Kai
Li
*ab,
Shu
Li
ab,
Ce
Sun
ab,
Tinglin
Huang
*ab,
Guibai
Li
c and
Heng
Liang
c
aKey Laboratory of Northwest Water Resource, Environment and Ecology, MOE, Xi'an University of Architecture and Technology, Xi'an, 710055, PR China. E-mail: likai@xauat.edu.cn; likai@xauat.edu.cn; sisl2018@163.com; wandou1996@126.com; huangtinglin@xauat.edu.cn; Tel: +86 29 82207886
bShaanxi Key Laboratory of Environmental Engineering, Xi'an University of Architecture and Technology, Xi'an, 710055, PR China. Tel: +86 29 82201038
cState Key Laboratory of Urban Water Resource and Environment (SKLUWRE), Harbin Institute of Technology, Harbin 150090, PR China. E-mail: liguibai@vip.163.com; hitliangheng@163.com
First published on 4th November 2019
The integration of adsorption with ultrafiltration (UF) is a promising technology for the production of high quality drinking water, but it is still under debate whether the addition of an adsorbent can mitigate membrane fouling. This study was conducted to investigate the impact of properties of the adsorbent and natural organic matter (NOM) in feed water on membrane fouling in an integrated adsorption–UF system. A continuous-flow hollow fiber UF set-up with periodic backwash and aeration was used to imitate the operation of full-scale integrated adsorption–UF systems. Two types of adsorbents, commercial powdered activated carbon (PAC) and homemade mesoporous adsorbent resin (MAR), were examined. As for Songhua River water dominated by allochthonous NOM, membrane fouling of the PAC/UF system was increased by 39.5% compared with that of UF alone, whereas the addition of MAR mitigated membrane fouling by 66.4%. For the synthetic water composed of algal organic matter (AOM), membrane fouling of PAC/UF and MAR/UF systems was reduced by 40.7% and 83.3%, respectively. The results suggested that deposition of adsorbent particles and fouling resistance of the cake layer were determined by the hydrophobicity and molecular weight distribution of NOM as well as the properties of adsorbents. This study highlights the significance of adsorbent selection for the integrated adsorption–UF process according to the source and properties of NOM.
Water impactThe integrated adsorption–UF process is a promising technology for the production of high quality drinking water. In spite of periodic air sparging and backwashing, a dense cake layer can still form on the membrane and result in significant irreversible fouling in some cases, which could be avoided by careful adsorbent selection according to the source and properties of NOM. |
Adsorption can be combined with UF in several configurations, including an adsorption unit placed prior to UF, integration of an adsorption unit with UF, and an adsorption unit following UF.11 For the integrated adsorption–UF process, powdered adsorbents are added to the feed water of UF, and are retained by the membrane, with direct contact between adsorbents and the membrane. In this configuration, the adsorbents can be completely retained by the membrane, and high concentrations of adsorbents can be maintained prior to UF to enhance the removal of contaminants.15,16 Besides, for a submerged UF membrane, the footprint of the process can be reduced by using the membrane tank as the adsorption contactor.17 Therefore, integrated adsorption–UF is a promising technology due to its high purification efficiency and compact footprint.18
However, for the integrated adsorption–UF process, the participation of adsorbent particles makes membrane fouling more complicated.19,20 Although many studies demonstrated that adsorbent particles exerted minor influence on pure water permeability due to their larger size compared with the membrane pore size,21,22 the reported results regarding the impact of adsorbent particles on membrane fouling were controversial. Taking the most common adsorbent in water treatment, powdered activated carbon (PAC), as an example, several studies proved that PAC alleviated organic fouling by removing some NOM,23,24 but many studies indicated that the addition of PAC resulted in serious cake layer fouling and exacerbated membrane fouling ultimately.25,26 Actually, the influence of adsorbent addition on membrane fouling can be regarded as a trade-off between the decrease of organic fouling and the increase of external fouling related to adsorbent particles.20 Adsorbent particles may result in cake/gel layer fouling in combination with organics, and the structure and resistance of the cake/gel layer are closely related to the properties of NOM and the adsorbent.19
Moreover, deposition of adsorbent particles on the membrane surface depends on the configuration and hydrodynamic conditions of the membrane system.27 Many studies regarding the combined fouling of adsorbent particles and organics were conducted using flat-sheet membranes in laboratory filtration cells in dead-end mode or with stirring to simulate the shear stress on the membrane surface.12,28,29 However, in practical application of the integrated adsorption–UF system, hollow fiber membrane systems are extensively utilized, and cross-flow or air sparging is usually applied in pressurized or submerged systems, respectively, to provide shear stress and prevent deposition of particles on the membrane surface.30,31 It is hard to predict to what extent periodic backwashing and cross-flow/air sparging would impact the deposition of adsorbent particles and NOM based on bench-scale short-term filtration tests using flat-sheet membranes. Therefore, it is necessary and meaningful to investigate membrane fouling in integrated adsorption–UF systems using membrane systems mimicking the operation conditions in full scale systems.
Mesoporous adsorbent resin (MAR) is an adsorbent specially developed for membrane fouling control as reported by Clark et al.32 It was made from polymeric membrane materials (e.g., polysulfone) through a process analogous to the non-solvent induced phase separation process. It has been reported to be selective for the adsorption of organic foulants.33,34 In our previous study,19 it has been proved that the structure and resistance of the fouling layer formed by MAR and NOM model foulants during dead-end filtration were different from those formed by PAC and model foulants. The main objective of this study was to obtain a deeper understanding of the effects of characteristics of the adsorbent and NOM on membrane fouling in the integrated adsorption–UF system. A continuous-flow hollow fiber UF set-up with periodic backwashing and aeration was used to imitate the operation of full-scale integrated adsorption–UF systems. Two types of adsorbents, MAR and PAC, and two water samples dominated by allochthonous and autochthonous NOM, respectively, were examined. TMP evolution, fouling resistance distribution and adsorbent particle deposition were quantified, and the roles of adsorbent and NOM properties were highlighted.
A lab-scale submerged hollow fiber UF system with automatic periodic backwashing and aeration was employed to simulate the operational procedure and hydrodynamic conditions in full-scale systems (Fig. 1). The filtration cycle consisted of 118 min filtration and 2 min backwashing, while the aerator was turned on for 0.5 min every 5 min. The filtration flux was kept at 30 L m−2 h−1 using a peristaltic pump (BT-100, Longer Pump, Baoding, China), and backwashing was conducted by reverse rotation of the peristaltic pump. A pressure transducer (PTP708, Tuopu Electric, Foshan, China) was mounted between the membrane module and the peristaltic pump to monitor the trans-membrane pressure (TMP). The transducer was connected to a computer, and the data were automatically logged every five seconds. An aeration intensity of 35 m3 m−2 h−1 (calculated based on the bottom area of the membrane tank) was used to prevent the sedimentation of adsorbent particles at the bottom of the membrane tank. The sludge was discharged every six filtration cycles and a fresh adsorbent was added at the beginning of each sludge discharge cycle at a dose of 50 mg L−1 based on the volume of inflow water in the whole cycle. The system was operated in constant flux mode with automatic backwashing and periodic aeration, and therefore the membrane fouling and adsorbent particle deposition can be regarded as hydraulically irreversible from the perspective of full-scale operation.
At the end of the experiment, the remaining fouling resistance was distinguished as external and internal fouling by determining the filtration resistance before and after wiping the membrane surface with a wet sponge.35 Briefly, after emptying the membrane tank and refilling with pure water, filtration for 5 min was conducted, and the average of TMP values was recorded as TMP1. Then the membrane module was taken out and the surface of each fiber was wiped with a wet sponge carefully. Afterwards, the module was reinstalled and filtration of pure water was carried out again with the average of TMP values recorded as TMP2. Therefore, with the TMP of the new membrane module denoted as TMP2, the resistance of external fouling (Ref) and internal fouling (Rif) can be calculated by eqn (1) and (2), respectively.
Ref = (TMP1 − TMP2)/(μ·J) | (1) |
Rif = (TMP2 − TMP0)/(μ·J) | (2) |
Parameters | Unit | Songhua River water | AOM-containing water |
---|---|---|---|
pH | — | 7.8 ± 0.2 | 7.5 ± 0.1 |
Turbidity | NTU | 6.35 ± 0.24 | 1.53 ± 0.17 |
TOC | Mg L−1 | 6.2 ± 0.4 | 3.0 ± 0.4 |
DOC | Mg L−1 | 5.9 ± 0.3 | 2.9 ± 0.4 |
UV254 | Cm−1 | 0.161 ± 0.005 | 0.042 ± 0.004 |
SUVA | L (mg−1 m−1) | 2.62 ± 0.03 | 1.38 ± 0.02 |
MAR was made from polyethersulfone (Veradel 3000P, Solvay, USA) by following the method proposed by Clark, et al.32 Wood-based PAC purchased from Bench Chemicals Co. (Tianjin, China) was used without further purification. The particle size and zeta potentials of the two adsorbents were measured using a laser particle size analyzer (MasterSizer 2000, Malvern, UK) and a zeta potential analyzer (Zetasizer Nano S90, Malvern, UK), respectively. The surface areas and pore size distributions of MAR and PAC were measured using a surface area and porosity analyzer (ASAP 2020, Micromeritics, USA) with the N2 adsorption method. The amounts of surface functional groups were determined by Boehm titration. The physical and chemical properties of MAR and PAC are shown in Table S1 in the ESI.† The FTIR spectra of MAR and PAC were acquired by using an infrared spectrometer (Nicolet iS50, Thermo Scientific, MA, USA), and the results are presented in Fig. S1 in the ESI.† It can be seen that MAR and PAC have similar particle sizes (25.2 versus 32.1 μm) and surface zeta potentials (−22.4 versus −23.9 mV), but have distinct specific surface areas and pore structures. The surface area of PAC (1219 m2 g−1) was much larger than that of MAR (108 m2 g−1), and the pore structures of MAR and PAC were dominated by mesopores and micropores, respectively.
Fig. 2 TMP evolution in integrated adsorption–UF systems during treatment of Songhua River water (a) and AOM-containing water (b). |
For Songhua River water, both the TMP increase within a filtration cycle and the irreversible TMP increase among filtration cycles were significantly mitigated by the addition of MAR. In contrast, the addition of PAC exerted minor influence on the TMP increase in the first two filtration cycles, but it increased irreversible fouling significantly in the following cycles. At the end of the experiment, compared with UF alone, the irreversible TMP increase of MAR/UF was reduced by 66.4%, whereas that of PAC/UF was increased by 39.5%.
As for AOM-containing water, the TMP increase of both PAC/UF and MAR/UF systems was lower than that of UF alone, and MAR controlled membrane fouling more efficiently. Based on the irreversible TMP increase at the end of the experiment, the addition of MAR and PAC reduced membrane fouling by 83.3% and 40.7%, respectively.
Fig. 3 Distribution of irreversible fouling resistance at the end of the experiment for Songhua River water (a) and AOM-containing water (b) (n = 3). |
The amount of adsorbent particles suspended in the membrane tank and deposited on the membrane surface at the end of the experiment was determined by a gravimetric method and mass balance, and the ratios are shown in Fig. 4. For Songhua River water, although there was periodic backwashing and aeration, more than 90% of PAC deposited on the membrane surface, while more than 60% of MAR was suspended in the membrane tank. The results were consistent with the significantly higher external fouling for PAC/UF (Fig. 3(a)). With respect to AOM-containing water, though the percentage of deposited PAC was still a little higher than that of MAR, it was much lower than that for Songhua River water.
Fig. 4 Distribution of adsorbent particles at the end of the experiment for Songhua River water (a) and AOM-containing water (b) (n = 3). |
The fluorescence EEM spectra of NOM in Songhua River water and AOM-containing water are shown in Fig. 5. The NOM in Songhua River water exhibited four major peaks: peaks A and C representing humic-like substances and peaks T1 and T2 which represent protein-like substances. The intensity of peak C was much higher than the other peaks, suggesting the predominance of allochthonous humic substances.38 In previous studies, it has been demonstrated that the mixture of PAC particles and humic substances can form a dense cake layer on the membrane surface, while the cake layer composed of MAR and humic substances was much porous.19,20 This study suggested that the dense cake layer formed by PAC and humic substances cannot be readily removed by backwashing and shear stress induced by aeration. For AOM-containing water, the intensity of peak T1 was much higher than those of peaks A and C, indicating the dominance of autochthonous proteins.39 Although polysaccharides cannot be detected by EEM, they are usually concomitant with proteins in AOM.36,40 Compared with Songhua River water, the amount of deposited PAC was much lower for AOM-containing water, probably due to the lower hydrophobicity of autochthonous proteins and polysaccharides in comparison with allochthonous humic substances.40
The MW distributions of NOM in Songhua River water and AOM are shown in Fig. 6. It can be observed that most of the NOM in Songhua River water was smaller than 30 kDa, and fractions with MW 10–30 kDa, 3–10 kDa and <3 kDa accounted for 30.2%, 27.2% and 37.3% in terms of DOC, respectively. As shown in Fig. 5(a), the NOM in Songhua River was dominated by humic substances, and therefore the MW distribution in terms of UV254 was in accordance with that based on DOC. As for AOM, the MW distribution exhibited a typical bimodal mode based on DOC, and the MW of more than 47% AOM was higher than 100 kDa. As demonstrated by previous studies,36,40 the high-MW fraction in AOM was mainly attributed to autochthonous proteins and polysaccharides, which was consistent with the result of fluorescence EEM. But the MW distribution in terms of UV254 was distinct from that based on DOC, and most of the UV-absorbing molecules in AOM were smaller than 3 kDa. Because UV-absorption was mainly caused by humic substances, rather than proteins and polysaccharides,41 it can be inferred that most of the humic substances in AOM were smaller than 3 kDa.
Fig. 6 MW distribution of NOM in Songhua River water and AOM-containing water: (a) ratio based on DOC and (b) ratio based on UV254 (n = 3). |
The hydrophobicity distributions of NOM in Songhua River water and AOM are shown in Fig. 7. For both NOM in Songhua River water and AOM, the hydrophilicity distributions based on DOC and UV254 were analogous, suggesting that the hydrophilicity distribution of the UV-absorbing fraction was consistent with that of the non-UV-absorbing fraction. The hydrophobic (HPO) fraction accounted for more than 65% of NOM in Songhua River water in terms of DOC, which was consistent with the nature of allochthonous humic substances as shown in the fluorescence EEM spectra in Fig. 5(a). As for AOM, the hydrophilic (HPI) fraction accounted for 59% in terms of DOC, followed by the HPO fraction accounting for 36%, which was in accordance with the predominance of autochthonous proteins as shown in the fluorescence EEM spectra in Fig. 5(b). In short, there was more HPO organic fraction in Songhua River water than in AOM-containing water.
Based on the analysis of fluorescence EEM, MW distribution and hydrophobicity distribution, it can be concluded that the composition and properties of NOM in the two feed water samples were quite different. The NOM in Songhua River water was dominated by humic substances with low to medium MW and strong hydrophobicity, while there was more HPI fraction in AOM and the MW of most humic substances in AOM was less than 3 kDa. With respect to the properties of adsorbents, the key distinction was the pore structure, and the average pore size of MAR (16.4 nm) was much larger than that of PAC (2.2 nm) (Table S1†). The discrepant influence of MAR and PAC on fouling in the integrated adsorption–UF system for different feed water samples can be explained by the properties of NOM and adsorbents. For Songhua River water, organic molecules with various MWs as well as organic colloids can be removed by entering the internal pores of MAR, and the cake layer formed by MAR particles was porous because there were less organics in the gap between adsorbent particles.19 In contrast, the organics with medium and high MWs as well as organic colloids cannot enter the micropores of PAC due to size exclusion, but their interactions with PAC particles were strong due to the predominance of the HPO fraction, making a dense cake layer composed of adsorbent particles and organics on the membrane surface.26 Therefore, large amounts of PAC particles deposited on the membrane surface and resulted in irreversible fouling, whereas MAR particles deposited on the membrane surface did not lead to obvious cake fouling. With respect to AOM, much less PAC particles were deposited on the membrane surface and the cake layer resistance was small because there was more HPI fraction in AOM and the interactions between the high-MW fraction (i.e., proteins and polysaccharides) and PAC particles were weak.33,40 Meanwhile, the addition of MAR mitigated fouling better because its mesoporous structure favored more removal of high-MW organics.33,34 Overall, in the integrated adsorption–UF system, although periodic air sparging and backwashing can prevent particle deposition to some extent, a dense cake layer composed of adsorbent particles and organics can still form and result in significant irreversible fouling in some cases, which could be avoided by careful adsorbent selection according to the source and properties of NOM.
It should be noted that PAC was more efficient than MAR in reducing DOC concentration in membrane permeate (Fig. S2†), which can be attributed to the much higher specific surface area of PAC compared with MAR (Table S1†). Meanwhile, the mesoporous structure of MAR was the major reason why MAR outperformed PAC in organic fouling control. Therefore, taking into account the mitigation of membrane fouling and the decrease of permeate DOC concentration at the same time, developing novel adsorbents with high specific surface area and abundant large mesopores (10 < w < 50 nm) is of great significance for the integrated adsorption-UF process.
For surface water dominated by allochthonous NOM (Songhua River water), the addition of PAC exacerbated membrane fouling due to the formation of a dense cake layer composed of PAC particles and hydrophobic organics. In contrast, less MAR particles were deposited on the membrane surface and the cake layer was much more porous due to its mesoporous structure. For the AOM-containing water, the MW and hydrophobicity distribution of AOM prevented the formation of a dense PAC cake layer. The results suggested that the deposition of adsorbent particles and the fouling resistance of the cake layer were influenced by the hydrophobicity and MW distribution of NOM as well as the properties of adsorbents. This study highlights the significance of adsorbent selection for the integrated adsorption–UF process according to the source and properties of NOM.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c9ew00843h |
This journal is © The Royal Society of Chemistry 2020 |