coma- mondaceae, Xanthobacteraceae, Xanthomonadaceae, Bradyrhizobiaceae, and Sphingomonadaceae were the 5 leading bacteria at the family level. Most of the dominate bacteria came from the phylum Proteobacteria, and the same composition can be found for E1 to E6, and a significant difference can be found for M1-M3. At the genus level, the distribution of the microbes was not so centralized, which can be seen in the microbial distribution image and heatmap in Fig. 9. The first 10 microbes occupied approximately Fig. 7. IR spectra of all foulants and fouled membrane by ATR-FTIR. 60–90% of the microbial community as shown in Table S1. The composition of E1 to E6 was similar, except for E5, where a subtle differ- phlyogenetic tree and spatial distribution graphs. That clearly points ence can be found as confirmed by the microbial spatial distribution. out the spatial distribution of microorganisms in the cross-section of the The Acidovorax was the major bacteria occupied in 20–25% of E1 to E6, membrane foulants, which had not been reported to the authors’ which was a gram-negative, catalase- and oxidase-positive, rod-shaped knowledge. The hierarchical structure in the foulant can be explained bacterium from the Comamonadaceae family that was isolated from the by the long-term operation, including the filtration, online cleaning, off- activated sludge [58]. It is an acetate-utilizing denitrifier, which can line cleaning, and also standby status, which could significantly influ- also use fatty acids, sugars, and proteins/ amino acids. The nitrogen and ence the microbial growth and metabolism. The nutrient and dissolved OMs in the RO influent after the UF and cartridge filter were the reason oxygen during the filtration or cleaning processes can also influence the for the growing Acidovorax bacteria. The bacteria with the second microorganisms and promote the formation of hierarchical biofilm. abundance was Starkeya, which belongs to class α-proteobacteria and family Xanthobacteraceae. The bacteria in genus Starkeya was related to 3.5.2. Microbial community composition nitrogen fixation and heterocyclic compounds with nitrogen. The third The microbial community composition with a classification in the abundant genus was Luteimonas, which is affiliated to the phylum phylum, class, order, family, and genus level were investigated to Proteobacteria from the family Xanthobacteraceae. It is a gram-negative analyze which contribute more on membrane fouling. It was clear that and aerobic bacterium which is a cold-resistant and hydrocarbon- proteobacteria and bacteroidetes were two leading bacteria in all fou- bearing microorganism. It can survive in extreme environments and lants, which have a percentage of 87–98% and 1–8% of the whole low carbon but high ammonia wastewater. In E1 and M3, Starkeya microbial community at the phylum level as shown in Table S1 and Fig. occupied 43.8% of the whole microbial community, which was attrib- S4. That explained why the rod-shaped bacteria were the dominate cells uted to the high concentration of nitrogen in the RO influent and ac- in the SEM micrographs. Among them, α-proteobacteria, β-proteo- counted for the high concentration of protein, as shown in the 3DEEM bacteria, γ-proteobacteria and δ-proteobacteria were the four major bac- image. teria at the class level, which accounted for 87–98% of the whole mi- As shown in the spatial distribution graphs, a significant difference crobial community, which was similar to other studies [30–32]. It can can be found for M1, M2, and M3. The diversity of the microbial also be found that the composition of E1 to E6 was similar, and β- community showed a sequence of M1 > M2 > M3, indicating the proteobacteria was the key composition in the class level, while for the vertical distribution of bacteria in the membrane fouling layer. other fixed foulants, α-proteobacteria was the key composition. How- Starkeya, Acidovorax, and Luteimonas were the 3 leading microbes in ever, γ-proteobacteria showed significantly high content in M1 with a M3, which was the same as E1 due to the same circumstances for the sequence of M1 > M2 > M3; that means it was the pioneering bacteria microorganisms. The difference can be attributed to the concentration on the membrane surface, as reported by previous work [29]. Mean- difference because of the concentration polarization. Note the sig- while, it can survive and stay on the membrane because it has a higher nificant decrease of starkeya (from 48.4% to 4.65%) and acidovorax

Fig. 8. Neighbor joining phlyogenetic tree (A) and spatial distribution graph (B) of the microbial community of all foulants.

853 L. Zheng et al. Journal of Membrane Science 563 (2018) 843–856

Fig. 9. Microbial community structure of the foulants on the membrane surface at the genus level. A: Histogram frequency distribution diagram; B: Heatmap. (Note: Top 50 bacteria are exhibited).

(from 17.64% to 5.61%) and increase of pseudoxanthomonas (from Thus, different bactericides or cleaning reagents should be applied to 1.04% to 12.89%) in M3. It also showed that the dominant bacteria are avoid microbial resistance and to benefit bio-fouling controlling. different in different depths, γ-proteobacteria showed more abundance Meanwhile, the bacteria flocs can also be formed in the endcap and the in the bottom layer. It indicate that γ-proteobacteria is the pioneering membrane entrance. However, these bacteria can be easily removed bacteria on the membrane fouling, and the high anti-bactericide during chemical cleaning; they have more similarity to the bacteria on property of γ-proteobacteria was another reason for the higher abun- the upmost layer of the membrane foulants. dance on the deeper layer. However, the aerobic bacteria occupied almost all of the composition in all of the samples; this, means that the 4. Conclusions aerobic environment covers the whole cross-section in the depths of the membrane foulant and was not suitable for anaerobes. In a word, the Membrane fouling was the largest obstacle facing the RO applica- diversity and composition of the microbial community, especially the tion in the MWR, especially in the summer. This report presents a vertical microbial distribution, was important for controlling the systematic investigation of the RO membrane fouling in terms of the membrane fouling and fouled membrane cleaning. inorganic, organic, and microbial composition, and analyzes the key contributors to membrane fouling. The conclusions are as follows: 3.6. Membrane fouling formation mechanism 1. The elements Ca, Mg, Si, Al, and Fe were the main inorganic scales, while proteins, humic substances, and polysaccharides were the Biofouling was believed to be the key contributor for the severe main components of organic fouling. The bacteria-derived OMs membrane fouling in summer, and stronger biological activity in showed a greater contribution to membrane fouling. Bio-fouling is summer may be an important reason because the variation in water the key contributor in the full-scale RO reclamation process, the quality of the SE and the advanced treatment process showed no sig- main bacteria were α-proteobacteria, β-proteobacteria, γ-proteo- nificant variation. As indicated by microbial community, γ-proteobacteria and δ-proteobacteria at the class level, while arkeya, bacteria, which has high resistance to sterilization and chemical Acidovorax, Luteimonas, and Hydrogenophaga were the leading bac- cleaning, was the pioneering bacteria to attach onto the membrane teria in the genus level. The higher microbial activity in summer was surface and propagated by assimilating the OMs and inorganic nutrients an important cause for the severe membrane fouling. in the RO influent. The bacteria with metabolic processes related to 2. The vertical distribution of the microbial community was found in nitrogen fixation and proteolysis showed higher abundances due to the the membrane fouling layer on the membrane surface. The γ-pro- higher concentration of nitrogen and protein in the RO influent, and the teobacteria was the pioneering bacteria on the membrane surface concentration polarization process was an accelerant for the accumu- and showed the highest abundance in the bottom layer of the lation of bacteria. Thus, the inorganics and organics were also absorbed membrane foulant for its high tolerance to bactericides. to the membrane surface due to their affinity and the microbial as- Pseudoxanthomonas and Sphingopyxis presented a high abundance in similation. Afterwards, the bacteria released SMP and EPS to protect the the bottom layer and Starkeya, Acidovorax, and Hydrogenophaga cells, and biofilms with hierarchical microbial structures formed. The showed a higher abundance in the upper layer. inorganic salts can deposit on the surface of the fouling layer due to the 3. The sterilization and chemical cleaning processes during long-term concentration polarization. During the chemical cleaning process, the operations act as the systematizer for the vertical distribution of bacteria with high affinity and bactericides resistance has a greater bacteria on the membrane fouling layer, while the concentration chance to stay on the membrane surface. That is, long-term operations polarization process acts as a promoter. including sterilization and chemical cleaning acting as the systematizer 4. The foulants on the endcap and entrance of the membrane system for the vertical distribution of bacteria on the membrane fouling layer.

854 L. Zheng et al. Journal of Membrane Science 563 (2018) 843–856

influence the TMP increase; the microbial community of the foulant (EfOM): relating major components of EfOM to their characteristic fouling beha- in the membrane entrance was similar to the upmost layer of the viors, J. Membr. Sci. 349 (2010) 75–82. [22] R.D. Cohen, R. Probstein, Colloidal fouling of reverse osmosis membranes, J. foulant on the membrane surface. Colloid Interface Sci. 114 (1986) 194–207. 5. The key process for RO fouling control was the sterilization process [23] R.Y. Ning, T.L. Troyer, Colloidal fouling of RO membranes following MF/UF in the and chemical cleaning process, especially for the inactivation and reclamation of municipal wastewater, Desalination 208 (2007) 232–237. γ [24] R. Komlenic, Rethinking the causes of membrane biofouling, Filtr. Sep. 47 (2010) removal of -proteobacteria. 26–28. [25] R.A. Al-Juboori, T. Yusaf, Biofouling in RO system: mechanisms, monitoring and Acknowledgments controlling, Desalination 302 (2012) 1–23. [26] J. Vrouwenvelder, S. Manolarakis, J. Van der Hoek, J. Van Paassen, W. Van der Meer, J. Van Agtmaal, H. Prummel, J. Kruithof, M. Van Loosdrecht, Quantitative This work was supported by the National Key Research and biofouling diagnosis in full scale nanofiltration and reverse osmosis installations, Development Program of China (Grant no. 2016YFD0501405) and Water Res. 42 (2008) 4856–4868. Major Science and Technology Program for Water Pollution Control and [27] Y. Khambhaty, J. Plumb, Characterization of bacterial population associated with a brackish water desalination membrane, Desalination 269 (2011) 35–40. Treatment of China (Grant no. 2015ZX07203-005). The authors also [28] H. Bae, D. Jeong, H. Kim, S. Kim, S. Lee, Dynamic shift in community structures of appreciate Tsinghua University and Beijing Institute of Technology for biofilm-forming bacteria by the pre-treatment systems of seawater reverse osmosis – the AFM and FT-IR test and data analysis. processes, Desalination 343 (2014) 17 25. [29] Y.J. Tan, L.J. Sun, B.T. Li, X.H. Zhao, T. Yu, N. Ikuno, K. Ishii, H.Y. Hu, Fouling characteristics and fouling control of reverse osmosis membranes for desalination of Appendix A. Supplementary material dyeing wastewater with high chemical oxygen demand, Desalination 419 (2017) 1–7. [30] H. Ivnitsky, I. Katz, D. Minz, G. Volvovic, E. Shimoni, E. Kesselman, R. Semiat, Supplementary data associated with this article can be found in the C.G. Dosoretz, Bacterial community composition and structure of biofilms devel- online version at http://dx.doi.org/10.1016/j.memsci.2018.06.043. oping on nanofiltration membranes applied to wastewater treatment, Water Res. 41 (2007) 3924–3935. [31] J. Lee, J.Y. Jung, S. Kim, I.S. Chang, S.S. Mitra, I.S. Kim, Selection of the most References problematic biofoulant in fouled RO membrane and the seawater intake to develop biosensors for membrane biofouling, Desalination 247 (2009) 125–136. [1] I. Safari, M.E. Walker, M.-K. Hsieh, D.A. Dzombak, W. Liu, R.D. Vidic, D.C. Miller, [32] B. Wu, S.R. Suwarno, H.S. Tan, L.H. Kim, F. Hochstrasser, T.H. Chong, J. Abbasian, Utilization of municipal wastewater for cooling in thermoelectric M. Burkhardt, W. Pronk, A.G. Fane, Gravity-driven microfiltration pretreatment for power plants, Fuel 111 (2013) 103–113. reverse osmosis (RO) seawater desalination: microbial community characterization [2] J.A. Veil, Use of Reclaimed Water for Power Plant Cooling, Argonne National Lab. and RO performance, Desalination 418 (2017) 1–8. (ANL), Argonne, IL (United States), 2007. [33] C. Ayache, C. Manes, M. Pidou, J.P. Croue, W. Gernjak, Microbial community [3] M.E. Walker, R.B. Theregowda, I. Safari, J. Abbasian, H. Arastoopour, analysis of fouled reverse osmosis membranes used in water recycling, Water Res. D.A. Dzombak, M.K. Hsieh, D.C. Miller, Utilization of municipal wastewater for 47 (2013) 3291–3299. cooling in thermoelectric power plants: evaluation of the combined cost of makeup [34] H. Chu, F. Zhao, X. Tan, L. Yang, X. Zhou, J. Zhao, Y. Zhang, The impact of tem- water treatment and increased condenser fouling, Energy 60 (2013) 139–147. perature on membrane fouling in algae harvesting, Algal Res. 16 (2016) 458–464. [4] A. Kavitha, T. Vasudevan, H.G. Prabu, Evaluation of synthesized antiscalants for [35] T. Miyoshi, T. Tsuyuhara, R. Ogyu, K. Kimura, Y. Watanabe, Seasonal variation in cooling water system application, Desalination 268 (2011) 38–45. membrane fouling in membrane bioreactors (MBRs) treating municipal wastewater, [5] S.R. Pandey, V. Jegatheesan, K. Baskaran, L. Shu, Fouling in reverse osmosis (RO) Water Res. 43 (2009) 5109–5118. membrane in water recovery from secondary effluent: a review, Rev. Environ. Sci. [36] L. Dudley, Membrane autopsies for reversing fouling in reverse osmosis, Membr. Bio/Technol. 11 (2012) 125–145. Technol. 1998 (1998) 9–12. [6] M.A. Sari, S. Chellam, Reverse osmosis fouling during pilot-scale municipal water [37] M. Pontié, S. Rapenne, A. Thekkedath, J. Duchesne, V. Jacquemet, J. Leparc, reuse: evidence for aluminum coagulant carryover, J. Membr. Sci. 520 (2016) H. Suty, Tools for membrane autopsies and antifouling strategies in seawater feeds: 231–239. a review, Desalination 181 (2005) 75–90. [7] L. Malaeb, G.M. Ayoub, Reverse osmosis technology for water treatment: state of [38] K. Chon, S.J. Kim, J. Moon, J. Cho, Combined coagulation-disk filtration process as the art review, Desalination 267 (2011) 1–8. a pretreatment of ultrafiltration and reverse osmosis membrane for wastewater [8] M.L. Pype, M.G. Lawrence, J. Keller, W. Gernjak, Reverse osmosis integrity mon- reclamation: an autopsy study of a pilot plant, Water Res. 46 (2012) 1803–1816. itoring in water reuse: the challenge to verify virus removal–A review, Water Res. [39] J. Kim, F.A. DiGiano, R.D. Reardon, Autopsy of high-pressure membranes to com- 98 (2016) 384–395. pare effectiveness of MF and UF pretreatment in water reclamation, Water Res. 42 [9] T. Wintgens, T. Melin, A. Schäfer, S. Khan, M. Muston, D. Bixio, C. Thoeye, The role (2008) 697–706. of membrane processes in municipal wastewater reclamation and reuse, [40] N. Melián-Martel, J. Sadhwani, S. Malamis, M. Ochsenkühn-Petropoulou, Structural Desalination 178 (2005) 1–11. and chemical characterization of long-term reverse osmosis membrane fouling in a [10] J. Zhang, K. Northcott, M. Duke, P. Scales, S.R. Gray, Influence of pre-treatment full scale desalination plant, Desalination 305 (2012) 44–53. combinations on RO membrane fouling, Desalination 393 (2016) 120–126. [41] F. Tang, H.Y. Hu, L.J. Sun, Q.Y. Wu, Y.M. Jiang, Y.T. Guan, J.J. Huang, Fouling of [11] S. Jiang, Y. Li, B.P. Ladewig, A review of reverse osmosis membrane fouling and reverse osmosis membrane for municipal wastewater reclamation: autopsy results control strategies, Sci. Total Environ. 595 (2017) 567–583. from a full-scale plant, Desalination 349 (2014) 73–79. [12] M. Asadollahi, D. Bastani, S.A. Musavi, Enhancement of surface properties and [42] T. Yu, L. Meng, Q.B. Zhao, Y. Shi, H.Y. Hu, Y. Lu, Effects of chemical cleaning on RO performance of reverse osmosis membranes after surface modification: a review, membrane inorganic, organic and microbial foulant removal in a full-scale plant for Desalination 420 (2017) 330–383. municipal wastewater reclamation, Water Res. 113 (2017) 1–10. [13] Y. Shim, H.-J. Lee, S. Lee, S.-H. Moon, J. Cho, Effects of natural organic matter and [43] F. Tang, H.Y. Hu, L.J. Sun, Y.X. Sun, N. Shi, J.C. Crittenden, Fouling characteristics ionic species on membrane surface charge, Environ. Sci. Technol. 36 (2002) of reverse osmosis membranes at different positions of a full-scale plant for muni- 3864–3871. cipal wastewater reclamation, Water Res. 90 (2016) 329–336. [14] A. Zularisam, A. Ismail, R. Salim, Behaviours of natural organic matter in mem- [44] X. Wang, F. Wei, W. Qi, Water and Wastewater Monitoring and Analysis Methods, brane filtration for surface water treatment-a review, Desalination 194 (2006) China environmental science press, Beijing, China, 2002. 211–231. [45] J.G. Caporaso, C.L. Lauber, W.A. Walters, D. Berg-Lyons, C.A. Lozupone, [15] A. Jawor, E.M. Hoek, Effects of feed water temperature on inorganic fouling of P.J. Turnbaugh, N. Fierer, R. Knight, Global patterns of 16S rRNA diversity at a brackish water RO membranes, Desalination 235 (2009) 44–57. depth of millions of sequences per sample, Proc. Natl. Acad. Sci. USA 108 (2011) [16] M. Goosen, S. Sablani, H. Al-Hinai, S. Al-Obeidani, R. Al-Belushi, D. Jackson, 4516–4522. Fouling of reverse osmosis and ultrafiltration membranes: a critical review, Sep. Sci. [46] T. Magoč, S.L. Salzberg, FLASH: fast length adjustment of short reads to improve Technol. 39 (2005) 2261–2297. genome assemblies, Bioinformatics 27 (2011) 2957–2963. [17] H.J. Lee, H.E. Kim, C. Lee, Combination of cupric ion with hydroxylamine and [47] R.C. Edgar, B.J. Haas, J.C. Clemente, C. Quince, R. Knight, UCHIME improves hydrogen peroxide for the control of bacterial biofilms on RO membranes, Water sensitivity and speed of chimera detection, Bioinformatics 27 (2011) 2194–2200. Res. 110 (2017) 83–90. [48] Q. Wang, G.M. Garrity, J.M. Tiedje, J.R. Cole, Naive Bayesian classifier for rapid [18] S. Jeong, S.J. Kim, L.H. Kim, M.S. Shin, S. Vigneswaran, T.V. Nguyen, I.S. Kim, assignment of rRNA sequences into the new bacterial taxonomy, Appl. Environ. Foulant analysis of a reverse osmosis membrane used pretreated seawater, J. Microbiol. 73 (2007) 5261–5267. Membr. Sci. 428 (2013) 434–444. [49] X. Liu, L. Jungang, Z. Qianya, F. Jinlai, L. Yingli, S. Jingxin, The analysis and [19] M. Yan, D. Wang, G.V. Korshin, M.F. Benedetti, Quantifying metal ions binding onto prediction of scale accumulation for water-injection pipelines in the Daqing dissolved organic matter using log-transformed absorbance spectra, Water Res. 47 Oilfield, J. Pet. Sci. Eng. 66 (2009) 161–164. (2013) 2603–2611. [50] S.H. You, D.H. Tseng, G.L. Guo, A case study on the wastewater reclamation and [20] S. Jeong, G. Naidu, R. Vollprecht, T. Leiknes, S. Vigneswaran, In-depth analyses of reuse in the semiconductor industry, Resour., Conserv. Recycl. 32 (2001) 73–81. organic matters in a full-scale seawater desalination plant and an autopsy of reverse [51] M.T. Khan, M. Busch, V.G. Molina, A.H. Emwas, C. Aubry, J.P. Croue, How different osmosis membrane, Sep. Purif. Technol. 162 (2016) 171–179. is the composition of the fouling layer of wastewater reuse and seawater desali- [21] Y. Zhao, L. Song, S.L. Ong, Fouling of RO membranes by effluent organic matter nation RO membranes? Water Res. 59 (2014) 271–282.

855 L. Zheng et al. Journal of Membrane Science 563 (2018) 843–856

[52] H.L. Yang, C. Huang, J.R. Pan, Characteristics of RO foulants in a brackish water [56] H.Z. Shafi, A. Matin, S. Akhtar, K.K. Gleason, S.M. Zubair, Z. Khan, Organic fouling desalination plant, Desalination 220 (2008) 353–358. in surface modified reverse osmosis membranes: filtration studies and subsequent [53] W. Chen, P. Westerhoff, J.A. Leenheer, K. Booksh, Fluorescence excitation-emission morphological and compositional characterization, J. Membr. Sci. 527 (2017) matrix regional integration to quantify spectra for dissolved organic matter, 152–163. Environ. Sci. Technol. 37 (2003) 5701–5710. [57] W. Zhang, B. Cao, D. Wang, T. Ma, H. Xia, D. Yu, Influence of wastewater sludge [54] R. Bi, Q. Lu, T. Yuan, S. Zhou, Y. Yuan, Y. Cai, Electrochemical and spectroscopic treatment using combined peroxyacetic acid oxidation and inorganic coagulants re- characteristics of dissolved organic matter in a forest soil profile, J. Environ. Sci. 25 flocculation on characteristics of extracellular polymeric substances (EPS), Water (2013) 2093–2101. Res. 88 (2016) 728–739. [55] L. Li, M. Liu, M. Wu, C. Jiang, X. Chen, X. Ma, J. Liu, W. Li, X. Tang, Z. Li, Effects of [58] K. Heylen, L. Lebbe, P. De Vos, Acidovorax caeni sp. nov., a denitrifying species duckweed (Spriodela polyrrhiza) remediation on the composition of dissolved or- with genetically diverse isolates from activated sludge, Int. J. Syst. Evolut. ganic matter in effluent of scale pig farms, J. Environ. Sci. 55 (2017) 247–256. Microbiol. 58 (2008) 73–77.

856