Membrane distillation of wastewater: comparison of model and real organics

Erratic rainfall patterns and associated runoff owing to climate change are one of the main causes of water scarcity, resulting in an uneven distribution of water resources worldwide. This necessitates the development of simple and holistic solutions to ensure a continuous and reliable supply of freshwater (Manikandan et al. 2022). In areas where potable water is in short or limited supply, the treated effluent (TE) generated by wastewater treatment plants (WWTPs) can be used as a viable alternative (Zhang & Liu 2021). Owing to its reduced sensitivity to weather and seasonal fluctuations, reclaimed water, which is an underutilised resource, should be considered for potable and non-potable water reuse applications (Machineni 2020). To meet the growing demand for clean water and reduce the associated environmental implications, state-of-the-art membrane-based separation processes such as reverse osmosis (RO) and nanofiltration (NF) have already been applied (Al-Araji et al. 2022; Abbas et al. 2023). Notably, in recent years, increased focus has been placed on membrane distillation (MD) technology for advanced wastewater treatment (Ramlow et al. 2017; Liu et al. 2020).

MD is a thermal technique that operates based on the principle of pressure difference between the hot feed and cold permeate separated by a microporous hydrophobic membrane (Ramezanianpour & Sivakumar 2014a). It is theoretically possible to obtain a removal efficiency of 100% for non-volatile contaminants because water penetrates the MD membrane as vapour (Ramlow et al. 2017). The MD method operates at a lower pressure (around 2.5 bar) and produces a high-quality permeate (>98% rejection) with reduced fouling propensity compared to conventional distillation processes (Eykens et al. 2017). Compared with pressure-driven membrane processes (e.g., NF), the minimal vapour pressure differential necessitates less stringent mechanical qualities in the membrane materials (Yan et al. 2019). To date, MD has been the subject of substantial research on brackish water treatment (Ramezanianpour & Sivakumar 2014b; Sivakumar et al. 2014) and desalination (Jia et al. 2021). However, its application in the reclamation of TE remains largely unexplored (Liu et al. 2020; Jeong et al. 2021).

When using a low saline feed, such as TE, organic fouling could become dominant and affect the MD performance. This could be attributed to the significant presence of effluent organic matter (EfOM) (Kumar et al. 2020). Organic fouling is primarily caused by EfOM owing to physical sorption and electrostatic interactions with the hydrophobic MD membrane (Tibi et al. 2021). According to our literature survey, recent studies have been conducted on organic fouling using MD technology (Cho et al. 2018; Charfi et al. 2021). However, organic fouling mechanisms have been predominantly assessed and identified using individual model organic foulants such as humic acid (HA) (Cho et al. 2018; Liu et al. 2018). The findings obtained using model foulants may not be linearly applicable to the fouling caused by real wastewater owing to its complex composition. In this context, the organic fouling of MD membranes in model foulant scenarios versus real wastewater scenarios (i.e., TE) needs to be explored and elucidated. With regards to the EfOM, it has been acknowledged that the main organic foulants found in TE include humic substances (HS) and polysaccharides (Zheng et al. 2019; Charfi et al. 2021). The significance of HS is because of its sorption onto the hydrophobic MD membrane surface (Naidu et al. 2015), whereas polysaccharides, such as alginates, promote a dense fouling layer due to their high compressibility and deformability (Tibi et al. 2021). Previously, Cho et al. (2018) observed the formation of a fouling layer caused by HA deposition, which reduced the direct contact MD (DCMD) flux by 40%. Moreover, Liu et al. (2018) revealed the development of a dense gel layer after exposure to sodium alginate (SA) because of its hydrophobic affinity to the MD membrane, resulting in a rapid flux decline in the early phase of operation. Owing to the interaction of various organics, Qin et al. (2017) revealed that fouling by a mixture of organics was more severe than fouling caused by individual organics. However, to the best of our knowledge, the comparative evaluation of model and real foulants on organic fouling of MD membrane has yet to be systematically investigated.

Although several studies have been conducted on the fouling of MD membranes, they have predominantly focused on the fouling caused by individual foulants (Han et al. 2017; Cho et al. 2018; Couto et al. 2019). Only a few studies have employed real feed (e.g., TE) (Jeong et al. 2021; Ji et al. 2023), probably because the initial research on MD was focused on its desalination performance. Moreover, the majority of fouling experiments in an MD system have been performed using model organics, which should not be expected to mimic the fouling behaviour in the real wastewater scenario (Fortunato et al. 2018). For practical implications, it is necessary to use a real feed for acquiring a better understanding of fouling behaviour among various foulants. Due to the complexity of organics, Yan et al. (2021) called for a comprehensive evaluation and research on the impacts of organics on the MD performance during the treatment of real wastewater. A complex and thick membrane fouling layer may form during the treatment of TE, which is promoted by complex organic interactions. For instance, Kumar et al. (2020) confirmed that the EfOM in TE has 50% reduction on the permeate flux of the MD system. Similarly, Jeong et al. (2021) reported that a 30% flux decline in MD caused by membrane fouling. It is important to note that most MD investigations using real wastewater as a feed have represented membrane fouling as a decline in permeate flux that might not correspond to the complete extent of organic fouling (Naidu et al. 2015; Fortunato et al. 2018; Asif et al. 2021). A valuable resource for understanding organics and their behaviour in membrane fouling could be achieved by employing a more responsive and sensitive analytical approach, such as liquid chromatography–organic carbon detection (LC-OCD). Research on membrane fouling in MD is based on short-term experiments (12–24 h) conducted mostly in the concentration mode (Asif et al. 2021; Ji et al. 2023). To trace the development of organic fouling, it was envisioned to evaluate the long-term performance of DCMD in a continuous-flow system for TE treatment.

Based on the research gaps highlighted above, this study aims to provide an effective understanding of the organic fouling behaviour caused by HS and polysaccharides in a DCMD process, including their rejection. We selected the DCMD configuration owing to its operational simplicity. The DCMD module was operated under three distinct scenarios: (i) individual model organics at different low to high concentrations (20–180 mg/L), (ii) a combination of model organics at concentrations comparable to EfOM in TE (5–20 mg/L each), and (iii) TE collected from a local WWTP. The influence of individual foulants and their combinations on fouling propensity was assessed and compared with that caused by TE. It is essential to note that the study of model organics was conducted in the absence of ions under low-salinity feed conditions. Several characterisation and analytical tools were employed to gain a deep understanding of the correlation between the fouling propensity and flux decline. In this context, HS were given special attention since they could thermally disintegrate into low-molecular-weight (LMW) organics. We believe that the results of this study have practical implications and will serve as a valuable reference for future studies of TE treatment using DCMD.

Two different model foulants of analytical grade were selected (SA and HA) Sigma-Aldrich (St. Louis, MO, USA). These were selected because they are predominantly present in TE to represent polysaccharides and HS. Deionised water was used to prepare stock solutions. Effluent treated biologically was collected from a municipal WWTP (NSW, Australia), and the basic characterisations were carried out. The wastewater had a total organic carbon (TOC) concentration of 40.29 ± 0.1 mg/L, a pH of 6.36, and a conductivity 0.929 mS/cm.

A total of 13 DCMD experiments were conducted for the model organics, as shown in Table 1. Experiments were conducted with the aim of exploring the impact of model organic fouling over a diverse range of concentrations and combinations. As a model HS, five different concentrations of HA (i.e., 20, 60, 100, 140, and 180 mg/L) were used in the first five runs. Five runs were performed with different concentrations of SA (i.e., 20, 60, 100, 140, and 180 mg/L) as model polysaccharides. Three different combinations of SA and HA were investigated to explore their interactions between HA and SA in the MD system. For these three runs, the total concentration was maintained at 20 mg/L with ratios of HA:SA (25/75, 50/50, and 75/25). The initial volume of the feed was maintained at 3 L for all model organics studies and was conducted in concentration mode with 80% recovery. The feed solution was filtered through a 0.45 μm membrane filter paper to prevent the initial deposition of suspended particles on the MD membrane.

To elucidate the fouling effect of real wastewater on the membrane, the MD module was run in two distinct modes: (i) concentration mode and (ii) continuous mode. TE was used as the feed in concentration and continuous modes, and the volume of feed tank was 3 L. The MD system was operated at an 80% recovery rate using an identical feed. A concentration mode cycle was conducted to evaluate the impact of the bulk organics in the feed. Ten cycles of the concentration mode were designed to achieve a 60% reduction in permeate flux, which served as the basis for a comparative analysis of the projected performance of a high concentration of model organics. The continuous mode was then studied by maintaining the volume of the feed tank at 3 L, which was controlled by a level sensor connected to a peristaltic pump fed by a TE reservoir tank. This mode was operated for 22 days until the flux decline was 90%. This showed the long-term effect of real wastewater (TE) on MD membrane fouling.

Permeate and feed samples were analysed for aromaticity (UV₂₅₄) at 254 nm using a UV–visible spectrometer (UV-2600, Shimadzu, Japan). The EfOM was analysed as TOC using a TOC analyser (Shimadzu, Japan). The morphology and composition of the fouling layers on the MD membrane surface were determined by scanning electron microscopy (SEM) combined with energy-dispersive X-ray spectroscopy (EDS) (JCM-600, JEOL, Tokyo, Japan). As the membrane samples were non-conductive, they were coated with platinum (Pt) for SEM-EDS analysis. Using attenuated total reflection-Fourier transform infrared (ATR-FTIR) spectroscopy, the chemical composition of the fouling layer was analysed (IRAffinity-1, Shimadzu, Kyoto, Japan). Using a goniometer (Model250, Rame-Hart, Netcong, NJ) based on the usual sessile drop method, the contact angle was measured. Using LC-OCD, the detailed organic characterisation of the feed, permeate, and the membrane fouling layer was assessed (Model 8, DOC-Labor, Karlsruhe, Germany). The major fractions detected by LC-OCD include biopolymers (BP), building blocks (BB), LMW acids, and neutrals (Naidu et al. 2015; Fortunato et al. 2018; Asif et al. 2021).

Notably, at HA concentrations of 20, 60, and 100 mg/L, the flux reduction was insignificant and remained below 10% (Figure 2(a)). It is evident from the experiment that when the concentration of the individual HA feed is low, they do not exhibit notable bonding with the membrane surface. This is in accordance with a previous investigation, in which a feed containing 100 mg/L of HA, even in the presence of salts, did not affect the permeate flux, i.e., less than 5% flux decline (Han et al. 2017). Therefore, it can be inferred that high HA concentrations (140 and 180 mg/L) play a significant role in the fouling behaviour of the MD membrane. However, the initial decline in flux for all concentrations of HA was due to the initial deposition of HA on the membrane surface, which was caused by its hydrophobic interactions with the MD membrane surface (Han et al. 2017; Cho et al. 2018). Subsequently, a significant initial flux decline was observed only when the HA concentrations in the feed were 140 and 180 mg/L (Figure 2(a)). Interestingly, this initial flux decline only occurred within the first 100 min of the DCMD operation. This is because HA forms a fouling layer via hydrophobic interactions, which seems to limit further HA deposition on the membrane surface (Couto et al. 2019). A previous study (Cho et al. 2018) showed that with the variation in HA concentration (0–200 mg/L) in feed solutions during short-term experiments, the extent of flux decline was governed by the concentration of HA, which is consistent with the findings of this study.

The behaviour of the flux decline in DCMD by SA (Figure 2(b)) was observed similar to that obtained for different HA concentrations (Figure 2(a)). There was also an initial drop in permeate flux across all SA concentrations. Similarly, the initial decline in flux was comparatively higher (5–60%) for SA concentrations of 140 and 180 mg/L. Moreover, this initial flux decline also occurred within the first 100 min, after which the permeate flux remained almost constant owing to the homogeneity of the feed solution or the hydrophilic nature of the fouling layer. When the flux declines in HA and SA (Figure 2) were compared, the flux decline in SA was greater than that in HA, although both tended to achieve a pseudo-steady state. The presence of carboxylic groups in SA causes it to be deposited on the MD membrane as a gel-like structure (Liu et al. 2018). This trend was observed for all concentrations of the SA feed, indicating that polysaccharides caused far more severe membrane fouling. Our findings concur with the existing literature, indicating that polysaccharides are the primary constituents of fouling in natural water (Katsoufidou et al. 2010). In a study conducted by Kim et al. (2019), it was observed that MD feed containing 50 mg/L of SA did not result in any decline in flux, which aligns with the findings of the present study (Figure 2). In brief, it was observed that an initial higher feed concentration results in a higher permeate flux decline, a phenomenon that appears to occur only during the early stages of DCMD operation. We suggest that the change in membrane hydrophobicity caused by the hydrophilic and negatively charged fouling layer may limit further deposition of model organic foulants after a certain operating time.

To investigate the applicability of the fouling behaviour assessed using model organic foulants and actual wastewater, the DCMD was operated with real feed (i.e., TE) in the concentration and continuous modes. The permeate flux results for the real feed (Figure 3(b)) showed a significantly different trend from that of the model organics. The gradual decrease in the permeate flux observed in the TE samples was not present in the model organics, even at high concentrations. This disparity can be attributed to the presence of complex organics in the TE, which led to the continuous deposition of EfOM. Moreover, the deposition of foulants is facilitated by inorganics even when the ionic strength of the feed is very low. The findings of this study regarding the DCMD flux behaviour for wastewater treatment are consistent with those reported in several previous studies (El-Abbassi et al. 2013; Lin et al. 2015).

The EfOM in TE comprises complex organics (Manikandan et al. 2022), which results in a significant decline in the permeate flux. Moreover, compared to model organics, TE also contains ions (such as Mg²⁺ and Ca²⁺) that interact with the negatively charged organics adsorbed on the membrane surface and may neutralise the ions (Yan et al. 2019). It is well known that divalent scale-forming cations act as a bridge between organic foulants, providing active sites for organic foulant attachment or disposition. This resulted in the formation of a dense membrane fouling layer. In addition, complex organic foulants can form a cation-stabilised gel, termed a Begg-box-shaped gel (Ji et al. 2023). Subsequently, the scaling agents might be further attached via electrostatic interactions to an already fouled layer, turning the membrane surface even less negatively charged. This continuous deposition significantly lost the hydrophobicity of the membrane surface. Despite the severe and steady build-up of the fouling layer, the DCMD was able to remove more than 98% of organics, as evident from the LC-OCD analysis.

The concentrate mode experiments revealed a steeper flux decline during all 10 runs over a 9-day period (Figure 3(b)). In contrast, the long-term study performed in continuous mode demonstrated a gradual decline in flux for the first 10 days. A slight increase in flux decline was observed between days 10 and 18. Furthermore, an abrupt flux decline was observed between days 18 and 21. However, the flux remained stable between days 21 and 25. As reported by others, this trend is caused by the breakdown of high-molecular-weight organics into low-molecular-weight organics between days 18 and 21. The deposition of organics with LMW was evident based on the LC-OCD analysis, as explained in subsequent sections. It is possible that small particles might penetrate or adhere to the membrane surface, which is likely the cause of the substantial reduction in permeate flux between days 18 and 21 (Liu et al. 2020). In summary, it can be deduced that the fouling behaviour during the treatment of real wastewater differs significantly from that observed in the case of synthetic wastewater containing model foulants.

When the DCMD was run with real wastewater (i.e., TE), a dense fouling layer was formed. The SEM images revealed that the organics significantly affected the morphology of the fouling layer. EDS was used to detect the presence of many elements (i.e., C, O, P, Cl, Fe, S, Na, Ca, Mg, K, and Si) in the fouling layer. With TE as a feed, the fouling layer was mostly composed of O (54%) and Ca (11.3%), whereas the cumulative concentration (% by weight) of other components, including P, Cl, Na, S, and Mg, was less than 10%. This proves that inorganic foulants, predominantly Ca²⁺, act as a bridge between organic foulants and hydrophobic membrane, resulting in severe and rapid organic fouling in MD (Liu et al. 2018). The SEM images showed that the deposits were extremely dense on the membrane surface. The insignificant presence of divalent ions (e.g., Ca²⁺ and Mg²⁺) can result in the formation of metal–natural organic matter complexes, which may contribute to highly compacted fouling deposits and a decrease in membrane performance (Silva et al. 2018). Based on the results of the EDS analysis of the fouled membrane, calcium (Ca) was present, which promoted the formation of calcium precipitates. The rapid conversion of HCO₃⁻ to CO₃²⁻ in the feed solution under heating conditions could expedite calcium deposition on the MD membrane (Tijing et al. 2015). In conclusion, HA and SA compounds individually resulted in initial (uneven) deposition governed by their initial concentrations because of the interaction between the hydrophobic membrane and the organics. Even a combination of model organics did not result in considerable fouling development. Considerable fouling development was observed in MD when TE was used as the feed. Considering the distinct organic fouling patterns observed on hydrophobic membranes, this study conducted additional research with a feed solution and characterised organic foulants. This is especially pertinent because most real feeds contain complex organic compounds.

The chromatogram obtained from the LC-OCD analysis of the DCMD permeate demonstrated notable rejection of EfOM. The permeate TOC concentration was found to be 1.23 mg/L, indicating a retention of 98% organics by the MD membrane. The MD process is supposed to only allow volatiles to pass; however, as shown in Figure 7, the prevalent organic compound present in the permeate was HS, constituting 36.8% of the total organics, with a concentration of 0.4 mg/L. This is due to the reason that HS can pass through the membrane to permeate via adsorption–desorption inside the unwetted membrane pores (Li et al. 2023). The overall results of LC-OCD in the DCMD permeate showed LMW neutrals (0.21 mg/L, 17%), BB (0.12 mg/L, 10%), and BP (0.09 mg/L, 7.4%). This shows that there was a high rejection of organics by the PTFE membrane and no pore wetting occurred.

To further identify the organics which lead to fouling, LC-OCD analysis was performed on the membrane foulant layer. The LC-OCD chromatogram of the membrane foulants shown in Figure 7 shows a high peak at an RT of 110 min, which indicates a high deposition of LMW neutrals and BB on the membrane surface. The presence of LMW neutrals and BB components in the total organic deposition was reported at 29.8 and 25%, respectively. Surprisingly, LMW organics were not considerably present in the TE feed, this phenomenon has previously been reported as high feed temperatures thermally disaggregate HS to LMW compounds (Fortunato et al. 2018). In membrane filtration, LMW neutrals have been recognised as one of the most prominent organic foulants (Li et al. 2016). The hydrophobicity of the MD membrane promoted the adsorption of organic materials on its surface. According to Zheng et al. (2019), proteins tend to accumulate on hydrophobic membranes (Naidu et al. 2014).

The TE feed contains complex organics with both hydrophilic and hydrophobic molecules (Naidu et al. 2014). The LC-OCD results showed that there were 47.5% (37.31 mg/L) hydrophobic and 52.5% (41.20 mg/L) hydrophilic compounds in the TE feed. The interaction between the hydrophobic and hydrophilic organic compounds forms a mechanism for the layer-by-layer effect. Initially, hydrophobic molecules are adsorbed onto the membrane surface, followed by the accumulation of hydrophilic organics, resulting in a dense layer on the membrane surface, leading to flux decline. Naidu et al. (2017) demonstrated that LMW organics possess hydrophilic propensities causing a substantial reduction in membrane hydrophobicity, causing a significant deposition of hydrophilic LMW organics. Moreover, Zheng et al. (2021) demonstrated the hydrophobic–hydrophilic interaction as ‘brick-layering’ process, where hydrophobic foulants acts as concrete and hydrophilic foulants act as bricks. Thus, it can be concluded that model organics (such as HA and SA) do not accurately replicate real humic compounds and polysaccharides.

This study was carried out to better understand the fouling behaviour of organic foulants in a direct contact MD (DCMD) system. To assess the impact of organic fouling on the MD process, a comparison between the model and real organics was performed. The findings presented here indicate that, in the DCMD system, model organics behave differently from real organics. Even a high concentration of individual model organics only caused an initial flux decline, after which the flux became stable. Moreover, there was no considerable decline in the flux with the combination of model organics. However, the real organics (EfOM) in the TE contained complex organics, including hydrophobic and hydrophilic organics. Organic fouling development using TE as the feed is a phenomenon of hydrophobic–hydrophilic interaction, which is initiated by the deposition of LMW hydrophobic organics, followed by interaction with hydrophilic organics. Thus, model organics may not be directly used in future research. Moreover, the DCMD system was also proven to be an effective way to reclaim water from the TE, as evident from the permeate quality (>98% organic rejection). The results of this study will not only aid future investigations into wastewater reclamation by DCMD systems but will also contribute to the development of high retention membrane technology for wastewater reclamation. Further investigation of the impacts of biofouling and pretreatment options using real wastewater is also recommended.

This research was carried out with the support of the Australian Government Research Training Program Scholarship to RH through the University of Wollongong (UOW). The authors acknowledge the use of facilities available within the School of Civil, Mining, Environmental and Architectural Engineering of UOW. We acknowledge access to FTIR at the School of Chemistry and Molecular Bioscience of UOW, Australia.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.