Volume reduction and water reclamation of reverse osmosis concentrate from coal chemical industry by forward osmosis with an osmotic backwash strategy

The coal chemical industry (CCI) is developing rapidly and demands tremendous water consumption (Li et al. 2015; Jia et al. 2016). CCI is also characterized as pollution-intensive for high strength wastewater, which contains a large number of hazardous organic substances with high COD value and chromaticity (Wang & Han 2012; Chen et al. 2019b). After pretreatment, biological and advanced treatment of the process wastewater, the effluent together with the saline wastewater produced from the circulating water system and desalination process is generally further treated by using ultrafiltration and reverse osmosis (UF + RO double-membrane process) for water reuse purposes (Zhu et al. 2018). However, the double-membrane process can only recover water with limited efficiency; that is, 40–85%, and thus leaves RO concentrate (ROC) with refractory organic pollutants and total dissolved solids (TDS) on the order of magnitude of ∼10⁴ mg·L⁻¹ (Pramanik et al. 2017). Adopting the evaporation pond for ROC final disposal will lead to 15–60% waste of water resources and potential risk of groundwater contamination (Katzir et al. 2012). Therefore, more efficient strategies are required for ROC volume reduction, water reclamation, and decontamination.

Both thermal and membrane-based processes have been developed for ROC volume reduction and zero liquid discharge (ZLD) in recent years (Tong & Elimelech 2016). Thermal processes working upon a brine concentrator and a brine crystallizer are generally expensive and energy-intensive (McGinnis et al. 2013). In comparison, a membrane-based process; that is, the secondary RO in this case, does not require phase transition for separation, making it more energy effective than a thermal process (Elimelech & Phillip 2011). Whereas, practical application of RO appears to be inherently hindered by serious membrane fouling and the upper limitation of external pressure that can be applied to the system. In recent years, electric voltage driven electrodialysis (ED) and hydrophobic membrane-based membrane distillation (MD) have emerged for volume reduction of ROC. Nevertheless, ED and MD are constrained by issues associated with membrane fouling (Korngold et al. 2009; Naidu et al. 2017) and passage of volatile pollutants into the permeate (Shaffer et al. 2013).

In FO, water molecules are transported across the semi-permeable membrane from the feed solution (FS) to the draw solution (DS) driven by osmotic pressure difference, while other components are non-permeable (Cath et al. 2006; Zhao et al. 2012). Without external pressure, FO is more attractive by virtue of less energy input, lower membrane fouling, and the loosely packed fouling layer (Shaffer et al. 2015). Up to now, FO has been extensively investigated for various applications including wastewater treatment (Sun et al. 2018; Gao et al. 2019; Maknakorn et al. 2019), liquid food concentration (Rastogi 2016; Menchik & Moraru 2019) and seawater/brackish water desalination (Chen et al. 2019a; Giagnorio et al. 2019).

Forward osmosis has also been successfully adopted for minimization of ROC from water reclamation plants (Kazner et al. 2014; Jamil et al. 2015, 2016) and groundwater desalting facilities (Martinetti et al. 2009; Wang et al. 2016; Li et al. 2019). It is noteworthy that the ROC used in these studies was either low in TDS (<3,000 mg·L⁻¹) (Kazner et al. 2014; Jamil et al. 2015, 2016) or low in TOC (0.19 mg·L⁻¹) (Wang et al. 2016). In addition, Wang et al. demonstrated the potential of FO in treating ROC (TDS of 5,800 mg·L⁻¹ and TOC of 200 mg·L⁻¹) from textile wastewater (Wang et al. 2020). However, to the best of our knowledge, there are no reports of CCI ROC minimization by FO.

Although FO is considered to be a membrane technology with low fouling risk, membrane fouling of the FS side was observed in most of the studies, causing flux decline and membrane deterioration (Martinetti et al. 2009; Kazner et al. 2014; Jamil et al. 2015, 2016; Wang et al. 2020). Osmotic backwash (OB), which utilizes osmotically induced reverse permeate flow to remove the foulants in the membrane pores and to lift the cake layer, has been proved as an efficient cleaning method to restore the membrane performance (Martinetti et al. 2009; Yip & Elimelech 2013; Motsa et al. 2017; Haupt & Lerch 2018).

In this paper, FO with a commercially available membrane was adopted to evaluate the feasibility of CCI ROC volume reduction. The effects of membrane orientation and DS concentration on the FO performance were investigated. The volume reduction capacity of FO was determined via extended period tests and the specific energy consumption was calculated. To eliminate the membrane fouling, the fouling layer was characterized and OB was assessed for membrane cleaning with comparison to two traditional methods, namely, physical cleaning and chemical cleaning.

The ROC used for the current study was collected from a CCI plant in Heilongjiang Province. In this plant, mixed wastewater from the coal gasification process and coal coking process are treated successively through pretreatment, biological process, and advanced treatment. Afterwards, the effluent and saline wastewater are subjected to a UF + RO double-membrane process for water reclamation. ROC was sampled from the brine stream exiting the RO facility. Table 1 displays the main characteristics of the ROC water samples.

The FO membrane used in the current study was provided by Fluid Technology Solutions (FTS-H₂O, USA) and consists of an active rejection layer made of cellulose triacetate (CTA) as well as a polyester support layer. Basic parameters and guidelines of the membrane are provided in Table S1 (Supplementary Data). NaCl was adopted as the DS. DS and the cleaning solution used in OB were prepared by dissolving NaCl in DI water, and HCl solution (pH = 3) by diluting the concentrated HCl with DI water. All the chemicals used are of analytical grade and were provided by Damao Chemical Reagent Factory, Tianjin, China. DI water was produced by PureLab Prima (ELGA LabWater, UK).

FO experiments were conducted with a laboratory-scale membrane module (Figure S1 Supplementary Data) as described in our previous study (You et al. 2012). The membrane was sealed by rubber gaskets between the DS channel and FS channel, which are designed to have a size of 7.2 cm × 3.5 cm × 0.7 cm with an effective membrane area of 25.2 cm². Both AL-FS mode (active layer facing FS) and AL-DS mode (active layer facing DS) were evaluated. DS and FS flowed concurrently (200 mL·min⁻¹) in the channels, assisted by two peristaltic pumps (WT600-2 J, LongerPump, Baoding, China). The system maintained a constant temperature (25 ± 1 °C) via a thermostatic bath (DC-0510, Scientz, Ningbo, China) and the weight of DS was recorded at regular intervals by an electronic balance (DT5000A, Xinsheng, Shanghai, China), which was connected to a computer.

During OB, DI and NaCl solution (4 M) were introduced to the DS side and the FS side, respectively, with the same flow rate as that in the fouling experiments. Since the osmotic pressure gradient was in the opposite direction compared with the FO process, water moved across the membrane from DS to FS, which enabled pore flushing and cake layer lifting (Motsa et al. 2017). For physical cleaning, DI water circulated on both sides of the membrane with a flow rate twice as large as that in the fouling experiments. As for chemical cleaning, HCl is considered to be efficient for inorganic foulant removal and pH of 3–7 is recommended for CTA membrane operation by the manufacturer, so HCl solution (pH = 3) was used to remove the foulants on the membrane surface on the FS side while DI was circulated on the DS side.

The concentration of anions and cations was determined by ion chromatography (Dionex Integrion HPIC, Thermo Fisher Scientific, USA) and inductively coupled plasma optical emission spectrometry (ICP-OES, Optima 5300 DV, PerkinElmer, USA). pH and TDS values were obtained by a pH detector (FiveEasy, Mettler Toledo, Switzerland) and a conductivity meter (CON 2700, Eutech Instruments, Thermo Fisher Scientific, USA). TOC values were measured by a TOC analyzer (TOC-VCPN, Shimadzu, Japan). The osmotic pressure of FS and DS was calculated by OLI System software.

Morphology of the membrane and the fouling layer was observed with a scanning electron microscope (SEM, Sigma 500, Zeiss, German) and an atomic force microscope (AFM, Dimension Icon, Bruker, German), while the chemical and elemental compositions of the fouling layer were provided by Fourier-transform infrared spectroscopy (FTIR, Spectrum One, PerkinElmer, USA) and X-ray energy dispersive spectrometry (EDS, X-MAX 50, Oxford, UK). All the membrane samples were dried in an oven for at least 24 h before these measurements.

The flux data were obtained under both modes at various DS concentrations; namely, 0.5 M (24.6 bar), 1.0 M (51.6 bar), 2.0 M (13.8 bar), 3.0 M (186.7 bar) and 4.0 M (270.5 bar). With the elevation of the DS concentration from 0.5 M to 4.0 M, the flux increased from 5.3 LMH to 14.8 LMH (179% increase) for AL-DS mode and from 6.4 LMH to 26.4 LMH (313% increase) for AL-FS mode (Figure 1(a)), mainly as a result of increased driving force. In addition, slower increase with DS concentration was observed for both modes due to the combined effects of concentration polarization (CP) and reverse salt flux (McCutcheon & Elimelech 2006).

It is also observed that the flux under AL-DS mode is much lower than under AL-FS mode, especially in the range of higher DS concentration. On the one hand, since the support layer has a rougher surface (Rq = 12.3, Ra = 9.85) than the active layer (Rq = 2.44, Ra = 1.82), as is shown in Figure 1(b), the foulants in FS would accumulate more easily on the rougher support layer (cake formation) under AL-DS mode. On the other hand, the foulants could enter the porous structure of the support layer (pore blocking) under AL-DS mode, impeding the water permeation and increasing the mass transfer resistance for the salt ions, leading to more serious CP. Furthermore, faster water permeation under higher DS concentration leads to a larger drag force of foulants to the membrane, thus accelerating both cake formation and pore blocking. Compared with the relatively clean membrane under AL-FS mode after 1 h operation, a much darker color was observed for AL-DS mode, indicating more serious fouling (Figure S2).

Although it is common practice to employ AL-FS mode for FO, AL-DS mode is favored by pressure retarded osmosis (PRO) using seawater/river water for energy harvest (Chou et al. 2012; Han et al. 2013) as well as applications where FS has relatively low fouling propensity (Kim et al. 2018). To further examine the effect of membrane orientation on long-term FO performance, the experiment was conducted under both modes for five continuous cycles (5 × 4 h) using 4 M NaCl as DS (Figure 1(c)). The initial water flux under AL-DS mode (10.9–18.4 LMH) is much lower compared with AL-FS mode (27.2–28.2 LMH), which is consistent with the results in Figure 1(a). The flux decline within a single cycle was mainly ascribed to the decrease of osmotic pressure difference, since the present study allowed the dilution of DS (osmotic dilution mode) instead of keeping a constant DS concentration (constant DS mode). In comparison, the flux difference between different cycles showed the sole effect of membrane fouling. As we can see, the initial flux declined by about 4% (from 28.2 to 27.2 LMH) after five cycles for AL-FS mode, while 41% (from 18.4 to 10.9 LMH) for AL-DS mode, indicating fast and severe foulant accumulation on the membrane surface and in the porous structure.

The membrane had a much darker color under AL-DS mode than under AL-FS mode after the cyclic experiments, whereas the SEM images showed no obvious difference of foulant accumulation on the membrane surface between the two modes (Figure 1(d)). Therefore, the inferiority of AL-DS mode in the present study is considered mainly derived from pore blocking instead of cake formation. According to the above results, AL-DS mode is unsuitable for CCI ROC volume reduction by FO and AL-FS mode is employed for the following experiments.

To ensure good quality of product water, we also determined the rejection of TOC, SO₄²⁻, Mg²⁺, Ca²⁺, and Si (Figure 2) under different DS concentrations. The membrane could reject more than 98% of SO₄²⁻, Mg²⁺ and Ca²⁺ with negligible effects of DS concentration on the rejection rates. As for Si, a little lower rejection (92–98%) was observed because Si was a major component of the fouling layer (which will be discussed later in this study) and the rejection was calculated using the FS concentration (C_t) after FO treatment (Equation (2)). In contrast, only 33–55% of TOC was rejected due to the existence of small neutral organics in ROC (Alturki et al. 2013; Werber et al. 2016; Zheng et al. 2019). Since the penetration of TOC will deteriorate the product quality, we need to utilize proper technologies before FO to lower the contents of TOC and other hazardous substances, especially when considering salt recovery.

Unlike RO, in which the highest salinity of treated water is restricted by RO equipment and the upper level of feasible applied pressure, FO is able to attain a larger extent of volume reduction with appropriate DS. In this part, FO was operated for extended periods to further reduce the volume of ROC and reclaim more water. To be specific, the FO system recovered 72.1% (1 M), 84.3% (2 M), 90.9% (3 M) and 92.5% (4 M) of water from FS (Figure 3(a)). The recovery rates obtained in the experiment were slightly lower than the theoretical values represented by the dashed lines, which is due to the assumptions made and the fact that osmotic pressure equilibrium across the membrane was not yet achieved. Besides, constant DS mode displayed higher theoretical recovery than osmotic dilution mode, since the former maintained the initial osmotic pressure throughout the operation, which resembled the continuous flow operation in practical applications. The flux declined gradually with increased recovery caused by both the decrease of osmotic pressure difference and membrane fouling; in addition, dramatic flux decline was observed beyond 80% recovery for 2/3/4 M due to the relatively large degree of volume reduction (Figure 3(b)). As we can expect, a higher recovery rate and a faster water reclamation are achievable by utilizing a higher initial osmotic pressure of DS. However, further elevation of recovery rate to values close to 100% leads to a dramatic increase of osmotic pressure, thus requiring a large dose of DS solutes. Therefore, a tradeoff between water reclamation and chemical dosage needs to be considered under practical situations.

In this part, TDS of the FS increased from 1.3 g·L⁻¹ to 4.6 g·L⁻¹ using DS of 1 M NaCl and further increased to 17.4 g·L⁻¹ with 4 M NaCl, indicating that FO possesses a volume reduction capacity similar to a brine concentrator, which is much higher than RO or mechanical vapor compression (Figure 4). As to the energy consumption of the current FO process with possible DS regeneration, 1 M NaCl resulted in an SEC of 0.105 kWh·m⁻³, which declined to 0.065 kWh·m⁻³, 0.056 kWh·m⁻³ and 0.050 kWh·m⁻³ for 2 M, 3 M and 4 M, respectively, making FO much more energy efficient compared with other brine concentration technologies (Figure 4). As presented by Equation (4), a higher DS concentration produces a higher J_w and thus a larger Q_p, which in turn leads to a lower SEC at fixed P_pump. Similar SEC values for FO were previously reported in a study by Lambrechts et al. (Lambrechts & Sheldon 2019). According to the above results, FO is capable of replacing the brine concentrator for ROC volume reduction with very low energy consumption compared with other technologies, meanwhile significantly cutting down the equipment investment. By using CO₂-NH₃ DS of higher concentration with waste heat for DS regeneration, a much larger potential is expected for water reclamation by FO. Although in practice, the system will not work under the maximum recovery condition, FO could still largely reduce the volume of ROC as well as the size of the subsequent concentrator with relatively low energy consumption.

To understand the fouling characteristics of the FO membrane during volume reduction of ROC, accelerated fouling experiments were run for ten cycles (10 × 20 h) with DS of 4 M NaCl, after which the flux declined by about 30% (Figure S3). Then, we employed one of the cleaning methods for 1 h to remove the foulants on the membrane.

Figure 5 displays the photographs of the new membrane, fouled membrane and membranes after cleaning. The covered parts represent areas on the membrane that are covered by foulants visible to the naked eye, while the uncovered parts represent the other areas. The percentages of covered parts and uncovered parts were calculated by an image processing software and are shown in Figure S4. As we can see, the membrane was completely covered by foulants with a dark brown color after the ten-cycle experiment, and the flux declined to 69.1% of the initial value (Figure 6). During chemical cleaning, only about 9% of foulants were removed with a recovered flux of 73.1%, which, however, is most likely to be the result of shear force instead of chemical interaction. Physical cleaning with a higher flow rate was able to wash away 37% of the foulants, restoring 92.6% of the initial flux due to an enhanced shear force on the membrane surface. During OB, water permeates through the membrane from the DS side to the FS side, pushing out the foulants in the micropores as well as exerting a lifting effect on the fouling layer, which could then be washed away by the influent. A recovered flux of 97.1% was achieved by OB, and only 20% of foulants remained due to poor hydraulic conditions near the membrane edges. Therefore, the foulants were removed from the membrane surface, mainly due to hydraulic interaction between the foulants and the influent, while HCl showed insignificant effect on the removal of Si-dominated foulants (details discussed later in the paper). Since physical cleaning adopted a higher flow rate, better cleaning efficiency was achieved compared with chemical cleaning. Although OB utilized a lower flow rate, the reverse permeate flow induced by the osmotic pressure difference exerted a lifting effect on the fouling layer, weakening the foulant adsorption on the membrane, which contributed to an even stronger overall hydraulic interaction and thus a higher cleaning efficiency.

In order to understand the properties of the fouling layer, the morphology and elemental composition were obtained via SEM-EDS. As is shown in Figure 7(a), the surface of the new membrane is smooth and clean, while the slightly bulgy parts originate from the embedded polyester mesh. The atomic ratio of C/O is 68.5:31.5, which is a reasonable value contributed to by both the polyester support layer and CTA active layer. In contrast, the fouled membrane has a very rough surface (Figure 7(b)), indicating a complete coverage of the foulants, which would significantly compromise the FO performance. According to the EDS results of the fouled membrane, At% of O rises to 42.6%, while At% of C declines to 29.2%, with the emergence of a similar amount of Si (27.3%). Besides, S, Mg and Ca account for 0.9% in total.

The changes of surface properties due to the foulant accumulation were further investigated via FTIR. The results illustrated absorption peaks (C = O, C-O-H, and C-O-C) characteristic of the CTA membrane structure for the new membrane (Figure S5). However, these peaks show lower or negligible intensity for the fouled membrane, indicating significant foulant adsorption on the membrane. Meanwhile, the foulants are further identified by the emergence of Si-O-Si absorption peaks, which is consistent with the EDS results in Figure 7(b), where a significant rise of At% is observed for O and Si. Since Si comes from the foulants instead of the membrane and the amount of Si is much larger than S, Mg and Ca, the C:O:Si ratio could serve as an indicator in this study to determine the degree of fouling and the efficiency of cleaning.

To further evaluate the cleaning efficiency of the three approaches, attention was paid to the covered parts and the uncovered parts. Since the new membrane is uncovered with 100% of initial flux and the fouled membrane is fully covered with 69.1% of initial flux, we can assume that the uncovered parts have a flux of 100% and covered parts 69.1%. Then, we can obtain the calculated flux_c after membrane cleaning according to the area percentages (Figure S4) and compare it with the real flux values flux_r in order to determine the permeability change of the fouling layer. For chemical cleaning, the real normalized flux (flux_r) is similar to flux_c, which means that the covered parts remained unchanged after chemical cleaning. In contrast, flux_r is much larger than flux_c after physical cleaning and OB, indicating less foulant accumulation and significantly restored water permeation at the covered parts (considering the lower percentage of covered parts for OB). Figure 7(c) and 7(d) display the SEM-EDS results of the covered and uncovered parts. After chemical cleaning and physical cleaning, the covered parts showed similar morphology and C:O:Si ratios to the fouled membrane. In contrast, only small amounts of foulants were adsorbed on the membrane after OB and the C:O:Si ratio is close to that of the new membrane. As for the uncovered parts, visible foulants remained on the membrane after chemical cleaning, while no obvious foulant accumulation was observed after OB and physical cleaning, and the C:O:Si ratios are nearly the same as that of the new membrane. Thus, OB was further proved to be an efficient cleaning method for CCI ROC volume reduction by FO.

This study investigated the potential of the FO process for volume reduction and water reclamation of CCI ROC. AL-FS mode outperformed AL-DS mode and achieved a flux of 26.4 LMH and a recovery rate of 92.5% at a DS concentration of 4 M NaCl with energy consumption of 0.050 kWh·m⁻³. A flux decline appeared in long-term treatment of ROC mainly due to the formation of Si-containing foulants, which, however, could be removed almost completely by OB owing to the synergistic effect by hydraulic shear force of the influent and lifting force of the reverse permeate flow. FO was proved to be an effective technology for volume reduction of CCI ROC, with OB acting as a promising cleaning method. For the ongoing and upcoming studies, proper pretreatment of ROC needs to be investigated for improved FO performance and better quality of product water. The fouling characteristics in accelerated fouling experiments may be different from real applications, so we still need to explore a suitable chemical cleaning method and combine it with OB to facilitate the foulant removal in real applications.

This work was supported by the National Key Technology R&D Program (No. 2019YFC0408503) and State Key Laboratory of Urban Water Resource and Environment (No. 20180X09).

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