Membrane cleaning in membrane distillation of reverse osmosis concentrate generated in landfill leachate treatment

As a thermally induced membrane separation process, membrane distillation (MD) has drawn more and more attention for the advantages of treating hypersaline wastewaters, especially the concentrate from reverse osmosis (RO) process. One of the major obstacles in widespread MD application is the membrane fouling. We investigated the feasibility of direct contact membrane distillation (DCMD) for landfill leachate reverse osmosis concentrate (LFLRO) brine treatment and systematically assessed the efficiency of chemical cleaning for DCMD after processing LFLRO brine. The results showed that 80% water recovery rate was achieved when processing the LFLRO brine by DCMD, but the membrane fouling occurred during the DCMD process, and manifested as the decreasing of permeate flux and the increasing of permeate conductivity. Analysis revealed that the serious flux reduction was primarily caused by the fouling layer that consist of organic matters and inorganic salts. Five cleaning methods were investigated for membrane cleaning, including hydrogen chloride (HCl)-sodium hydroxide (NaOH), ethylene diamine tetraacetic acid (EDTA)-NaOH, critic acid, sodium hypochlorite (NaClO) and sodium dodecyl sulphate (SDS) cleaning. Among the chemical cleaning methods investigated, the 3 wt.% SDS cleaning showed the best efficiency at recovering the performance of fouled membranes.


INTRODUCTION
Sanitary landfill is one of the most common methods for the ultimate disposal of municipal solid wastes (MSW) nowadays due to its economic advantages, and the MSW generation is increasing due to the growing population and industrialization processes. Thus, the leachate released from sanitary landfills has also increased and become the subject of interest as highly contaminated wastewater. Landfill leachate is defined as a mixture of rainwater percolation through wastes, water produced from wastes by a series of physical, hydrolytic and fermentative degradation and the inherent water content of wastes (Renou et al. 2008). Landfill leachate contains high concentration of organic and inorganic contaminants, including humic acid, ammonia nitrogen, heavy metals, and inorganic salts. With the increasingly strict discharge standard of landfill leachate, the treatments based on membrane technology become the feasible alternative to the conventional physico-chemical methods of landfill leachate treatment (Li et al. 2010). The reverse osmosis (RO) was more and more used in recent years due to its great advantages in inorganic salts rejection and water recovery (Ince et al. 2010;Kuusik et al. 2014;Rukapan et al. 2015).
During the RO process, a kind of brown solution will be produced continuously -the landfill leachate reverse osmosis concentrate (LFLRO) brine, which represents typically 13-30% of total landfill leachate entering the treatment system and contains high concentration of refractory organic and inorganic salt contaminants (Van der Bruggen et al. 2003). Discharge of LFLRO brine without effective treatment would cause substantial environment pollution. Currently, common treatment methods for LFLRO brine include further treatment to remove contaminants (Labiadh et al. 2006;Wang et al. 2016;Fernandes et al. 2017;Mojiri et al. 2017;Chen et al. 2019), thermal incineration or drying (Ye et al. 2017;Bai et al. 2021) and re-infiltration into the landfills (Calabro et al. 2010). However, there are many limitations and disadvantages to these methods. For example, recycling back to landfills will increase leachate salinity and cause negative effects on the leachate treatment system. During mechanical vapor recompression (MVR), the high salinity may lead to severe scaling of treatment facilities and reduce the heat transfer efficiency.
Membrane distillation (MD) is an emerging technology that combines thermally driven distillation and membrane separation and has shown great promise in the treatment of hypersaline solutions (Khayet et al. 2011). The MD process may be used as a substitute for conventional separation processes such as multistage vacuum evaporation, RO, and distillation (Lawson & Lloyd 1997). Compared to MVR and incineration, MD process requires lower operating temperature, vapor space and energy consumption (Alkhudhiri et al. 2012). Besides, unlike RO, MD is hardly affected by osmotic pressure, thus it can further recover pure water from RO brine. Other advantages of MD include potential for 100% rejection of nonvolatile solutes and compact configuration (Ding et al. 2006a(Ding et al. , 2006bKhayet 2011;Chung & Wang 2015). In recent years, there have been extensive studies on the application of MD in water desalination and treatment (Gryta 2005(Gryta , 2015Martinetti et al. 2009;Hickenbottom & Cath 2014;Chew et al. 2019). Thus, MD might be an effective method for the further treatment of LFLRO brine. However, membrane fouling and membrane wetting are the major technical challenges, which can cause several negative consequences on MD performance, including flux reduction and salt leakage (Zou et al. 2018;Chang et al. 2021).
To achieve long stable operation of MD process in LFLRO brine treatment, effective membrane cleaning methods will be necessary. The potential pollutants in LFLRO brine that may cause membrane fouling include inorganic and organic matters. For inorganic pollutants, acids such as hydrogen chloride (HCl), oxalic acid and citric acid have been proven to achieve effective removing (Gryta 2008). Elena et al. reported that the mixed cleaning solution composed of 0.1 wt.% oxalic acid and 0.8 wt.% citric acid can remove the sodium chloride (NaCl) scale layer that covered the fouled membranes (Guillen-Burrieza et al. 2014). Besides, ethylene diamine tetraacetic acid (EDTA) forms strong complex with Ca 2þ , thus can be used to remove Ca 2þ precipitation on the membrane. Peng et al. used five cleaning agents to washing calcium sulfate (CaSO 4 ) deposits on the fouled membrane and found that the flux recovery rate after cleaning followed an order of ethylene diamine tetraacetic acid tetrasodium (EDTA-4Na) . NaCl . citric acid . NaOH . KCOOH (Peng et al. 2015). The pollutants in LFLRO brine are more complicated than pure inorganic salts ( Van der Bruggen et al. 2003), especially the organic pollutants that may cause serious membrane fouling in MD process. Thus, it is necessary to consider the removal of organic pollutants.
For organic pollutants, sodium hypochlorite (NaClO) has been proved to be an effective cleaning solution (Paugam et al. 2010;Porcelli & Judd 2010;Lee et al. 2013). Due to its oxidizability, NaClO can react with the foulants on membrane surface (Cai & Liu 2016). Besides, Wang et al. found that large organic colloids can be disintegrated into fine particles and/or soluble organic matters at caustic conditions (Wang et al. 2014). More important, Ca 2þ can greatly enhance organic matter fouling by complexation and subsequent formation of intermolecular bridges among organic foulant molecules (Li & Elimelech 2004;Srisurichan et al. 2005), thus, breaking the Ca 2þ -organic complexation is also important for the removal of organic pollutants (Li & Elimelech 2004). Study showed that sodium dodecyl sulphate (SDS) and ethylene diamine tetraacetic acid (EDTA) were effective in disrupting the complexes formed by the organic foulants with Ca 2þ when cleaning organic fouled membranes, including RO, nanofiltration (NF) and forward osmosis (FO) membranes (Li & Elimelech 2004;Ang et al. 2006Ang et al. , 2011Beyer et al. 2010;Wang et al. 2015).
Given the absence of cleaning methods for direct contact membrane distillation (DCMD) treatment of LFLRO brine. The objective of this study was to investigate the feasibility of DCMD for LFLRO brine treatment and evaluate the cleaning efficiency of different cleaning methods. First, during the DCMD treatment of LFLRO, the permeate flux and conductivity variations were studied. In addition, the membrane fouling was also investigated. For the membrane cleaning, five cleaning methods were evaluated and the effects of each membrane cleaning method on membrane performance were analyzed in detail.

Water source
The municipal landfill considered in this study locates in the north of Beijing, China, which has been in operation since 1996. Leachate treatment process flow diagram for this plant is shown in Figure 1, the treatment system contains biological treatment process and advanced membrane separation processes. In membrane separation processes, landfill leachate is pumped to the first stage RO membrane module units (RO1) after ultrafiltration. The RO1 permeate is fed to the second stage RO membrane module units (RO2) for purifying and the RO2 permeate is discharged to Wenyu River at last. Concentrate from the RO2 units is fed back to RO1 units and mixed with landfill leachate as the feed solution of RO1, and the RO1 concentrate is sent to the concentrate tank for further treatment. The recovery rate of the RO1 units is 50-80%, which is adjusted in accordance with total salinity of the inlet water. The RO2 units are operated under the recovery rate of 80-90%. The total recovery rate of the whole landfill leachate treatment system is in the range of 70% to 85%. The chemical physical characteristics of the RO concentrate are listed in Table 1.

Chemical cleaning solutions
HCl (pH ¼ 2), NaOH (pH ¼ 12), EDTA-Na (pH ¼ 11.5), 2 wt.% citric acid, 2 wt.% NaClO and 3 wt.% SDS were used as membrane cleaning agents. All the chemicals were certified analytical reagent grade and supplied by Sinopharm Chemical Reagent Co., Ltd China. The chemical cleaning solutions were prepared by dissolving the chemical in deionized water without further purification.

Membrane and membrane module
The membrane fabricated from polyvinylidene fluoride (PVDF)/N, N-Dimethylacetamide/ lithium chloride (LiCl)/ethylene glycol (12/80/5/3 wt. %) dope via phase inversion process was chosen to fabricate membrane modules. The hydrophobic hollow fibers in the number of 50 pieces were assembled into a polyester tube (diameter d in /d out ¼ 15/20 mm/mm) with two unplasticized polyvinyl chloride (UPVC) T-tubes and two ends of the bundle of fibers were sealed with solidified epoxy resin to compose a membrane module. The effective membrane length was 100 mm for each membrane module.

Uncorrected Proof
The effective membrane length and the total membrane area were 100 mm and 125.6 cm 2 for each membrane module, respectively. The characteristics of the membrane and membrane module are presented in Table 2.

Membrane distillation setup
The DCMD experimental setup is schematically shown in Figure 2. The feed wastewater heated by heater, which was stirred continually by a magnetic stirrer, flowed through the lumen side of the hollow fibers, and the cold distillate flowed through   Uncorrected Proof the shell side. Both solutions were circulated in the membrane module with the help of two magnetic pumps (MP-15RN, Shanghai Seisun Pumps, China). The feed temperature was controlled by a Pt-100 sensor and a heater connected to an external thermostat (XMTD-2202, Yongshang Instruments, China). The distillate temperature was adjusted through a spiral glass heat exchanger immersed in the constant temperature trough of the cooler (SDC-6, Nanjing Xinchen Biotechnology, China). The temperature of both fluids was monitored at the inlet and outlet of the membrane module using four Pt-100 thermoresistances connected to a digital meter (Digit RTD, model XMT-808, Yuyao Changjiang Temperature Meter Instruments, China) with an accuracy of +0.1°C. An electric conductivity monitor (CM-230A, Shijiazhuang Create Instrumentation Technologies, China) was used to monitor the distillate conductivity.

Membrane fouling experiments
Fouling tests were carried out using the LFLRO brine as feed. The feed and the distillate flowed co-currently through the membrane module, and the circulation feed rate was fixed at 45 L h À1 , while the cold side was set at 27 L h À1 . The feed temperature was fixed at 53°C and the distillate temperature kept at constant 20°C. The initial volumes of the feed and the distillate were 2.5 L and 0.25 L, respectively. During membrane fouling tests, there was no make-up water added into the feed tank, which meant that the feed was concentrated gradually. The permeate flux was calculated by the following equation: where J is the permeate flux (kg m À2 h À1 ), ΔW is the mass of the permeate (kg), A is the effective area of the hollow fiber membranes (m 2 ) and Δt is the time interval (h). The concentration factor K can be calculated by the following equation: where Q o and Q p are the initial quantity of feed (kg) and the cumulative permeate production (kg), respectively.

Membrane cleaning experiments
The total operating time for each DCMD experiment of LFLRO brine was 30 hours. Afterwards, the membrane modules were retrieved and cleaned separately according to the following four cleaning procedures ( (1) Deionized water cleaning: the membranes were rinsed with deionized water for 30 min to remove the loose deposits on the membrane surface.
(2) Chemical cleaning: the membranes were cleaned independently with different cleaning methods at ambient temperature.
The cleaning solution was pumped into the membrane module and flowed through the lumen side of the hollow fibers with the circulation flow rate at 45 L h À1 . The chemical solutions and cleaning schemes for five cleaning methods used in chemical cleaning step are presented in Table 3. (3) Deionized water flushing: the membranes were flushed with deionized water for 1.0 h at 50°C followed by chemical cleaning. The hot deionized water flowed co-currently through the membrane module, the flow rate was 45 L h À1 in lumen side of the hollow fibers. At the completion of membrane flushing, the membrane module was removed from Uncorrected Proof experimental setup and excess liquid on membrane surface was allowed to drain off by gently tilting the membrane module. (4) Dry-out: the membranes were dried in air blowing drying cabinet for 2.0 h at 50°C to remove the residual liquid on membrane surface and in membrane pores.
To evaluate the efficiencies of different cleaning methods, the DCMD experiments were carried out using the same membrane module, the 4 wt.% NaCl solution was used as the feed and operating parameters (i.e., initial volumes of feed and distillation, circulation rates and operating temperatures) were in accordance with LFLRO treatment experiment.

Membrane surface analysis
Membrane morphology was investigated with a HITACHI S-3000N scanning electron microscope (SEM) (Hitachi Ltd, Japan). Membrane samples were frozen in liquid nitrogen, fractured to obtain fragments, and sputtered with gold using a HITACHI E-1010 Ion Sputtering device for SEM observation. Membrane contact angles were tested using a contact angle measurement instrument (OCA 15EC, Dataphysics, GER). The membrane samples fouled by LFLRO brine were handled gently and without any excessive forces to ensure that the fouling layer remained intact.

Analytical methods
The total dissolved solids (TDS) (GB 11901-1989), chemical oxygen demand (COD) (HJ 828-2017) and ammonia nitrogen (NH 4 þ -N) (HJ 535-2009) were measured according to the standard methods. The pH was measured using a pH meter (Five Easy, MMETTLER TOLEDO, USA). Metal ions were analyzed by ICP-AES (1200, Agilent, USA) and anions such as sulfate and chloride were measured by ion chromatograph (861, Metrohm, Switzerland). The conductivity of the LFLRO and permeate of MD process were measured using a conductivity meter (CO150, HACH, USA).

Membrane fouling experiment
As shown in Figure 3, during the first 18 hours of operation, the feed was concentrated gradually, with significant permeate flux decline and conductivity increase, indicating the accumulation of membrane fouling. Over this period, the permeate flux declined about 55.5% and the permeate conductivity increased from 36 to 118 μS cm À1 . In addition to the membrane fouling, the permeate conductivity increase in the concentration stage could also be attributed to the migration of NH 3 and volatile organic compounds through the membrane pores (Ding et al. 2006a(Ding et al. , 2006bLin et al. 2018;He et al. 2019). When the concentration factor reached 5.0, the permeate was fed back to the feed tank every half an hour to keep the concentration factor constant in the last 12 hours. The permeate flux slightly declined and stabilized at about 4.70 kg m À2 h À1 as Uncorrected Proof the experiment terminated, which was only 39% of the initial permeate flux. Besides, the permeate conductivity rose to 269 μS cm À1 , suggesting that the permeate quality was deteriorated continuously.
Physical photos of the membrane before and after DCMD experiment were shown in Figure 4. Compared with virgin membrane module (Figure 4(a)), the membrane module after treating LFLRO brine showed visual signs of fouling where brownish deposits covered and clogged some hollow fibers (Figure 4(b)), which would lead to the decline of the feed flow rate and caused the permeate flux decrease.
The SEM images of the fouled membranes and element distribution corresponding to these images were showed in Figure 5. At low magnification a relatively even distribution of a fouling layer and apparently salt crystals can be seen on the inner surface of the fouled membrane ( Figure 5(b)). Figure 5(c) shows a closer image of the fouling layer where three different deposit morphologies were revealed: a porous underlying deposit which was composed of organic scaling and inorganic elements identified by energy dispersive spectroscopy (EDS) (Figure 5(f)), a small number of relatively pure salt crystals on the fouling layer was identified as NaCl by EDS (Figure 5(d)), and some lumpy scaling with rough surface identified by EDS as the mixture of inorganic salts primarily consisted of O, S, K, Na, Mg, Ca ( Figure 5(e)). Furthermore, the cross-sectional image ( Figure 5(a)) showed the presence of internal crystals and organic scaling, suggesting that scaling not only occurred on the surface, but also within the pores and internal structure of the membrane. With deposits inside membrane pores, membrane  wetting would occur, and as a result, the permeate conductivity increased as shown in Figure 3. Besides, the decline of the permeate flux shown in Figure 3 was also associated with the pore clogging by scale deposits formed on the membrane surface and the pore narrowing or blocking by the adsorption of foulants to pores walls.

Chemical cleaning assessment
Five cleaning methods, 2 wt.% NaClO, 3 wt.% SDS, HCl (pH ¼ 2)-NaOH (pH ¼ 12), EDTA (pH ¼ 11.5)-HCl (pH ¼ 2) and 2 wt.% citric acid cleaning, were tested in this study. To evaluate the performance of membranes cleaned by five different cleaning methods, DCMD experiments were performed using 4 wt.% NaCl as feed solution, and the variations of permeate flux and permeate conductivity were analyzed. For comparison, the performance of pristine membrane was also tested via MD experiment. Figure 6 showed the permeate flux recovery rates after membrane cleaning, and the variations of permeate flux and permeate conductivity during the DCMD experiments with 4 wt.% NaCl solution were shown in Figure 7. The initial permeate flux was 12.5 kg m À2 h À1 and kept stable throughout the MD process when using pristine membrane, besides, the permeate conductivity stayed lower than 10 μS cm À1 , indicating the nearly 100% salt rejection. Among all the five cleaning methods investigated, SDS cleaning showed the best cleaning efficiency. Following SDS cleaning, nearly 100% of the initial flux was restored and the permeate flux remained stable above 12 kg m À2 h À1 during the MD process, it was noteworthy that no significant salt leakage was observed, and the permeate conductivity kept stable below 20 μS cm À1 . A similar result was found in the initial flux recovery for the NaClO cleaning method. Although 97.9% initial permeate flux recovery was achieved and the permeate flux remained above 10 kg m À2 h À1 , the permeate conductivity rose rapidly throughout the MD process, indicating the serous salt leakage and the occurrence of membrane wetting. For acid-base solutions cleaning, only 62.8% initial permeate flux recovery was achieved after the HCl-NaOH cleaning, and slightly higher rates were gained after EDTA-HCl cleaning and citric acid cleaning as 71.3% and 77.7%, respectively. Besides, similar variation trend of permeate flux and permeate conductivity was shown after acid-base solutions cleaning. The permeate flux gradually decreased, along with the rapid increase of permeate conductivity, indicating the pore clogging by scale deposits formed on the membrane surface and membrane wetting caused by foulants inside the membrane pores.
The SEM images and contact angles of fouled membranes after five cleaning methods and pristine membrane were shown in Figure 8. After SDS cleaning, there was no obvious fouling layer on the membrane surface, and membrane pores can be clearly seen in the SEM image, suggesting the effective removing of foulants by SDS, which was consistent with others' findings (Li & Elimelech 2004;Ang et al. 2006Ang et al. , 2011. It is noteworthy that EDTA, which was confirmed to be effective for removing organic-Ca 2þ composite pollutants in results of Ang et al., was found inefficient in our study (Ang et al. 2006(Ang et al. , 2011. Considering the pollutants Ang et al. used were mixture of sodium alginate, humic acid and Ca 2þ , there were other organic and divalent cation pollutants (e.g., Mg 2þ ) in the LFLRO brine that would form organic-inorganic complexes,  Uncorrected Proof thus, we thought that under the experimental conditions of this study, SDS may be more effective in breaking up organicmetal ion bindings than EDTA due to its hydrophilic and hydrophobic structures . Besides, SDS can enter the membrane pores and remove absorbed foulants (Liikanen et al. 2002), thus obtained better cleaning efficiency.
As observed by Mo et al. and Yu et al., NaOH and acid cleaning were inefficient and removed only part of organicinorganic composite pollutants (Mo et al. 2010;Yu et al. 2013). After HCl-NaOH and critic acid cleaning methods, a notable amount of scaling was observed on the membrane surface.
Though only relatively few precipitations can be seen on the membrane surface after NaClO cleaning, the fuzzy membrane pores in the SEM image indicated that there were still pollutants inside or around membrane pores. Besides, the obvious signs of membrane damage were shown on the membrane surface, which may be due to the strong oxidizability of NaClO (Wang et al. 2018).
Contact angle measurements of membranes also demonstrated the variations in the efficiency of the five cleaning methods (Figure 8). The virgin membrane showed good hydrophobicity with high contact angle of 118.6°. After DCMD treatment of LFLRO brine, the membrane surface became completely hydrophilic, and the contact angle decreased to 0°. The strong reduction of contact angle suggested the occurrence of serious membrane fouling, which was consistent with the results in section 3.1. Nearly 100% of contact angle was restored (116.4°) after SDS cleaning. In contrast, NaClO, EDTA-HCl, critic acid and HCl-NaOH cleaning restored contact angle to 109.2°, 99.5°, 97.2°and 42.5°, respectively.
Collectively, these results reported here suggest that the membrane fouling during DCMD treatment of LFLRO can be controlled by effective cleaning methods. We took both inorganic and organic foulant on the membrane surface into account and found that the cleaning efficiency of cleaning methods investigated in this study followed an order of SDS . NaClO . citric acid . EDTA-HCl . HCl-NaOH.

CONCLUSIONS
DCMD was effective in concentrating LFLRO brine with high concentration of organic and inorganic contaminants. Although the water recovery can reach as high as 80%, the organic and inorganic contaminants caused severe membrane fouling both on the membrane surface and inside membrane pores.
The fouled MD membranes were cleaned with different cleaning methods, acid-base and citric acid solutions cleaning was ineffective in restoring the permeate flux and the salt rejecting ability of membrane. 2 wt.% NaClO cleaning achieved higher permeate flux recovery rate, but poor cleaning efficiency in removing the foulants inside the membrane pores. Besides, the NaClO cleaning method may cause damage to the membrane. The 3 wt.% SDS cleaning method gained nearly 100% permeate flux recovery rate and no significant salt leakage was observed during the MD process. These results showed that the SDS can be used to restore the membrane performance after MD processing LFLRO brine, thus achieve long stable MD operation.