Coupled effect of organic fouling and scaling in the treatment of hyper-saline produced water using forward osmosis

Abstract Hyper-saline water streams, such as the produced water from the oil industry, cannot be treated via the conventional desalination method, such as reverse osmosis (RO), due to the extreme required pressure to overcome the osmotic pressure. This research investigates the effect of oil fouling and cleaning behaviors when the forward osmosis (FO) process is applied to treat hyper-saline-produced water. Synthetic produced water with total dissolved solids (TDS) concentration of 240 g/l (contained Na, Ca, Mg, Fe, Cl, SO4, and CO3 ions) and oil content of 100 ppm was used as a feed solution. Also, two draw solutions, namely: ammonium bicarbonate (NH4HCO3) and magnesium chloride (MgCl2) of concentrations 6 M and 4.8 M, respectively, were utilized for the FO operation. Results showed that the MgCl2 draw solution provided significantly higher water flux than NH4HCO3 due to MgCl2’s high osmotic pressure and low scaling influence. Furthermore, the individual impact of oil was found to be low on the FO’s water flux. However, a synergistic effect between the oil and inorganic scaling was noticed. SEM technique was used to comprehend the observation and analyze the fouling content. Although the feed was a highly polluted stream, it was found that osmotic backwashing can recover more than 90% of the initial flux.


Introduction
Due to the increase in water demands and depletion in traditional water resources, unconventional water resources, such as produced water and wastewater, have gained attention worldwide to fulfill the need . Oilfieldproduced water (PW) can be considered a potential clean water resource if treated efficiently and economically (Waisi et al. 2015). The PW may contain different kinds of inorganic and organic constituents, which makes it irregular compared with other sources like seawater Fakhru'l-Razi et al. 2009). Such a complex solution, especially with high levels of salinities, cannot be treated by conventional methods like reverse osmosis but requires particular curing methods (Shaffer et al. 2013). Forward osmosis (FO) is a membrane-based desalination process that depends on the osmotic pressure potential on drawing water molecules from saline water by a higher salinity solution called draw solution. This process has the inverse concept of the reverse osmosis (RO) process and, hence, no hydraulic pressure is needed (Kadhom and Deng 2018). The FO process gained its significance due to the low required energy for its operation compared with the other desalination processes. In terms of equivalent work (which is the energy required to desalinate 1 m 3 of seawater), Figure 1 shows five desalination processes including the forward osmosis, which requires the lowest value. In the figure, the MSF is multi-stage flash distillation, MED is multi-effect distillation, LT is low temperature, and TVC is thermal vapor compression (McGinnis and Elimelech 2007). Hence, the FO process was applied in many processes, but we will focus on its role in desalination. When the feed is very concentrated saline, the frequently used RO process becomes deficient because it needs very high pressure to overcome its osmotic pressure. Thereby, alternative solutions and processes should be introduced to treat this type of feed, with the aim of using the lowest operational energy ). However, FO can potentially deal with highly impaired liquid streams, such as the PW (Coday et al. 2014;Al-Furaiji et al. 2020).
The presence of low solubility inorganic salts of carbonate and sulfate at high levels cause deposition of these salts on the membrane's surface and form scaling. As one of the fouling types, scaling causes severe problems in piping and pumping, increases the operational cost, and decreases the permeable water flux across the membrane. Here, Mi and Elimelech investigated the inorganic scaling of the FO process using gypsum as a model. They found that more than 90% of the initial flux can be recovered in gypsum scaling by rinsing with DI water (Mi and Elimelech 2010). At the same time, Choi et al. demonstrated that gypsum scaling is fully reversible by only applying ultrasound treatment (Choi et al. 2014). However, organic fouling has a different mechanism than that of inorganic fouling, where multiple physical and chemical parameters could affect it. Crystallization of the low solubility salts on the membrane's surface governs the inorganic fouling, whereas the organic fouling represents adhesion forces between the organic species and membrane surface. Generally, there are three types of materials that are used as models to study the organic fouling in surface water and wastewater; namely, bovine serum albumin (BSA), humic acid (HA), and alginate. It has been reported that organic fouling results in a reduction in the water flux of the FO process and surly other membranebased processes (Mi and Elimelech 2008). However, the flux can be entirely recovered by increasing the cross-flow velocity, indicating that the fouling layer is loose compared to the RO, in which fouling is irreversible (Lee et al. 2010).
The coexistence of scaling and organic fouling leads to forming a complex fouling effect, which leads to a further reduction in water flux compared to the impact of individual fouling. Furthermore, the combined fouling is not fully recoverable by rinsing the membrane with DI water (Liu and Mi 2012). Conversely, Parida and Yong Ng indicated that the presence of Ca þ2 ions does not significantly influence the organic fouling in the FO mode (Parida and Ng 2013). This could be attributed to the lower crossflow velocity used as they ran their experiment with high cross-flow velocity (i.e., 50 cm/s) compared to the tests of Liu and Mi (i.e.,8.5 cm/s) (Liu and Mi 2012). Using oil as a source of fouling has been studied by Duong and Chung; they found that the oil tends to form a fouling layer on the membrane surface and results in a decline in water flux (Duong and Chung 2014). So far, there has been no study on the combined effect of oil fouling and scaling, which is the actual case in the PW produced from the oil industry and wastewater of petroleum refinery and vegetable oil industry.
The main objective of this work is to study the effect of oil on the inorganic fouling (scaling) in the FO process. Synthetic produced water was prepared to simulate the hyper-saline produced water that contains elevated concentrations of the inorganic salts and a low oil concentration. Two different draw solutions were utilized in the experiments: magnesium chloride (MgCl 2 ), and ammonium bicarbonate (NH 4 HCO 3 ). The osmotic backwashing of the combined fouling was also investigated in this research.

Feed and draw solutions
The feed solution was prepared to simulate the real composition of produced water from the oil fields in the south of Iraq. Detailed analysis of produced water from the super-giant oilfields in the south of Iraq is provided elsewhere (Al-Rubaie et al. 2015). The used salts in the synthesized PW were purchased from different vendors, where NaCl (99.9%), CaCl 2 (98.0%), and MgSO 4 .7H 2 O (99.3%) were purchased from Fisher. However, MgCl 2 .6H2O (99%) and FeCl 3 .6H 2 O (99þ %) were obtained from Acros organics, while NaHCO 3 (99%) was provided from J.T. Baker. Oil exists in the produced water mainly in the emulsion phase (d p < 20 mm), where free and dispersed oil (d p > 20 mm) are separated in earlier treatment stages of skimming, gravity separation, dissolved air floatation (DAF), and coagulation and flocculation [12]. Petroleum and Tween 80 were purchased from Sigma-Aldrich and were used as an oil source and emulsifier, respectively. The emulsion was prepared by mixing oil with the emulsifier in a ratio of 9:1, respectively, and added to DI water in a blender (Waring Products Division, Torrington, CT, USA) at a speed of 20,000 rpm for 3 minutes. Two oil-water solutions were synthesized, 100 and 1000 ppm, in this study. Then, the emulsion was mixed with the solution that contains the other salts described in Table 1 to form the synthetic produced water. The osmotic pressure of feed and draw solutions can be indicated here in the text and in Table 1. To eliminate the effect of scaling and oil, baseline tests were conducted with only NaCl and NaCl þ oil as feed solutions. However, the osmotic pressure of the fouling and baseline tests was kept the same.
Two draw solutions were used at concentrations near their saturation: 6 M ammonium bicarbonate (NH 4 HCO 3 ) and 4.8 M magnesium chloride (MgCl 2 ).

Characterization methods
The membrane surface was analyzed by Scanning Electron Microscopy-Energy Dispersive X-ray (SEM-EDX) Apparatus JSM-6335F (FESEM, JEOL Ltd., Japan). Prior to imaging, the samples were stuck on a carbon-made sticking pan and sputtercoated with a thin layer of gold at 20 milliAmp for 60 s. Imaging was done using an accelerating voltage of 10 kV and a current of 12 mA. Particle size distribution was measured by dynamic light scattering using a submicron particle sizer (NICOMP, Santa Barbara, California, USA).

Forward osmosis test
The FO tests were conducted using a custombuilt experimental setup, where two tanks were used for the feed and draw solutions. The membrane was installed in a flat sheet module with an effective area of 20 cm 2 , while the flowrate of feed and draw solutions was 1 L/min (25 cm/s). Two two-lit tanks were used for the feed solution and draw solution. All experiments were conducted for 20 h. of operation using FO mode (feed solution facing the active layer).
All experiments were conducted at room temperature (25 C). Figure 2 shows a diagram of the process.
Two commercially available membranes were purchased from Hydration Technology Innovations (HTI) and used in this research. The first one is made out of a cellulose triacetate (CTA) active layer supported by an embedded polyester mesh. The other is a thin film composite (TFC) membrane consisting of a selective polyamide layer supported by a polysulfone porous layer and a polyester mesh.
The osmotic backwashing was conducted by replacing the draw solution with DI water and running the FO experiment for 1 h. Then, the FO test was repeated using the same feed solution and draw solution.

Effect of draw solution
Two draw solutions were tested in this research, ammonium bicarbonate and magnesium chloride. Also, adding 100 ppm of oil to the NaCl feed solution reduced the water flux slightly. For the more practical case, the NaCl baseline was replaced with the PW, and adding oil to it was considered. Similar findings were observed, where the MgCl 2 showed a higher average water flux than the NH 4 HCO 3 . However, a remarkable drop in the water flux can be observed when the NH 4 HCO 3 was applied as a draw solution and the PW as a feed solution. When oil was added, a drop in the flux was recorded, which means that the fouling increased, where the lowest water flux was calculated when the feed was PW þ oil, and the draw solution was NH 4 HCO 3 . In other words, the MgCl 2 draw solution performed higher water flux than the NH 4 HCO 3 by 0.70, 0.69, 3.55, and 5.52 folds when the feed solution was NaCl baseline, NaCl baseline þ oil, PW, and PW þ oil, respectively. Figure 3 shows the water flux of different feed solutions using the two draw solutions. The SEM images of the fouled membranes for both draw solution is shown in Figure S1.

Effect of oil concentration
In order to confirm the stability of the emulsion, the particle size distribution of the synthetic produced water was measured before and after the FO experiment as shown in Figure S2. From the figure, the average particle size of freshly prepared solution and after 24 h were 600 and 800 nm, respectively. These values of the feed solution showed that the average oil droplets are way below 20 mm, which is in the range of the emulsified oil form. Results showed that increasing oil concentrations in the feed from 100 to 1000 mg/L decreased the water flux by 15.6% when MgCl 2 as a draw solution and HTI-CTA membrane were employed. Figure 4 shows a drop in the water flux from 3.2 to 2.7 LMH when the oil content increased from 100 to 1000 ppm, respectively. The oil droplets' impact appears due to the precipitation and coagulation of oil droplets on the membrane's surface. Hence, increasing oil concentration in the feed resulted in increasing the fouling on the membrane's surface, which is attributed to the increased organic matter and foulants' positive charge. The variation of the water flux with time for 20 h of operation is depicted in Figure S3. Figure 5 shows SEM images that clarify the occurrence of a visible fouling layer on the membrane surface because of the agglomerations of participating emulsified oil droplets. In the figure, the fouling was on the CTA membrane when the feed was PW þ 1000 ppm oil and the draw solution was MgCl 2 .

Membrane type
Each membrane has its unique physiochemical surface properties, including the hydrophobicity Figure 2. A schematic diagram of the lab-scale FO system. and surface roughness related to the water flux variation. It was observed that the water flux of the TFC membrane was higher than the CTA one during the test by 11.11%, as shown in Figure 6. The membranes were operated at 25 C using a feed of PW þ 100 ppm oil and a draw solution of MgCl 2 . This difference can be attributed to the fact that the structural parameter of the TFC membrane is higher than that of the CTA membrane (Figure 7). Although this would lead to a higher internal concentration polarization (inside the support), the selective layer of TFC is much thinner, which leads to higher water flux. Figure 8 compares SEM images of the formed fouling layers on the CTA and TFC membranes under the same operating conditions. In spite of the fouling layer on the TFC membrane appearing more intensive than that formed on the CTA membrane, the TFC membrane showed better performance in terms of water flux. Similar findings were obtained by Li et al. (2018) when they studied the effect of soluble algal products (SAPs) on membrane fouling during the FO operation. The TFC membrane showed a rougher morphology with a higher porosity structure than the CTA one, which could explain the water flux outcomes. However, they also used MgCl 2 as a draw solution, in addition to NaCl and CaCl 2 . When the MgCl 2 was operated as a draw solution, a cake layer formed on    the face next to the feed solution when the TFC membrane was operated.
Nevertheless, no notable layer formed at the draw solution side, and no interaction between the salt ions and SAP occurred. When CaCl 2 was applied as a draw solution, a cake layer was generated on both sides, which led to a reduction in water flux. In addition to the fouling layers, blocking within the transport channels of the support layer forms. When the morphology was examined by SEM, a denser fouling layer was observed on the TFC membrane. The results concluded that smoother morphology and low reverse flux rates from the draw solution contribute to decreasing the fouling layer thickness. These findings are identical to ours, where we found that the MgCl 2 draw solution gave higher flux than the NH 4 HCO 3 ; also, the MgCl 2 resulted in a thinner fouling layer. It is worth mentioning that even though the TFC membrane is hydrophilic, the support layer is hydrophobic (Kadhom and Deng 2019). Hence, no crossed oil droplets were detected in the output due to the different hydrophilicity nature between oil and the TFC membrane's surface. Also, the molecular weight/size of oil molecules is high compared with salts and other ions (Closmann and Seba 1990). It was found that the penetration of organics through a polyamide layer was slower than other materials (Flaconn eche et al. 2001). Here, oil droplets could accumulate on the surface due to the interaction between the droplets and the surface's active groups, salts ions, and other fouling scalents. The SEM test showed an intense cake layer on the TFC membrane, which could be attributed to the interaction of oil droplets with the feed salts ions on one hand and the carboxylic acid groups on the membrane's surface on the other hand. Also, the negative charge of the membrane's surface could generate an   interaction force with the positively charged oil droplets (Hou et al. 2021).

Osmotic backwashing
In order to measure the impact of treatment on membrane fouling, osmotic backwashing was applied, and the water flux values before and after osmotic backwashing were compared. PW with two oil concentrations, 100 and 1000 ppm, were used as feed and the two draw solutions, MgCl 2 and NH 4 HCO 3 , were employed. However, when the MgCl 2 was applied to draw a PW of 100 ppm, the generated fouling was not thick; hence, the recovered water flux value was not much different from the fouled membrane's flux. On the other hand, when the draw solution NH 4 HCO 3 was employed with PW of 1000 ppm, the fouling layer was thick and it extended to the other side; hence no recovery trails succeeded.
The best scenarios to study the backwashing effect were listed in Figure 8, where two cases of MgCl 2 and PW þ 1000 ppm oil and NH 4 HCO 3 and PW þ 100 were illustrated. In the figure, the normalized flux values, which are calculated by dividing the case flux by the original value of the flux, were listed. In the MgCl 2 case, the recovered normalized flux was more than 25% higher than the fouled membrane normalized flux. Also, in the NH 4 HCO 3 case, the recovered normalized flux was higher by around 35% than the fouled membrane normalized flux. Each of the previous cases is compared with itself and cannot be compared with the other one due to the difference in oil content in the feed. These values indicate the usefulness of using osmotic backwashing; though, it could be durable up to a limit, where the membranes could have irreversible fouling. Figure 9 shows a diagram of the fouling mechanism when the two draw solutions were applied with 100 and 1000 ppm oil content feeds. The SEM images of the CTA FO membranes after backwashing can be seen in Figure S4.

Conclusions
In this work, a synthetic produced water was prepared to simulate the oilfield produced water in the south of Iraq to investigate the fouling behavior on the used membrane. As this type of feed is very salty, the FO process was used to purify water, whereas the conventional RO process became incapacitated. The feed was mixed with oil at different concentrations, 100 and 1000 ppm, that emulsified with the PW to mimic the real PW. Two types of membranes were used, namely: TFC and CTA membranes. Also, two draw solutions were compared, MgCl 2 and NH 4 HCO 3 of concentrations 4.8 M and 6 M, respectively. Findings showed a higher water flux when the MgCl 2 was applied as a draw solution which was assigned to its high osmotic pressure and low scaling impact. Also, a synergic influence for the oil and inorganic scaling was detected.
Although the SEM images proved that the fouling on the TFC membrane was more intensive than the CTA one, the TFC membrane resulted in higher water flux which could be attributed to the porous structure of its support layer. Future work should focus on actual testing samples of high salinity-produced water in long-term operation in the forward osmosis process.