The forward osmosis (FO) membrane was provided by Hydration Technology, Inc. (Albany, OR). The membrane had an asymmetric structure and made of Cellulose Triacetate (CTA) with embedded polyester screen support. The maximum operating temperature was 160˚F (71˚C), maximum transmembrane pressure was 10 psi and operating pH range was 3 to 8, according to the manufacturer. The membrane distillation (MD) membrane was asymmetric polypropylene (PP) membrane. It composed of pure polypropylene and offer broad chemical compatibility allowing its use with aqueous and organic solvent samples. It was obtained from GE Osmonics (Minnetonka, MN). The effective area for both FO and MD membranes were 20.02 cm² in the membrane cell.

The forward osmosis (FO) membrane module consisted of a custom-built cross-flow membrane cell with two channels for feed and draw solutions. Feed was the wastewater and draw was 1 M NaCl solution. The water transfer took place across the FO membrane due to the osmotic pressure gradient between the feed and draw solutions, and weight changes in the draw solution were monitored by a computer to record the FO flux. Because pure water was transported from feed to draw by the FO process, the volume of the feed tank decreased and the volume of draw solution increased over time. Two variable speed gear pumps (Micropump, Vancouver, WA) were used to generate cross-flows, forming separate closed loops for the feed and draw solutions. A constant cross-flow rate of 0.4 L/min was maintained among the two loops using general purpose acrylic flow-meter (Cole-Parmer, Vernon Hills, IL). The feed solution was completely mixed by a stirrer (Corning, Tewksbury, MA) to avoid settling of particles. The temperature of the feed and draw solutions were maintained at room temperature (20˚C ± 1˚C) by using a water bath (Neslab, Newington, NH). The configuration of the membrane distillation (MD) system was similar to the FO system. The two channels in the MD system were feed, which was the wastewater, and permeate, which was purified water. The feed was operated at an elevated temperature of 50˚C ± 1˚C, while the permeate was maintained at room temperature (20˚C ± 1˚C). Temperature gradient between the two channels transported the water vapor from feed to permeate, which then became condensed by the low temperature permeate solution. Eventually the feed volume reduced and the permeate volume increased by the MD process. Because the feed was operated at an elevated temperature, heavy-duty PTFE/PFA flow- meter (Cole-Parmer, Vernon Hills, IL) was used. In the investigation of combined FO-MD system, draw solution of the FO system was used as a feed in the MD system. The schematic diagrams of the FO and MD setups are shown in Figure 1.

A batch anaerobic digester was constructed of high density polyethylene (HDPE) 25 L brewer tanks (Hobby Beverage Equipment Company, Temecula, CA). The digester was covered with aluminium foil and temperature adjustable heating tape (Thermolyne, Columbia, MD) was placed on top of the foil. The aluminium foil was used to ensure even heat distribution to the digesters and to provide protection from the heating tape so that physical failure of the polyethylene would not occur. The temperature of the digester was maintained at 35˚C.

Gas mixing was applied to the anaerobic digester by circulating the headspace gas to the bottom of digesters using a 600 rpm modular drive peristaltic pump (Cole-Parmer, Vernon Hills, IL). The digester was equipped with a gas collection flask and Wet Tip gas meter (Nashville, TN) to measure total gas volume production. Figure 2 shows the laboratory setup of the 35˚C anaerobic digester system.

The anaerobic digester system was started up with seed sludge from Alexandria Renew Wastewater Treatment Plant in VA. Seed sludge is beneficial due to abundant presence of useful microorganisms such as methane formers and acid formers bacteria. After approximately 3 days of seed feeding, the digester system was fed with raw municipal sludge collected from Blue Plains Advanced Wastewater Treatment Plant (AWTP) in Washington DC. Nitrogen gas (N₂) was purged for 2 - 3 min to mix the digester before collecting uniform and homogenous effluent and providean anaerobic condition within the digester.

The effluent collected from the digester was dewatered by centrifuge using AccuSpin™ 400 (Fisher Scientific, Pittsburg, PA) at 4000 RPM for 10 min. The supernatant was diluted with tap water to obtain the desired concentration of 200 to 2000 mg/L. This centrifuged supernatant was used as feed solution for the FO, MD and combined FO-MD system. The process flow diagram of the feed preparation is illustrated in Figure 3.

The protocol to investigate the fouling effects of solids on the membrane surfaces for FO and MD systems comprised of the following steps. New membranes were placed in the FO system before the experiments. The feed and draw tank were filled with 2 L solutions. The system was run for 15 min in closed loop without passing flow through the membrane cell to stabilize the temperature. After 15 min, the feed and draw flows were allowed to the membrane cell by closing the bypass valve. The weight change in the draw tank was continuously monitored by a computer and the water flux through the FO membrane was recorded. For the MD system, the experimental protocol was similar to FO system, except the temperatures of the feed and permeate tanks were maintained at 50˚C and 20˚C, respectively. To determine the rejection of solids, NH₃-N and acetic acid, the system was run for 48 h before collecting the samples for measurements. Total solids (TS) were measured according to the Standard Methods 2540B [15] .

Ammonia nitrogen was measured using the titrimetric methods, according to Standard Methods 4500-NH₃ C. Preliminary distillation in Standard Methods 4500-NH₃ B was followed prior to titration with 0.02 N H₂SO₄ (Fisher Scientific, Pittsburg, PA) [15] . Ammonia measured using this technique is the total ammonia nitrogen (TAN) concentration. Acetic acid was measured using a Shimadzu Gas Chromatograph Model GC-2010 (Shimadzu, Columbia, MD) with flame ionization detector (FID). The rejections of different contaminants were calculated by measuring the concentration in the feed and permeate tanks after continuing the experiment for 48 h. Equation (1) was used for calculating the rejection efficiency, R.

where C_f is the initial concentration of the contaminant in the feed tank, and C_p is the concentrations of contaminant in the permeate tank after 48 h of experiment.

In the membrane cleaning experiment, solids deposition on the membrane surfaces was measured and analyzed. For membrane cleaning, the feed and draw (or permeate) tanks were switched with tap water and cross- flow rates were increased to 1.0 L/min. The system was run for 15 min to allow the increased cross-flow to remove foulants from the membrane surface. Solids deposition and solids removal percentages were calculated by weighing the new membrane, fouled membrane and cleaned membrane. The percentage of flux recovery for the cleaning experiment was calculated using Equation (2).

where J_v is the water flux in (m³/m²・s) and superscript N, F and C are used for new membrane, fouled membrane and cleaned membrane, respectively. For the long-term experiments to check the reusability of the FO membrane, the cleaning was performed for 15 min in every 24 h interval. New feed and draw solutions were put in the tanks before starting the experiments after cleaning.

The digester was operated at 35˚C (mesophilic temperature) with a hydraulic retention time of 10 d. Because the digester was kept completely mixed throughout the study, solids retention time (SRT) of the digester was equal to its hydraulic retention time (HRT). The digester was fed with 5.4% solids with organic loading rate of 4.22 kgVS/m³・d. The performance of the anaerobic digester was continuously monitored and total gas production per day of 31.2 L/d was observed with 68.4% methane content. The digestion process was able to remove 38% of total solids (TS) and 48% of volatile solids (VS) from the feed sludge. The effluent from the digester was dewatered by centrifuge at 4000 RPM for 10 min. Different parameters measured for the feed sludge, digested sludge and centrifuged supernatant are summarized in Table 1. The centrifuged supernatant was used as feed to the membrane systems for further treatment.

The supernatant from the centrifuge contained 2020 mg/L to total solids (TS). To investigate the potential of using coagulants on the centrifuge supernatant, jar test was performed with alum (Al₂(SO₄)₃・18H₂O) and ferric

chloride (FeCl₃). The dosages of coagulants were varied from 10 mg/L to 80 mg/L in the standard jar test apparatus to reduce the amount of solids. However, it was found to be ineffective in reducing solids content indicated by the final turbidity value and visual observations. The higher dosage of ferric chloride actually increased the final turbidity in some cases. The difficulties in reducing the total solids from the centrifuged supernatant might be due to the fact that the organic contents were not in particulate form, but in the form of emulsion and thereby coagulation-flocculation process was ineffective.

The sidestream characteristics are highly variable and depend on many factors such as influent composition, treatment process, equipment performance, chemical use and point of applications. Based on different liquid treatment or sludge handling processes, the total solids present in the sidestream varies largely, from 50 mg/L to 20,000 mg/L [16] . In order to investigate the variability of the total solids on membrane processes, the centrifuged supernatant were diluted with tap water and used as feed for both FO and MD systems. The feed solution concentrations were varied from 200 mg/L to 2000 mg/L. The effects of solids on FO and MD membranes are presented in Figure 4.

For both FO and MD membranes, the flux decline is proportional to the amount of solids present in the feed. The higher amount of solids caused membrane fouling faster and resulted in faster flux decline. To compare the effect of solids between FO and MD membranes, the percentages of flux decline were measured from the initial flux after 20 h for both cases and presented in Table 2. The results showed that for feed solutions with different solids content, flux decline in FO membranes were much lower than that of MD. The initial fluxes for FO membrane were proportional to the solids content in the feed. This is understandable as higher amounts of solids in the feed increased the feed concentration, thereby reduced the osmotic pressure difference across the membrane surface and resulted in lower initial flux. On the other hand, initial fluxes for MD membranes were relatively constant because the partial vapor pressure of water is only minimally affected by increased concentration in the feed [17] . Overall, the fouling effect was found to be more severe in MD membrane as compared to FO membrane.

Centrifuged supernatants from the anaerobic digester, with total solids content of 2020 mg/L, caused significant fouling in both FO and MD membranes. Therefore feed solids concentration of 1000 mg/L, 400 mg/L and 200 mg/L were used in the rejection experiment. The rejection of total solids, NH₃-N and acetic acid were measured for the FO, MD and combined FO-MD system.

The centrifuged effluent from anaerobic digester contained approximately 2020 mg/L of total solids, 1485 mg/L of NH₃-N and 92 mg/L of acetic acid. The organic contents in the solution were in the form of emulsion and could not be further reduced by using coagulant like alum or ferric chloride. However, the combined FO-MD

system was able to achieve almost 100% rejection of total solids. For the rejection of NH₃-N, FO-MD system was more than 98% efficient when the solids content in the feed was less than 1000 mg/L, as shown in Table 3. Investigating the rejection for FO and MD separately revealed that the higher rejection in the FO-MD system was mainly achieved due to the FO process. For MD, the rejection was approximately 62% - 72%. The poor rejection is due to the fact that higher solids (organic content) in the feed caused wetting problem in the MD and allowed the contaminants to pass through. The volatile nature of the contaminants at pH 8 can also contribute to the rejection efficiency, as MD process allowed volatile matters to pass through.

High rejection (>98%) of acetic acid was also obtained by both FO and MD processes. The pH of the centrifuged supernatants was higher than 8.0. At this pH, the acetic acid was mainly in the form of negatively charged acetate ion in the solution (pKa value of 4.7). The higher pH have increased the negative charged surface density of the membranes, which increased the electrostatic interaction between the membrane surface and negative charged acetate ion, thus resulting in high rejection. The combined FO-MD process achieved almost 100% rejection of acetic acid.

Fouling in MD membrane was significantly high while treating sidestream from anaerobic digester. In the combined FO-MD system, FO was primarily used to treat the sidestream, while MD was regenerating the draw solution. Therefore, fouling was mostly associated with FO.

FO process is an ambient pressure operation. Foulants in FO membrane are loosely attached to the membrane surface and can be cleaned with tap water [18] . For the cleaning experiments, 15 min cleaning were performed after running the tests for 1 d. Figure 5 showed the measurement of weights for new membranes, fouled membranes and cleaned membranes with different feed concentrations. Table 4 presents the solids removal and flux recovery from the cleaning experiments. The results showed that 44% to 60% of the solids from membrane surface were removed by cleaning with tap water for 15 min, which resulted in a flux recovery of as high as 84%. The results showed great potential of reusing the FO membrane with continuous cleaning, which can significantly reduce the operating cost of the process.

To investigate the reusability of the FO membrane, 200 mg/L feed was used for a 4-d experiment with continuous cleaning. Membrane cleaning was performed for 15 min in every 24 h interval. The results in Figure 6

show that continuous flux recovery can be achieved by cleaning the membrane. The results also demonstrated that percentage flux recovery deteriorated with every cleaning. When comparing the results with a similar experiment without continuous cleaning, the FO membrane performed better in terms of water flux with cleaning. In this case, 961.2 mL of water was recovered by FO process with cleaning, as compared to 725.1 mL of water without cleaning, resulting 33% higher water recovery when the FO membrane was cleaned in every 24 h interval.