A Systematic Study of Ammonia Recovery from Anaerobic Digestate Using Membrane-Based Separation

Ammonia recovery from synthetic and real anaerobic digestates was accomplished using hydrophobic flat sheet membranes operated with H2SO4 solutions to convert ammonia into ammonium sulphate. The influence of the membrane material, flow rate (0.007, 0.015, 0.030 and 0.045 m3 h−1) and pH (7.6, 8.9, 10 and 11) of the digestate on ammonia recovery was investigated. The process was carried out with a flat sheet configuration at a temperature of 35 °C and with a 1 M, or 0.005 M, H2SO4 solution on the other side of the membrane. Polytetrafluoroethylene membranes with a nominal pore radius of 0.22 µm provided ammonia recoveries from synthetic and real digestates of 84.6% ± 1.0% and 71.6% ± 0.3%, respectively, for a membrane area of 8.6 × 10−4 m2 and a reservoir volume of 0.5 L, in 3.5 h with a 1 M H2SO4 solution and a recirculation flow on the feed side of the membrane of 0.030 m3 h−1. NH3 recovery followed first order kinetics and was faster at higher pHs of the H2SO4 solution and recirculation flow rate on the membrane feed side. Fouling resulted in changes in membrane surface morphology and pore size, which were confirmed by Atomic Force Microscopy and Air Displacement Porometry.


Introduction
A change in the perception of the uses of anaerobic digestion has occurred over past decades. Thus, anaerobic digestion was initially considered as a cost-competitive technology for organic matter stabilization in wastewaters and solid waste, then a sustainable platform for renewable electricity and heat generation via biogas production, and more recently, it has been regarded as the potential core of a multiproduct biorefinery. Energy and carbon in the form of biogas, along with nutrients dissolved in anaerobic effluents (typically called digestates), represent nowadays the main by-products from anaerobic waste treatment. NH 3 recovery from digestates will prevent the pollution of natural water bodies, which is desirable as this nitrogenous compound is toxic to fish, increases oxygen demand and induces eutrophication [1]. NH 3 is also harmful to humans as it causes respiratory problems and is considered a precursor of N 2 O, a potent greenhouse gas [2][3][4]. Therefore, NH 3 recovery from digestates is crucial to improve the environmental sustainability of anaerobic digestion processes and could bring additional economic benefits. Unfortunately, only 10% of the nitrogen present in wastewater is recovered by conventional wastewater treatment plants (WWTPs) [5][6][7]. Today, there is a wide range of commercial technologies to remove ammonia from wastewater, but most of them entail high operating costs [8] and environmental impacts resulting from nitrogen conversion and release to the atmosphere. Thus, conventional physical/chemical technologies are based on selective ion exchange, air stripping, chemical precipitation, adsorption, advanced oxidation processes, breakpoint chlorination, etc. [9][10][11][12]. For instance, NH 3 stripping ranks among the most popular, but energy demanding, technologies available to remove nitrogen from high-strength wastewater, with an average energy consumption of 4 kwh per kg of nitrogen [13]. On the other hand, conventional nitrogen removal in domestic WWTPs relies on nitrification, which is the conversion of ammoniacal nitrogen (NH 3 -N) to NO − 3 , and denitrification, which is the conversion of NO − 3 to N 2 gas [14][15][16]. In this context, 1.5 kWh of electricity are typically required to remove 1 kg of N in denitrification/nitrification or anammox processes [17]. In addition, conventional biological nitrogen removal in WWTPs entails 0.9 kg CO 2 eq per m 3 of treated water [18,19]. Therefore, there is an urgent need to develop and implement cost-competitive and sustainable NH 3 recovery technologies in high strength wastewater such as digestates [20][21][22].
Recently, the use of membranes for NH 3 recovery from high strength wastewater has been proposed as a cost-competitive and environmentally friendly approach to upgrade residual nitrogen. For example, Brennan et al. [23] made an extensive analysis of costs demonstrating the feasibility of NH 3 recovery by membranes over conventional methods at a pilot plant scale. Membrane contactors require less energy because of their high specific surface area, which provides a faster NH 3 separation. In this context, membrane contactors have been successfully implemented for NH 3 recovery from gaseous emissions [24][25][26]. Membrane contactors have also been implemented to support a direct NH 3 recovery from wastewater [27][28][29]. This approach is based on hydrophobic membranes made of polymers highly permeable to NH 3 , which is transported through the membrane pores into the receiving liquid phase recirculating through the permeate side [20,21,30]. This permeating NH 3 reacts with an acid contained in the permeate side of the membrane. For the process to be efficient, ammoniacal nitrogen has to be present in its volatile form (NH 3 ), which can be ensured by increasing pH and temperature [22]. Ammonium sulphate, a commercial chemical fertilizer, can be generated during membrane-based NH 3 recovery when using H 2 SO 4 to boost NH 3 diffusion (Equation (1)) [31]. The use of H 3 PO 4 and HNO 3 on the permeate side also results in the generation of chemical fertilizers such as ammonium phosphate and ammonium nitrate, respectively [27][28][29]. In this regard, Damtie and coworkers concluded that H 2 SO 4 mediates a more effective ammonia capture than H 3 PO 4 and HNO 3 [32].
In this work, the performance of commercial flat sheet membranes for NH 3 recovery from synthetic and real digestates was systematically evaluated. To optimize the operation conditions, the influence of digestate pH, within the range of coexisting NH 4 + , NH 3 [33], and recirculation flow rate, was investigated using different concentrations of sulfuric acid on the other side of the membrane. To make this new ammonia process accessible, our proposal involves the use of commercial membranes that were previously employed for other applications. By scanning the variables that can be controlled at the place of installation, we reached the optimal conditions, without using the latest generation and non-commercial materials. Ultimately, our objective was to propose a solution within the reach of the farmer who intends to implement it in their facilities. Finally, membrane fouling was studied using Atomic Force Microscopy (AFM) to examine the surface morphology of the membranes and Air Displacement Porometry (ADP) to determine the corresponding changes of the pore sizes of the membranes.

Real and Synthetic Digestate Characterization
The synthetic digestate (SD) composition mimicked a real digestate and consisted of 5.0 g NaHCO 3 , 0.85 g C 8 H 5 KO 4 , 0.73 g peptone from casein, 1.70 g NH 4 Cl, 0.90 g CO(NH 2 ) 2 , 0.224 g K 2 HPO 4 , 0.0175 g NaCl, 0.01 Ca 2 Cl and 0.005 g MgSO 4 (per litre of distilled water) [34]. All reagents were purchased from PANREAC (Panreac, Química SAU, Barcelona, Spain). The real digestate (RD) used in this study was supplied by Valladolid WWTP (Spain). This digestate was obtained from the anaerobic digestion of mixed sludge and on-site centrifuged prior to use to eliminate suspended solids. The concentration of NH 3 in SD and RD averaged 605.8 ± 0.1 and 678.9 ± 0.6 ppm, respectively, while pHs averaged 7.57 ± 0.05 and 8.99 ± 0.08, respectively.

Experimental Setup
A schematic representation of the experimental setup used for ammonia recovery from digestates is shown in Figure 1. The target digestate (SD or RD) was continuously circulated using a peristaltic pump (Watson-Marlow Sci-Q 323, Spirax-Sarco Engineering plc, Cheltenham, England, UK) over the active layer of the membrane in a customized cell module [35]. A sulfuric acid solution was recirculated using a similar peristaltic pump on the support layer of the membrane. Both digestate and sulfuric acid solutions were maintained at 35 • C in a thermostatic bath (HAAKE type E12, Thermo Fisher Scientific, Waltham, MA, USA) in 0.5 L enclosed Erlenmeyer bottles. 0.224 g K2HPO4, 0.0175 g NaCl, 0.01 Ca2Cl and 0.005 g MgSO4 (per litre of distilled water) [34]. All reagents were purchased from PANREAC (Panreac, Química SAU, Barcelona, Spain). The real digestate (RD) used in this study was supplied by Valladolid WWTP (Spain). This digestate was obtained from the anaerobic digestion of mixed sludge and on-site centrifuged prior to use to eliminate suspended solids. The concentration of NH3 in SD and RD averaged 605.8 ± 0.1 and 678.9 ± 0.6 ppm, respectively, while pHs averaged 7.57 ± 0.05 and 8.99 ± 0.08, respectively.

Experimental Setup
A schematic representation of the experimental setup used for ammonia recovery from digestates is shown in Figure 1. The target digestate (SD or RD) was continuously circulated using a peristaltic pump (Watson-Marlow Sci-Q 323, Spirax-Sarco Engineering plc, Cheltenham, England, UK) over the active layer of the membrane in a customized cell module [35]. A sulfuric acid solution was recirculated using a similar peristaltic pump on the support layer of the membrane. Both digestate and sulfuric acid solutions were maintained at 35 °C in a thermostatic bath (HAAKE type E12, Thermo Fisher Scientific, Waltham, MA, USA) in 0.5 L enclosed Erlenmeyer bottles.  Table 1. Polytetrafluoroethylene (PTFE) and polyvinylidene difluoride (PVDF) were selected as model commercial membranes based on their ideal properties for the separation of light molecules from water in a gaseous phase.  The membranes used and their main characteristics are shown in Table 1. Polytetrafluoroethylene (PTFE) and polyvinylidene difluoride (PVDF) were selected as model commercial membranes based on their ideal properties for the separation of light molecules from water in a gaseous phase.

Operational Conditions and Process Evaluation
An initial test series was carried out with all the membranes initially selected (Table 1)  and SD and the PTFE-0.22 membrane, at a pH of 10, with a sulfuric acid concentration of 1 M and recirculation flow rates of 0.007, 0.015, 0.030 and 0.045 m 3 h −1 . The membrane system was washed twice for 1 h after each experiment with distilled water when using SD, and with tap water (which contains sodium hypochlorite) when using RD. Samples of the digestate solution were drawn every 30 min over 3.5 h to analyse the concentration of NH 3 and obtain pH values. Finally, a systematic evaluation of membrane fouling was performed by determining the pore size distribution of PTFE-0.22 membranes after their use with SD and RD, at a pH of 10 and with a sulfuric acid concentration of 1 M at different recirculation flow rates (0.007, 0.015, 0.030 and 0.045 m 3 h −1 ). All experiments were carried out in duplicate.

Analytical Methods
The values of pH and temperature were monitored using a HI5522 pH meter Hanna Instruments (Woonsocket, RI, USA) and a Basic 20 pH meter (Crison Instruments, S.A., Alella, Barcelona, Spain) in the liquids on both sides of the membrane. NH 3 was measured using the Nessler analytical method in a Shimadzu UV-160 A spectrophotometer at 425 nm (Shimadzu, Kyoto, Japan).

Membrane Characterization Techniques
Surface morphology was analysed by using Atomic Force Microscopy. Images were obtained with a Nanoscope IIIA microscope (Digital Instruments, Veeco Metrology Group, Chadds Ford, PA, USA) using the Tapping mode. Pore size distribution was analysed by the extended bubble point method, or Air Displacement Porometry (ADP), using a Coulter ® Porometer-II manufactured by Coulter Electronics (Porometer, Aptco Invest, Dulles, VA, USA) [36]. Samples were first wetted with an electronic liquid FC-43 (Fluorinert TM , 3 M, St. Paul, MN, USA) of low surface tension (γ 1.6 × 10 −2 N m −1 ), low vapour pressure (192 Pa) and low reactivity, that can be assumed to fill all the pores given a zero contact-angle with the membrane material. The wetted sample was subjected to increasing pressure, applied by a compressed clean and dry air source. The pressure range for PTFE 0.22 membrane was 0.09-0.60 µm.

Theoretical Methods
The overall mass transfer coefficient of NH 3 was calculated with Equation (2). according to [21].
where k s and k m are the mass transfer coefficients in the digestate side and within pores, respectively. The resistance on the acid solution side can be considered negligible (i.e., mass transfer coefficient on the acid side is much larger than k s and k m ). In this context, the mass transfer coefficient on the digestate side can be estimated using Equation (3): where D AW , Sh and D H are: diffusion coefficient of NH 3 in water (calculated using the software ASPEN (AspenTech, Bedford, MA, USA) at 35 • C), the Sherwood dimensionless number [37], and the hydraulic diameter, respectively. Similarly, the mass transfer coefficient within pores can be calculated by Equation (4) [26,38]: where τ, δ, D ij and are: the tortuosity, the wall thickness of the membrane, the diffusion coefficient of NH 3 in the air gap within pores, and the porosity of the membrane, respectively. Accepting Knudsen regime [21]: 8RT πm (5) in terms of the pore diameter and NH 3 molar mass m. The molar theoretical flux of NH 3 across the membrane can be estimated using Equation (6) [39]: here, ∆C is the difference in NH 3 concentrations between the feed and permeate sides. Finally, NH 3 recovery seems to follow first-order kinetics, which can be estimated using Equation (7) [40,41]: where k ov , A m /V t and t are the overall mass transfer coefficient, the effective ratio of membrane area to feed volume of the digestate, and the elapsed time, respectively. C 0 and C t are the concentration of ammonia at time zero and time t in the digestate, respectively. NH 3 recovery (R) ranges between 0 and 1 (Equation (8)): here α = k ov A m V t is the rate of concentration decrease and recovery increase. Therefore, the time needed to reach a given R can be estimated using Equation (9):

Influence of Membrane Material, pH and H 2 SO 4 Concentration
Five flat sheet membranes (Table 1) were investigated at a recirculation flow rate of 0.030 m 3 h −1 , a pH of 10 and a concentration of H 2 SO 4 of 1 M, to elucidate the optimum material, pore size and wettability for NH 3 recovery from a synthetic digestate. PTFE membranes exhibited a superior performance to PVDF membranes in terms of NH 3 recovery at a pH of 10 ( Figure 2). In this case, the highest NH 3 removal recovery after 3.5 h (77.7% ± 0.4%) was provided by PTFE-0.22. This removal efficiency was 3.6-fold higher than those obtained for the two PVDF membranes tested. As shown in Table 1, PTFE-0.22 has a contact angle (θ) of 150 degrees [42], which corresponds to a rather hydrophobic surface [43]. In fact, a membrane with θ > 140 • could be considered super-hydrophobic according to Tylkowski et al. [43]. The superior performance of the PTFE-0.22 membrane could be also explained by its asymmetric structure, as membranes with symmetric structures, with similar thickness, have been consistently shown to be less permeable [20]. Therefore, the following experiments were also conducted using PTFE-0.22 membranes. Membranes 2022, 11, x FOR PEER REVIEW 6 of 14 The influence of the digestate pH was also studied. This parameter determines NH3 mass transfer as it shifts the equilibrium NH ↔ NH + H . In our study, higher ammonia recoveries were achieved by increasing the pH of the synthetic digestate because of the increase in the NH3 concentration gradient ( Figure S1 in the Supplementary material). Nevertheless, an increase in the pH of the digestate above 10 does not entail large improvement in the ammonia removal [20]. Operating the process at a pH of 11 with a PTFE-0.22 membrane, a recirculation flow rate of 0.030 m 3 h −1 and a concentration of H2SO4 of 1 M, resulted in a NH3 recovery of 82.8% ± 0.4%, only 10% higher than the recoveries recorded at a pH of 10. NH3 recovery was lower in real digestate than in synthetic digestate using a PTFE-0.22 membrane at pHs of 10 and 7.6, the latter pH supporting lower NH3 recoveries ( Figure S1 in the Supplementary material). Thus, the times needed to reach 95% NH3 recovery significantly increased by a factor of 6.5 and 2.3 in SD and RD, respectively, when digestate pH was increased from 7.6 to 10 ( Table 2). H2SO4 concentration was identified, also, as a key operational parameter for NH3 recovery. The performance of a PTFE-0.22 membrane at two concentrations of H2SO4 (0.005 and 1 M) was studied at pHs of 7.6 and 10, under a flow rate of 0.030 m 3 h −1 using synthetic digestate. Acid concentrations of 0.005 M resulted in NH3 recoveries of 12.2% ± 0.2% and 55.5% ± 0.1% at a pH of 7.6 and 10, respectively, after 3.5 h of operation at the same flow rate. Similarly, under the same conditions, NH3 recoveries of 22.9% ± 0.4% and 77.7% ± 0.4% were achieved using H2SO4 concentrations of 1 M (Figure 3). Therefore, it was concluded that the best operating conditions correspond to a pH of 10 and H2SO4 concentrations of 1 M. The influence of the digestate pH was also studied. This parameter determines NH 3 mass transfer as it shifts the equilibrium NH + 4 ↔ NH 3 + H + . In our study, higher ammonia recoveries were achieved by increasing the pH of the synthetic digestate because of the increase in the NH 3 concentration gradient ( Figure S1 in the Supplementary material). Nevertheless, an increase in the pH of the digestate above 10 does not entail large improvement in the ammonia removal [20]. Operating the process at a pH of 11 with a PTFE-0.22 membrane, a recirculation flow rate of 0.030 m 3 h −1 and a concentration of H 2 SO 4 of 1 M, resulted in a NH 3 recovery of 82.8% ± 0.4%, only 10% higher than the recoveries recorded at a pH of 10. NH 3 recovery was lower in real digestate than in synthetic digestate using a PTFE-0.22 membrane at pHs of 10 and 7.6, the latter pH supporting lower NH 3 recoveries ( Figure S1 in the Supplementary material). Thus, the times needed to reach 95% NH 3 recovery significantly increased by a factor of 6.5 and 2.3 in SD and RD, respectively, when digestate pH was increased from 7.6 to 10 ( Table 2). H 2 SO 4 concentration was identified, also, as a key operational parameter for NH 3 recovery. The performance of a PTFE-0.22 membrane at two concentrations of H 2 SO 4 (0.005 and 1 M) was studied at pHs of 7.6 and 10, under a flow rate of 0.030 m 3 h −1 using synthetic digestate. Acid concentrations of 0.005 M resulted in NH 3 recoveries of 12.2% ± 0.2% and 55.5% ± 0.1% at a pH of 7.6 and 10, respectively, after 3.5 h of operation at the same flow rate. Similarly, under the same conditions, NH 3 recoveries of 22.9% ± 0.4% and 77.7% ± 0.4% were achieved using H 2 SO 4 concentrations of 1 M (Figure 3). Therefore, it was concluded that the best operating conditions correspond to a pH of 10 and H 2 SO 4 concentrations of 1 M.

Influence of the Digestate Recirculation Flow Rate
Ammonia removal efficiencies of 84.6% ± 1.0% were achieved using the PTFE-0.22 membrane after 3.5 h of experiment in synthetic digestate, which were higher than the removal efficiencies in real digestate (71.6% ± 0.3%), at 0.045 m 3 h −1 and a pH of 10. The lower NH3 recoveries in real digestate were likely due to membrane fouling. This fouling affected the NH3 permeate flux mainly due to adsorption of colloids, or solutes including ammonia, within the membrane pores, with a subsequent reduction in mean pore sizes. According to Meng et al. [44], membrane fouling typically causes cake layer formation, deposition of sludge flocs and their deposition on the membrane. The influence of the velocity of the acid solution was not herein assessed as previous literature has consistently shown that the velocity of the acid solutions does not impact significantly on NH3 mass transfer [20,31].
The increase in the digestate recirculation flow rate brought about a rise in the pace of the NH3 recovery across the PTFE-0.22 membrane regardless of the type of digestate (real or synthetic). Figure S2a,b (in Supplementary Material) show the exponential decrease of the ammonia concentration, which resulted in a rapid increase of NH3 recovery (Equation (7)). In Figure 4a, the corresponding rate of concentration decrease and recovery increase α, is shown as a function of the recirculation flow rate of SD and RD (Equation (8)). The value of α was higher for the synthetic digestate, while the increase in digestate recirculation flow rate was positively correlated with NH3 recovery rate regardless of the type of digestate. Figure 4b shows the evolution of the time needed to reach 95% recovery (Equation (9)). This time is higher for the real digestate than for the synthetic one. It is worth noting that an increase of the recirculation flow of the digestate from 0.007 to 0.045 m 3 h −1 brought about a substantial decrease of the time needed to reach 95% recovery of NH3. In all cases, a pH of 10 was maintained. The duration of the experiment conducted at a liquid recirculation rate of 0.03 m 3 h −1 was extended to 6 h, in order to check how a longer time improved recovery. This experiment resulted in a final NH3 recovery of 94.3% ± 0.3% in synthetic digestate in accordance with Figure 4b. Therefore, we concluded that over long time intervals, the evolution of recovery follows the pace assumed here.

Influence of the Digestate Recirculation Flow Rate
Ammonia removal efficiencies of 84.6% ± 1.0% were achieved using the PTFE-0.22 membrane after 3.5 h of experiment in synthetic digestate, which were higher than the removal efficiencies in real digestate (71.6% ± 0.3%), at 0.045 m 3 h −1 and a pH of 10. The lower NH 3 recoveries in real digestate were likely due to membrane fouling. This fouling affected the NH 3 permeate flux mainly due to adsorption of colloids, or solutes including ammonia, within the membrane pores, with a subsequent reduction in mean pore sizes. According to Meng et al. [44], membrane fouling typically causes cake layer formation, deposition of sludge flocs and their deposition on the membrane. The influence of the velocity of the acid solution was not herein assessed as previous literature has consistently shown that the velocity of the acid solutions does not impact significantly on NH 3 mass transfer [20,31].
The increase in the digestate recirculation flow rate brought about a rise in the pace of the NH 3 recovery across the PTFE-0.22 membrane regardless of the type of digestate (real or synthetic). Figure S2a,b (in Supplementary Material) show the exponential decrease of the ammonia concentration, which resulted in a rapid increase of NH 3 recovery (Equation (7)). In Figure 4a, the corresponding rate of concentration decrease and recovery increase α, is shown as a function of the recirculation flow rate of SD and RD (Equation (8)). The value of α was higher for the synthetic digestate, while the increase in digestate recirculation flow rate was positively correlated with NH 3 recovery rate regardless of the type of digestate. Figure 4b shows the evolution of the time needed to reach 95% recovery (Equation (9)). This time is higher for the real digestate than for the synthetic one. It is worth noting that an increase of the recirculation flow of the digestate from 0.007 to 0.045 m 3 h −1 brought about a substantial decrease of the time needed to reach 95% recovery of NH 3 . In all cases, a pH of 10 was maintained. The duration of the experiment conducted at a liquid recirculation rate of 0.03 m 3 h −1 was extended to 6 h, in order to check how a longer time improved recovery. This experiment resulted in a final NH 3 recovery of 94.3% ± 0.3% in synthetic digestate in accordance with Figure 4b. Therefore, we concluded that over long time intervals, the evolution of recovery follows the pace assumed here.  The NH3 fluxes across the PTFE-0.22 membrane in both SD and RD are shown in Figure 5. They decreased as the recirculation flow rate increased because of fouling, probably induced by shear. In addition, NH3 fluxes were slightly higher in SD than in RD according to its lower induction of membrane fouling. By fitting theoretical to experimental fluxes and using Equations (2)-(6), the corresponding effective pore diameters were evaluated at the end of each experiment, as a function of the recirculation flow Q, which was 0.25 µm. The NH 3 fluxes across the PTFE-0.22 membrane in both SD and RD are shown in Figure 5. They decreased as the recirculation flow rate increased because of fouling, probably induced by shear. In addition, NH 3 fluxes were slightly higher in SD than in RD according to its lower induction of membrane fouling. By fitting theoretical to experimental fluxes and using Equations (2)-(6), the corresponding effective pore diameters The experimental NH3 fluxes herein obtained with flat plate membranes were parable to, but clearly higher than, those reported in the literature when using simila with hollow fibre or tubular membranes (Table 3). In contrast, the overall mass transfer coefficient increased at increasing digesta circulation flow rates. An analysis of the relative importance of the membrane mass t fer coefficient km, and that on the digestate side ks, compared to the overall mass tra coefficient kov, show that the largest restriction to mass transfer occurred on the dige The experimental NH 3 fluxes herein obtained with flat plate membranes were comparable to, but clearly higher than, those reported in the literature when using similar pHs with hollow fibre or tubular membranes (Table 3). In contrast, the overall mass transfer coefficient increased at increasing digestate recirculation flow rates. An analysis of the relative importance of the membrane mass transfer coefficient k m , and that on the digestate side k s , compared to the overall mass transfer coefficient k ov , show that the largest restriction to mass transfer occurred on the digestate side of the membrane, while through the pores, there was a very high NH 3 mass transfer. NH 3 mass transfer through the pores decreased steeply when the recirculation flow rate was increased, but with a concomitant increase of the mass transfer on the digestate side of the membrane. Values of Table 4 correspond to synthetic digestate.

Pore Size
The characteristics of the PTFE-0.22 membrane were studied by air displacement porometry after the operation of the process under different experimental conditions (pH of 10, concentration of H 2 SO 4 1 M, recirculation flow rates 0.007, 0.015, 0.030 and 0.045 m 3 h −1 ) using both synthetic digestate and real digestate. The increase in membrane fouling reduced the pore diameter, as a result of deposits that partially or totally blocked the membrane pores (Table 5) [47]. A higher reduction in pore diameter was observed when the PTFE-0.22 membrane was exposed to longer periods of filtration, lower recirculation rates and real digestate from Valladolid WWTP. Interestingly, the mean pore size was similar in the new membrane and in the membranes used twice with recirculation flow rates of 0.045 m 3 h −1 (0.3548 µm ± 0.0004 and 0.3521 µm ± 0.0004, respectively). This behaviour is in accordance with more specific fouling studies for PTFE in literature [48,49].  Figure 6a shows three-dimensional AFM pictures of PTFE-0.22 membranes that were unused or used under multiple operational conditions. More fouling was detected on the surfaces of the membrane operated with real digestate at three recirculation flow rates. The membranes operated with real digestate at the highest circulation flow rate presented a surface fouling level in agreement with the pore size results ( Table 4). The unused membrane has a topography with higher roughness than for the used membranes, in accordance with Zhang et al. [50]. In addition, the AFM topographic images of the support layer had a very similar appearance in all membranes tested, which suggested that the use of H2SO4 solutions did not significantly deteriorate the membrane surface (Figure 6b).

Membrane Morphology Analysis
In addition, phase contrast images were obtained to detect the presence of materials coating the membrane surface ( Figure S3 in Supplementary Material). The deeper brown tones detected in the membrane, used with real digestate and at three recirculation flow rates, would correspond to organic matter from the real digestate. Fouling was not significant in the rest of the samples and the coating effect was more homogeneous. In addition, the AFM topographic images of the support layer had a very similar appearance in all membranes tested, which suggested that the use of H 2 SO 4 solutions did not significantly deteriorate the membrane surface (Figure 6b).
In addition, phase contrast images were obtained to detect the presence of materials coating the membrane surface ( Figure S3 in Supplementary Material). The deeper brown tones detected in the membrane, used with real digestate and at three recirculation flow rates, would correspond to organic matter from the real digestate. Fouling was not significant in the rest of the samples and the coating effect was more homogeneous.

Conclusions
The feasibility of several commercial flat sheet membranes for ammonia recovery and their fouling were studied using synthetic and real digestate. PTFE was identified as the most efficient material for NH 3 recovery regardless of the type of digestate. The highest NH 3 recovery was obtained when the pH of the digestate was increased to 10 and 1 M sulfuric acid was used. NH 3 recovery also increased at higher digestate circulation flow rates because of the inherent reduction in mass transfer resistance. NH 3 recovery was shown to follow first order kinetics and to be faster under alkaline pH, high H 2 SO 4 concentration and high recirculation flow rate on the digestate side of the membrane. A PTFE-0.22 membrane operated under optimal conditions supported NH 3 recoveries of 71.6% ± 0.3% and 84.6% ± 1.0% after only 3.5 h using real and synthetic digestates, respectively. After 6 h of operation, 94.3% ± 0.3% NH 3 recoveries were reached in synthetic digestate under similar operational conditions in accordance with the kinetics assumed here. Membrane fouling was relevant when using real digestate and resulted in a decrease in membrane pore diameter and growing surface deposition.