Long-Term Robustness and Failure Mechanisms of Electrochemical Stripping for Wastewater Ammonia Recovery

Abstract

Nitrogen in wastewater has negative environmental, human health, and economic impacts but can be recovered to reduce the costs and environmental impacts of wastewater treatment and chemical production. To recover ammonia/ammonium (total ammonia nitrogen, TAN) from urine, we operated electrochemical stripping (ECS) for over a month, achieving 83.4 ± 1.5% TAN removal and 73.0 ± 2.9% TAN recovery. With two reactors, we recovered sixteen 500-mL batches (8 L total) of ammonium sulfate (20.9 g/L TAN) approaching commercial fertilizer concentrations (28.4 g/L TAN) and often having >95% purity. While evaluating the operation and maintenance needs, we identified pH, full-cell voltage, product volume, and water flux into the product as informative process monitoring parameters that can be inexpensively and rapidly measured. Characterization of fouled cation exchange and omniphobic membranes informs cleaning and reactor modifications to reduce fouling with organics and calcium/magnesium salts. To evaluate the impact of urine collection and storage on ECS, we conducted experiments with urine at different levels of dilution with flush water, extents of divalent cation precipitation, and degrees of hydrolysis. ECS effectively treated urine under all conditions, but minimizing flush water and ensuring storage until complete hydrolysis would enable energy-efficient TAN recovery. Our experimental results and cost analysis motivate a multifaceted approach to improving ECS’s technical and economic viability by extending component lifetimes, decreasing component costs, and reducing energy consumption through material, reactor, and process engineering. In summary, we demonstrated urine treatment as a foothold for electrochemical nutrient recovery from wastewater while supporting the applicability of ECS to seven other wastewaters with widely varying characteristics. Our findings will facilitate the scale-up and deployment of electrochemical nutrient recovery technologies, enabling a circular nitrogen economy that fosters sanitation provision, efficient chemical production, and water resource protection.

1. Introduction

Nitrogen in wastewater has negative environmental, human health, and economic impacts.¹⁻³ However, nitrogen as ammonia, which occurs in high concentrations in streams such as hydrolyzed (i.e., ureolyzed or after urea hydrolysis) urine,^4,5 can be harnessed as a valuable resource.⁶ In fact, nitrogen excreted via urine could replace about 15% of nitrogen fertilizers worldwide.^7,8 Incumbent municipal wastewater nitrogen removal requires substantial chemical and energy inputs, with wastewater treatment accounting for about 2% of electrical energy consumption in the United States.^9,10 Meanwhile, ammonia synthesis via the Haber-Bosch process accounts for about 1% of global fossil fuel energy consumption¹¹ and 1.4% of global carbon dioxide emissions.¹² Recovering nitrogen from wastewater, particularly from highly concentrated streams like urine, for beneficial reuse in a circular economy can reduce the environmental impacts and costs of wastewater treatment and chemical production.

Electrochemical processes comprise an emerging type of wastewater treatment technology that is well-suited for nutrient (e.g., nitrogen) recovery but requires further study.^13,14 Electrochemical processes can easily be scaled to treat varying flow rates and facilitate in situ production of chemicals, enabling decentralized treatment.¹⁵⁻¹⁸ While facilitating targeted total ammonia nitrogen [TAN; the sum of ammonia (NH₃) and ammonium (NH₄⁺)] recovery, some electrochemical treatment processes can simultaneously achieve other benefits, such as pharmaceutical removal, electricity generation, and pathogen disinfection.^13,15 Many electrochemical ammonia recovery technologies, including electrodialysis,^19,20 capacitive deionization,^21,22 and electrochemical stripping (ECS),²³ have been demonstrated at lab scale. Batch and continuous ECS treatment of urine has been demonstrated,²³ along with batch treatment of effluent from wastewater algae hydrothermal liquefaction²⁴ and anaerobic digestate from fecal sludge.²⁵ ECS application across a range of influent ammonium concentrations (30 to 3,000 mg/L TAN) also appears feasible based on experiments and an accompanying process performance model.²⁵ When treating urine, ECS recovers an ammonium sulfate product with negligible pharmaceutical contamination based on target analysis of 11 compounds.²³ These promising advances in electrochemical nutrient recovery motivate further investigation to accelerate technologies toward real-world application.

Several process engineering gaps must be addressed to facilitate scale-up and widespread implementation of electrochemical nutrient treatment technologies, including ECS. Common barriers include integration into treatment trains; product characteristics and acceptable applications; optimization of operating conditions, reactor configuration, and materials; and gaps between experimental data and modeling efforts related to life cycle and cost evaluation.^15,26 Moreover, long-term studies evaluating robustness in realistic operating environments are limited for electrochemical nutrient technologies compared to other nutrient technologies and electrochemical technologies. In terms of electrochemical nutrient technologies, electrochemical magnesium dosing for struvite (NH₄MgPO₄·6H₂O) recovery from urine has been applied to urine for 4 days,²⁷ a hydrogen-recycling electrochemical system coupled with membrane stripping has been tested for TAN recovery from urine for 7 days,²⁸ and electrodialysis has been demonstrated for anaerobic digestate from solids processing at a wastewater treatment plant for 3 days.²⁹ In comparison, other nutrient technologies, such as nitrification–distillation, have been operated for several months to 3.5 years in studies and longer in practice.³⁰⁻³² Electrochemical technologies with non-nutrient targets (e.g., electrochemical oxidation of wastewater) have also been demonstrated for about one month.³³ To date, ECS has been evaluated for up to 24 h with constant influent composition,²³ demonstrating the need for evaluation during longer operation.

Long-term investigations address barriers facing electrochemical nutrient recovery technologies by providing insights into reactor component degradation, failure mechanisms (e.g., ion exchange membrane fouling,³⁴ electrode inactivation³⁵⁻³⁸), and performance changes with longer operation. Such studies elucidate the impact of influent composition on performance, as well as pre- and post-treatment processes and maintenance tasks needed to sustain consistent operation and effluent quality despite varying influents. Long-term operation also requires measurable indicator metrics that help diagnose maintenance needs.³⁹ Some considerations of long-term operation are shared among both nutrient-focused electrochemical technologies and other electrochemical techniques, such as electrolysis (e.g., in the chlor-alkali industry⁴⁰), electrodialysis (e.g., for potable water production from brackish water⁴¹), and electrodeionization (e.g., at nuclear power plants⁴²). However, some aspects are unique to nutrient recovery and applications to excreta-derived waste streams. Thus, a long-term study of ECS could inform the development and implementation of other electrochemical nutrient treatment processes due to the similarity of operating mechanisms and reactor construction across many technologies.

This study demonstrates and investigates electrochemical nutrient recovery in diverse wastewaters under long-term operation. Specifically, we (1) evaluated nutrient removal and recovery performance when treating multiple wastewaters; (2) investigated performance stability, failure mechanisms, and monitoring methods during long-term operation; and (3) demonstrated the robustness of ECS performance when exposed to influent urine variability. The demonstrated operation of ECS for over a month informs understanding of costs, maintenance, process control, and pretreatment strategies to prevent performance degradation. The large data set under variable influent conditions generated in this study can support validation of future process models. By providing evidence for robust operation under relevant process conditions, this work advances ECS toward technology readiness level 5 (i.e., a laboratory-scale system in a simulated operational environment) as defined by the National Aeronautics and Space Administration and the Department of Defense.^26,43,44 Our long-term demonstration of ECS treating about half of the volume of urine excreted per person each day⁴⁵ will support future investigations of larger scales and other waste streams. These advancements will accelerate integration into robust treatment trains that facilitate the achievement of several United Nations Sustainable Development Goals (SDGs): provision of safe drinking water and sanitation (SDG 6), responsible consumption and production (SDG 12), and protection of water resources against pollution (SDG 14).^13,46

2. Materials and Methods

2.1. Electrochemical Reactors

ECS experiments were performed in a three-chambered parallel-plate reactor (Figure S1-1) constructed of three frames (internal dimensions 8 cm × 8 cm × 1.9 cm) compressed between two square plates (dimensions 18 cm × 18 cm × 1.9 cm). Buna-N rubber gaskets (Grainger, Lake Forest, IL) were used to separate all of the reactor components and prevent leaks. Chambers 1 and 2 were separated by a cation exchange membrane (CEM; FumaTech Fumasep FKS-50, Fuel Cell Store, Bryan, TX). Chambers 2 and 3 were separated by a membrane facilitating ammonia transport. Chamber 1 contained a dimensionally stable titanium-based anode (64 cm² geometric area). Chamber 2 contained a 304L stainless steel mesh cathode (64 cm² geometric area, 40 mesh, 0.425 mm opening, 190 μm wire diameter, 42% open area, Yikai Industrial Scientific, Shanghai, China) with a 304 stainless steel current collector (KC Sheetmetal, San Jose, CA). The cathode was used as the working electrode, and the anode was used as the counter electrode and reference electrode to facilitate continuous monitoring of the full-cell potential. We operated the reactor galvanostatically (i.e., chronopotentiometrically, CP), controlling the rate of reaction at the working electrode by maintaining a constant current density (Interface 1010E potentiostat, Gamry Instruments, Warminster, PA). Because we did not need to precisely measure or control the interfacial potential across the working electrode–electrolyte interface, a two-electrode system without a separate reference electrode was sufficient.⁴⁷ Masterflex peristaltic pumps (VWR International, Radnor, PA) controlled the flow into and out of reactors.

The anode and the ammonia-transporting membrane between Chambers 2 and 3 varied between experiments. The anode used in Chamber 1 was either a titanium mesh electrode coated with IrO₂–Ta₂O₅ mixed-metal oxide (Magneto Special Anodes, Schiedam, The Netherlands) for batch ECS experiments (Section 2.2) or a titanium mesh electrode coated with IrO₂ mixed-metal oxide (Optimum Anode Technologies, Camarillo, CA) for continuous ECS experiments (Sections 2.4 and 2.6). The ammonia-transporting membrane material varied depending on experiment duration because membrane wetting occurred during preliminary long-term experiments. Specifically, a polyethylene hydrophobic membrane (Aquastill, Sittard, The Netherlands) was used for the batch experiments with different wastewaters (Section 2.2) and for the continuous experiments with varying influent urine composition (Section 2.6). For the long-term continuous experiments (Section 2.4), we used a wetting-resistant omniphobic acrylic copolymer membrane cast on a nonwoven nylon support (Versapor RC450, Pall Corporation, Port Washington, NY). For continuous experiments discussed in Sections 2.4 and 2.6, the ammonia-transporting membranes were framed with polyethylene tape (VWR International, Radnor, PA) to protect them from contact with rubber gaskets; framing with tape resulted in about 43% lower active membrane area for ammonia-transporting membranes compared with the CEM (Figure S1-2).

2.2. Feasibility of Treating Multiple Wastewaters

Triplicate batch ECS experiments (Figure S1-3A) were conducted for six different wastewaters, including fecal sludge treatment plant (FSTP) effluent, reverse osmosis concentrate from full advanced treatment of municipal wastewater, anaerobically treated mainstream municipal wastewater, and three anaerobic digestates from solids processing at municipal wastewater treatment plants. Data from previous studies were also reported for two additional wastewaters: urine and hydrothermal liquefaction effluent.^23,24 The composition of these wastewaters is summarized in Table S1-1 with process flow diagrams illustrating the sources of these wastewaters shown in Figure S1-4. Experimental conditions (i.e., applied current density and duration) (Table S1-2) were adjusted for wastewaters based on influent TAN concentration and wastewater volume treated. Reactor chambers were recirculated at a rate of 75 mL/min. ECS performance in each wastewater was evaluated based on removal and recovery efficiency, energy required for removal and recovery, and current efficiency (all metrics defined by eqs S1-10 through S1-14 in Section S1.7.1).

2.3. Urine Collection for Continuous ECS Experiments

Urine was collected from July 3, 2021 to February 9, 2022 at Stanford University’s Shriram Center with approximately equal volumetric contributions from adult male and female donors (Internal Review Board Protocol 60601). Urine was stored in a closed container at room temperature (25 °C) and allowed to hydrolyze for at least 12 months. Hydrolyzed urine contained less than 0.35 mM urea-N compared to 360 ± 14 mM TAN for long-term continuous experiments (Section 2.4) and 366 ± 2.4 mM TAN for continuous experiments with varying influent urine composition (Section 2.6), indicating complete hydrolysis. For experiments involving partially hydrolyzed urine, fresh urine was collected in closed 8-ounce containers by individuals during June and July 2023 and stored at 4 °C within 3 h of collection. To minimize urea hydrolysis while allowing aggregation of samples from multiple donors, individual containers were combined into glass bottles every 1 to 2 days and stored at 4 °C until about 1 h before use in experiments.

2.4. Continuous Long-Term Urine Treatment

Two ECS reactors operated simultaneously to treat separately collected urine (Figure S1-3B). For each reactor, influent urine (Table 1) was supplied to Chamber 1 from a 10 L container (same for both reactors) and refilled daily to minimize the headspace and potential influent ammonia volatilization. Treated effluent urine from Chamber 1 was collected in a 20 L container (separate for each reactor). For each reactor, Chamber 1 was connected to a recirculation bottle initially containing 250 mL of acidified urine (pH 2.8, adjusted with H₂SO₄) to limit foaming in the reactor during early operation, likely caused by urine surfactants and bicarbonate neutralization. Chamber 2 was connected to a recirculation bottle containing 250 mL of 0.1 M NaCl, and this solution was replaced if its volume was low when the CEM was replaced (Section 3.2.1). Chamber 3 was connected to a recirculation bottle containing 500 mL of 1 M H₂SO₄. Sulfuric acid was chosen as the absorbent because it facilitated near-complete TAN recovery in previous studies.^22,23,25 The H₂SO₄ concentration was chosen to allow several days of operation before the solution was neutralized by recovered TAN and required replacement to continue providing an acidic sink that facilitates TAN recovery. Each chamber was recirculated at 59 mL/min. Influent and effluent urine were pumped at approximately 0.5 mL/min, equivalent to a hydraulic residence time (HRT) of about 4 h (eq S1-1), to achieve at least 80% removal efficiency (Figure S1-5, Section S1.3). Samples were collected from influent urine, effluent urine, combined effluent from each reactor, and all three ECS chambers at 4 and 8 h (i.e., every HRT initially) and then daily starting at 24 h. The duplicate reactors were operated for 35 and 37 days. Performance metrics (removal and recovery efficiency, energy required for removal and recovery, product purity, energy consumption per recovered product batch, and removal and recovery flux) were calculated (eqs S1-15 through S1-24 in Section S1.7.2).

Table 1. Influent Urine Characteristics for Long-Term Continuous Electrochemical Stripping (ECS) Treatment^a.

Urine
TAN (mg/L)	5,040 ± 200
Cations (mg/L)
Na+	1,620 ± 81
K+	1,640 ± 110
Mg2+	1.19 ± 0.64
Ca2+	11.0 ± 10
Anions (mg/L)
SO42–	740 ± 120
Cl–	2,300 ± 380
PO43–	640 ± 110
Other
pH	9.25 ± 0.10
Electrical conductivity (mS/cm)	30.6 ± 8.7
COD (mg/L)	4,040 ± 250
Total inorganic carbon (TIC, mg/L)	1,990 ± 88
Total organic carbon (TOC, mg/L)	1,550 ± 150

^aNO₃^–, NO₂^–, Br^–, and F^– were not detected. Urine compositions reported in literature vary, but the composition reported here falls within reported ranges.^4,45,48 Resuspension of precipitated salts during refilling of the influent container could cause variable Ca²⁺ concentration.

2.5. Electrochemical Monitoring of Long-Term Urine Treatment

Two electrochemical monitoring measurements, electrochemical impedance spectroscopy (EIS) and fast chronopotentiometry (CP), were performed daily for the first 14 days of the long-term experiment and every 2–4 days for the remainder. When performing electrochemical monitoring, we paused ECS operation and recirculated the solutions for 5 min to disperse dissolved hydrogen and oxygen. We then stopped recirculation, completed both electrochemical measurements, and resumed recirculation and ECS operation. Gamry Interface 1010E potentiostats were used for electrochemical monitoring measurements.

Galvanostatic EIS was used to determine the contributions of membranes and solutions to the overall resistance. This measurement was conducted when ECS operation was paused and repeated with the same settings about 2 min after ECS operation resumed. Platinum wire electrodes (0.5-mm diameter, Thermo Scientific, Waltham, MA) were immersed into the middle of Chambers 1 and 2 (Figure S1-6A), with the electrode in Chamber 2 acting as the working electrode and the electrode in Chamber 1 acting as the counter and reference electrode. An alternating current with a frequency of 0.01 Hz to 0.1 MHz and an amplitude of 0.1 rms mA (root mean squared; i.e., 0.14 mA peak current) was applied, and six measurements per decade were taken. The x-intercept of Nyquist plots, representing the total membrane and solution resistance,^49,50 was determined for each measurement. If the data crossed the x-axis, the x-intercept was interpolated using a linear fit defined by the two points closest to and on opposite sides of the x-axis. If the data did not cross the x-axis, the x-intercept was extrapolated by using a linear fit to the two points closest to and above the x-axis. To isolate the contribution of membrane resistance to total resistance, we subtracted solution resistance (calculated from Chamber 1 and 2 solution conductivities) and resistance of electrical patch cords from the total measured resistance (eqs S1-2 through S1-6, Section S1.4).

Fast CP was conducted to study changes in the mechanisms of ion transport that can be associated with membrane fouling. Current densities of 10 and 2 mA/cm² (i.e., operational ECS current density and a current density similar to that applied during CP in other ion exchange membrane investigations^51,52) were applied for 60 s, and full-cell voltage was measured every 0.01 s. We used the stainless steel cathode as the working electrode and the anode as the counter and reference electrodes (Figure S1-6B). The observed full-cell voltage was plotted over time, and two values were extracted from the curves: the initial voltage drop (full-cell voltage at 0.01 s) and the transition time (first inflection point).

2.6. Impacts of Urine Collection Factors on Performance

To investigate the impact of influent urine composition on ECS performance, we investigated three factors that vary with urine collection and storage procedures: (1) dilution of urine with flush water, (2) extent of divalent cation precipitation, and (3) degree of urea hydrolysis. Solutions in each chamber, recirculation flow rate, and influent and effluent flow rates were identical to those used in continuous long-term experiments (Section 2.4). Influent urine was supplied from a 2 or 10 L container, and treated effluent urine was collected in a 2 or 10 L container. Duplicate or triplicate 48 h continuous ECS experiments (Figure S1-3B) were conducted for each influent (Tables S1-3 through S1-5). Samples were collected from each chamber and the effluent at 0, 16, 20, 24, 36, 40, 44, and 48 h to capture multiple data points during transient and steady-state periods. Influent samples were collected at 0, 24, and 48 h to verify that influent composition remained consistent throughout each experiment. Performance metrics (removal and recovery efficiency, energy required for removal and recovery, current efficiency, and removal and recovery flux) were calculated (eqs S1-15, S1-17 through S1-19, and S1-22 through S1-24 in Section S1.7.2).

2.7. Membrane Characterization

After removal from the reactor, CEMs and omniphobic membranes were stored and dried in closed Petri dishes for 2 to 4 months before characterization. Scanning electron microscopy (SEM, Thermo Fisher Scientific Apreo S LoVac microscope) and energy-dispersive X-ray spectroscopy (EDS, Bruker Quantax XFlash 6|60 EDS detector) were used to understand the morphology and elemental distribution of the foulants on the membrane surface.^34,53−63 Membrane samples were mounted on aluminum pin stubs and sputter-coated with gold to improve the imaging surface conductivity. The SEM images and EDS elemental maps (C, O, Na, K, Ca, Mg, P, and S) were acquired at 12 kV and 50 pA. EDS data were collected at two points on each membrane sample. Functional groups of the foulants were analyzed using Fourier transform infrared spectroscopy (FTIR, Thermo Fisher Scientific Nicolet iN10) with an attenuated total reflection (ATR) accessory.^{34,55,57−59,62,64} Spectra were obtained by averaging 128 scans with a wavenumber resolution of 4.0 cm^–1 in the range of 680–2000 cm^–1. A high-sensitivity liquid nitrogen-cooled detector was used to distinguish small features due to a high background signal from the virgin membrane. ATR-FTIR data were collected at two or more points on each membrane, but we only showed multiple spectra if different features were detected at different points (Section 3.2.4).

Potentiostatic EIS was used to determine the ionic conductivity of the CEM.⁶⁵⁻⁶⁷ For the virgin CEM and each used CEM, four samples were cut slightly larger than the circular electrode (2 cm diameter) in the sample holder and equilibrated with 0.1 M NaCl solution overnight prior to measurement. Stacks consisting of 1, 2, and 3 samples were placed into a sample holder (BioLogic CESH-e, Seyssinet-Pariset, France) and compressed to allow through-plane measurement. Using a VMP potentiostat and EC-Lab software (BioLogic, Seyssinet-Pariset, France), an oscillating potential of 10 mV was applied with a frequency of 100 Hz to 3 MHz, and ten measurements per decade were taken for two replicate CEM stacks. Open- and short-circuit compensation tests were conducted to characterize the resistance contributions from the CESH-e device and wiring, and EIS spectra were corrected with the compensations. The thickness of each membrane stack was measured at five positions by using a digital micrometer (Mitutoyo America Corporation IP65, Aurora, IL). The x-intercept of the Nyquist plot was taken as the total ohmic resistance of the membrane, membrane–electrode interfaces, and electrolyte–surface interfaces. The measured area resistance, determined as the product of the electrode area and the measured total resistance, was plotted against the stack thickness, and ionic conductivity was calculated from the slope of a linear fit (eqs S1-7 through S1-9, Section S1.6).

2.8. Chemical Analyses

Sample pH was measured with a pH meter (FiveEasy F20 or SevenDirect SD20, Mettler Toledo, Columbus, OH) and sample conductivity was measured with a pH/conductivity meter (SevenDirect SD23, Mettler Toledo, Columbus, OH). TAN concentrations were determined by the indophenol method using flow-injection analysis (AA500 AutoAnalyzer, SEAL Analytical, Mequon, WI). A Dionex ICS-6000 was used to determine cation concentrations (IonPac SCS1 column, unsuppressed, 4 mM tartaric acid/2 mM oxalic acid eluent, 1.0 mL/min, 30 °C, Thermo Scientific, Waltham, MA) and anion concentrations (IonPac AS22-4 μm column, suppressed, 4.5 mM Na₂CO₃/1.4 mM NaHCO₃ eluent, 1.0 mL/min, 30 °C, Thermo Scientific, Waltham, MA). Chemical oxygen demand (COD) was determined colorimetrically by Hach Method 8000 using high-range (0–1,500 mg/L) reagent kits (CHEMetrics, Midland, VA) with a DRB200 digital reactor block and a DR1900 portable spectrophotometer (Hach Company, Loveland, Colorado). Total organic and inorganic carbon (TOC and TIC) were measured using a Shimadzu TOC-L autoanalyzer (Shimadzu Scientific Instruments, Inc., Columbia, MD).

2.9. Preliminary Cost Assessment

Capital costs for the ECS reactor (Table S1-6) were determined based on the reactor components for the long-term continuous experiment (Section 2.1). Operating costs (Table S1-7) accounted for electricity consumption by ECS and associated pumps, chemical inputs for periodic replacement of Chamber 2 and 3 solutions, and periodic replacement of key components (anode, cathode, CEM, and ammonia-transporting membrane). The cost of recovering ammonia or net present value (NPV) was calculated using eq 1:

1

where R_n is the cash flow (i.e., costs) during time period n, R_PV is the present value of a cash flow, r is the discount rate (10%) used to convert cash flows during time period n to present value, i is the interest rate (3%) used to determine cash flows during time period n based on the present value of cash flows, and t is the time period for the cost assessment (20 years). Operating costs of ECS were compared to those of other TAN removal and recovery processes.

3. Results and Discussion

3.1. Feasibility of Treating Multiple Wastewaters

Batch experiments demonstrated the applicability of ECS to multiple wastewaters with varying characteristics (Table S1-1). ECS achieved greater than 90% TAN removal in all wastewaters tested in this and previous studies, including urine, hydrothermal liquefaction effluent, FSTP effluent, reverse osmosis concentrate, anaerobically treated mainstream wastewater, and three anaerobic digestates from solid processing at municipal wastewater treatment plants. ECS also recovered greater than 75% of TAN in all wastewaters tested and greater than 80% of TAN in all wastewaters, except reverse osmosis concentrate (Figure 1). Reverse osmosis concentrate exhibited three to 30 times higher divalent cation (Mg²⁺, Ca²⁺) concentrations than other wastewaters (Table S1-1). Under basic conditions in Chamber 2, these ions precipitate (e.g., as phosphate and carbonate salts^68,69) and could deposit on membrane surfaces, interfering with TAN recovery.

Figure 1.

Total ammonia nitrogen (TAN) removal and recovery efficiency in multiple wastewaters with TAN concentrations (shown on the right axis) varying across 2 orders of magnitude. Wastewaters are arranged in order of increasing TAN concentrations from left to right. SAF-MBR effluent is a staged anaerobic fluidized membrane bioreactor effluent; ROC is reverse osmosis concentrate; FSTP effluent is fecal sludge treatment plant effluent; AD is anaerobic digestate; and HTL effluent is hydrothermal liquefaction effluent. All wastewaters were generated during the treatment of municipal wastewater with process flow diagrams shown in Figure S1-4. Dashed horizontal lines indicate the lowest removal (blue) and recovery (green) efficiencies observed across the wastewaters. Data for urine and HTL effluent are taken from previous ECS studies;^23,24 experiments for all other wastewaters were performed in the current study. Statistical analysis of differences between removal and recovery metrics is shown in Table S2-1.

Energy consumption for TAN removal and recovery ranged from 49.2 to 606 MJ/kg N for most wastewaters investigated (Figure S2-1), except for FSTP effluent, which required about 2,900 MJ/kg N. The high energy consumption for FSTP effluent can be attributed to a difference in reactor configuration and selection of current density (Table S1-2). For FSTP effluent, electrodes were placed at opposite ends of reactor chambers instead of immediately adjacent to the CEM (the case for all other wastewaters tested in this study), which caused solution resistance to contribute more significantly to the full-cell potential and energy demand. Moreover, a relatively high current density of 100 A/m² was used as in a previous study on urine treatment²³ despite the low conductivity of FSTP effluent (4.5 vs 24.7 mS/cm for urine). The resulting greater ratio of current density to wastewater conductivity (Figure S2-2) compared to other wastewaters contributed to the greater energy input required for FSTP effluent. These results illustrate the importance of tuning the reactor configuration and operating conditions to influent composition.

We demonstrated a proof-of-concept for applying ECS to various wastewaters with operating conditions selected based only on influent wastewater volume and TAN concentration. Further optimization of the operating parameters could reduce energy consumption. Current density normalized to TAN mass introduced in the influent, removed from the influent, or recovered to Chamber 3 (Figure S2-3A) reflected trends in energy consumption across wastewaters, as did the charge passed normalized to TAN mass removed and the current efficiency (Figure S2-3B). Researchers previously identified the load ratio (i.e., current density normalized to TAN loading rate) as a predictive parameter for removal efficiency and energy consumption for continuous electrodialysis–membrane stripping;⁷⁰ thus, load ratio could also help optimize batch operating conditions for energy efficiency based on influent composition.

3.2. Robustness during Continuous Urine Treatment

3.2.1. Nutrient Removal and Recovery Performance during Long-Term Operation

We focused long-term evaluation on urine treatment because urine is a high-conductivity, high-TAN waste stream amenable to energy-efficient electrochemical treatment^7,23,71 and because source separation of urine could provide many wastewater treatment benefits.⁷²⁻⁷⁶ When treating influent urine containing 5,050 ± 210 mg/L TAN for over 30 days, ECS achieved consistent steady-state removal and recovery efficiencies of 83.4 ± 1.5 and 73.0 ± 2.9%, respectively (Figure 2A), after an initial transient start-up phase (Figures S2-4 through S2-7). Most electrochemical nutrient recovery technologies, except bioelectrochemical systems, have been operated for much shorter times on the order of days.^27,29

Figure 2.

(A) Long-term performance metrics for electrochemical stripping (ECS). Daily influent TAN concentration for both reactors and average cumulative removal and recovery efficiency achieved by two reactors are shown along with the product TAN concentration achieved by one reactor (Reactor A). The product TAN concentration profile for Reactor B is shown in Figure S2-8B. Final product batch pH values are noted at each peak in the product TAN concentration profile, which corresponds to the time of trap solution replacement. (B) Purity of recovered product batches and energy required for generating each batch for Reactor A. Analogous data for Reactor B are shown in Figure S2-10. The vertical green line indicates replacement of the omniphobic membrane.

Both ECS reactors achieved reproducible TAN recovery, with each reactor generating eight batches of ammonium sulfate solution (product TAN concentrations in Figures 2A and S2-8). Product TAN concentration rose steadily until the initially supplied 1 M H₂SO₄ was neutralized (pH values in Figures 2A and S2-8), reaching a final pH of 8.59 ± 0.47 across both reactors. The products exhibited consistent TAN concentrations of 1.49 ± 0.15 M (20.9 ± 2.0 g/L) TAN across the eight batches recovered by each reactor (i.e., 16 batches total), which was slightly lower than commercial ammonium sulfate fertilizers (28.4 g/L TAN)⁷⁷ but similar to concentrations recovered from synthetic anaerobic digestate by an electrodialysis–membrane stripping system.⁷⁷ Therefore, ECS repeatably achieved a concentration factor of 4.1 ± 0.4 for the neutral product compared to the average influent concentration. In total, Reactors A and B recovered 107 and 103 g TAN from 27.8 and 26.4 L of urine over the entire operating period, respectively (Figure S2-9).

Implementation of ECS requires the recovery of a high-quality product. When considering anionic and cationic impurities measured by ion chromatography (IC) (specifically Na⁺, K⁺, Ca²⁺, Mg²⁺, Cl^–, PO₄^3–, NO₃^–), multiple consecutive batches of recovered ammonium sulfate met ACS reagent-grade purity guidelines of 95% (w/w) (Figures 2B and S2-10).⁷⁸ In decreasing concentration (mg/L) order, K⁺, Na⁺, and Cl^– were detected in the recovered product. K⁺ is an important mineral for plant growth⁷⁹ and, therefore, likely not a concerning contaminant for the application of recovered ammonium sulfate as a fertilizer, but Na⁺ and Cl^– can contribute to increasing soil salinity, which can reduce agricultural productivity.⁸⁰ Despite some K⁺ and Na⁺ transfer to the product chamber, ECS was highly selective for TAN recovery; K⁺ and Na⁺ were removed about 1.2 times as efficiently as TAN (99% K⁺ and Na⁺ removal vs 83.4% TAN removal), but TAN recovery efficiency was about 91 and 66 times the recovery efficiency of Na⁺ and K⁺, respectively (73.0% TAN recovery vs 0.80% Na⁺ recovery and 1.1% K⁺ recovery). After purity fell below reagent grade, replacement of the omniphobic membrane restored product purity for Reactor A. Reactor B did not exhibit the same purity effect likely because of an undersized replacement membrane (Figure S2-11).

To investigate contaminants beyond the inorganic ions measured by IC, we quantified TOC and TIC (Figures S2-12 and S2-13). TOC and TIC were present in the products starting with the first batch, but concentrations remained approximately constant for the first four batches (20 days). For batches five through eight, the TOC concentration increased significantly, while TIC remained similar or slightly greater than for the first four batches. The increase in TOC aligned with the increased presence of ionic contaminants and the associated drop in purity (Figure 2B). We observed a more dramatic increase in TOC for the final batch recovered in Reactor B, which provides further evidence of an undersized replacement membrane. Organic contamination could originate from degrading polymeric membranes,^81,82 transmembrane migration of urine constituents (e.g., amino acids, pharmaceuticals), and urine contamination due to compromised CEMs.^4,83−85 Inorganic carbon contamination could originate from bicarbonate in influent urine and equilibration with atmospheric CO₂, which is highly soluble in basic solutions.⁸⁶⁻⁸⁸ Although ECS with hydrophobic membranes showed negligible pharmaceutical migration to Chamber 3,²³ omniphobic membranes treating larger volumes of urine could have enhanced organic migration. More thorough characterization of carbon-containing constituents could prove valuable for addressing potential user concerns about product purity⁸⁹ and for informing mitigation strategies.

We evaluated electrical energy consumption for ECS throughout the long-term experiment because previous studies indicated that it was the predominant factor in overall energy demand.²³ Energy consumption for recovery was 70.1 ± 9.4 MJ/kg of N when averaged across all batches generated by both reactors (Figures 2B and S2-10). Daily evaluation of energy consumption for removal and recovery showed consistent values throughout the experiment (Figure S2-5A). From 48 h onward, average steady-state energy consumption was 61.2 ± 4.2 MJ/kg N removed and 69.6 ± 3.2 MJ/kg N recovered. These values were about twice the energy consumption (30.6 MJ/kg N removed) reported in a previous ECS study, in which only 60.6% of influent TAN was removed and 49.6% was recovered while continuously treating urine.²³ Thus, in this study, we inputted more energy (i.e., applied greater current density relative to flow rate) to achieve more complete treatment (83.4% removal and 73.0% recovery). We also compared removal/recovery efficiency and energy consumption observed in this study to values reported for other TAN removal and recovery technologies applied to urine or synthetic urine (Figure S2-14, Table S2-2). ECS achieved greater removal/recovery efficiencies than many other processes but also required more energy, further highlighting that more complete treatment (i.e., greater removal/recovery efficiency) often requires more energy. Further optimization and reactor engineering of ECS could improve ECS performance relative to other emerging urine treatment and nutrient recovery technologies and reduce ECS energy consumption toward that of Haber-Bosch ammonia synthesis (about 31.6 MJ/kg N)⁹⁰ or conventional ammonia stripping (32 MJ/kg N).⁹¹

3.2.2. Operation and Maintenance Tasks

O&M tasks for membrane-based and electrochemical processes are often tied to membrane fouling, which hinders stable long-term operation.^34,58 In addition to absorbent solution renewal and omniphobic membrane replacement (Section 3.2.1), lab-scale ECS operation showed that replacing failed CEMs was a major O&M task. CEMs were replaced every 5.2 ± 1.9 days (reactor A) and 8.4 ± 2.5 days (reactor B) when the effluent pH was 25% higher than the steady-state average (Figure S2-15). Although we could have continued operation longer per membrane (no decline in cumulative removal efficiency observed), we chose to replace the CEMs at the first sign of failure to ensure that removal and recovery efficiency remained stable and high. Frequently replacing the CEM also reduced contamination of Chamber 2 with urine constituents (i.e., components other than cations that we expect to transfer across the CEM), minimizing the potential for omniphobic membrane degradation and product contamination. Future work could focus on determining the upper limit of membrane lifetime instead of the upper limit of process operation explored here.

Replacing CEMs and omniphobic membranes required pausing operation, emptying the ECS reactors, and disassembling the reactors. The complexity and disruptiveness of this procedure demonstrated the need to design scaled-up reactors that are easily serviceable, for example, by facilitating membrane replacement without significant disassembly (e.g., sliding a framed membrane into the reactor), and highlights the importance of operating redundant ECS units to avoid treatment interruptions during maintenance.

3.2.3. Potential Process Monitoring Parameters

Ideal ECS monitoring parameters are (1) fast, requiring minimal interruption of treatment or allowing online measurement, and (2) inexpensive, requiring few additional parts or equipment beyond components in the ECS reactor and common at treatment plant laboratories. The following sections summarize the parameters studied for monitoring the overall process function, CEM integrity, and omniphobic membrane integrity. pH, full-cell voltage, product (Chamber 3) volume, and water flux across the omniphobic membrane were identified as promising process monitoring parameters, while in situ EIS, fast CP, and product conductivity require method modifications and further study to evaluate their utility as monitoring tools.

3.2.3.1. Overall: pH

Because ECS relies on pH swings to achieve TAN removal and recovery, pH is a critical monitoring parameter. TAN removal requires a low pH in Chamber 1 to facilitate the conversion of neutral NH₃ in the influent to cationic NH₄⁺, which can electromigrate across the CEM. TAN recovery requires a high pH in Chamber 2 to facilitate the conversion of NH₄⁺ to NH₃, which can cross the omniphobic membrane. TAN recovery also requires a low pH in Chamber 3 to facilitate the conversion of NH₃ back to NH₄⁺, creating an NH₃ sink and, therefore, a persistent driving force for NH₃ diffusion from Chamber 2 to 3.

pH trends and associated reactor O&M tasks are shown in Figures 3 and S2-16. During normal operation, the effluent pH remained stable at about pH 2. When effluent pH increased toward the neutral range despite no significant changes in influent pH, we replaced the CEM, which restored the effluent pH to typical values. Therefore, effluent pH is a valuable indicator for CEM failure because it demonstrates when significant leakage of hydroxide or NH₃ occurs from Chamber 2 to Chamber 1. Hydroxide ions (OH^–) are highly mobile and can cross CEMs, particularly if membrane degradation decreases permselectivity.⁹² NH₃, a small neutral molecule, can diffuse through CEMs.⁸³⁻⁸⁵ Increasing effluent pH was often correlated with increasing effluent TAN concentration and decreasing instantaneous removal efficiency (Figure S2-17). Therefore, online pH monitoring could be a valuable indicator for deterioration of removal performance, allowing quicker responses to diminished removal than offline TAN analysis.

Figure 3.

pH values in influent, effluent, and Chambers 2 and 3 over time with replacement of the cation exchange membrane (CEM, vertical blue lines) and replacement of exhausted Chamber 3 solution (vertical green lines) indicated for Reactor A. Data for Reactor B are shown in Figure S2-16.

Chamber 2 pH generally remained between pH 13 and 14, suggesting that damage to the CEM or omniphobic membrane was not sufficient to allow acid leakage into Chamber 2. However, catastrophic failure of the omniphobic membrane could likely be detected by a pH decrease in Chamber 2. Chamber 3 pH increased periodically every 4 to 5 days when TAN absorption neutralized the sulfuric acid in Chamber 3; thus, pH is an essential indicator for replacing Chamber 3 solution to continuously recover TAN. pH probes, which are readily available, could be integrated into ECS reactors, enabling real-time monitoring that signals the need for multiple O&M tasks.

3.2.3.2. Cation Exchange Membrane: Operational Full-Cell Voltage, EIS, and Fast CP

The full-cell voltage for ECS can reflect changes in the CEM, electrodes, and solution composition. We observed a rapid increase in full-cell voltage magnitude upon resuming operation after performing process monitoring measurements or routine maintenance tasks that required pausing the experiment (Figure S2-18). This pattern reflects the ohmic potential drop associated with the membrane and surrounding electrolyte, the rapid emergence of strong concentration polarization when an overlimiting current was applied, and the Nernstian shift for the pH-affected oxygen evolution and hydrogen evolution reactions.^51,52,93,94 In general, the magnitude of full-cell voltage increased over the course of each CEM’s use, suggesting that fouling contributed to increased membrane resistance. The degree of change varied across the reactors and membranes, but these differences could not be explained by CEM characterization (Section 3.2.4).

In situ EIS showed no consistent trend in the measured membrane or total resistance over the time for which each CEM was used (Figures S2-19 through S2-22, Table S2-3). Nevertheless, the similarity of results of non-operational and operational in situ EIS measurements suggests the feasibility of online EIS monitoring without interruption of ECS treatment. We determined the ionic conductivity of each membrane after removal from the reactor using a higher fidelity, more controlled EIS measurement method compared with our in situ method. The ionic conductivity of the used membranes was significantly lower than that of the virgin membrane, and ionic conductivity generally decreased as the duration of installation increased (Figures S2-23 through S2-25). These results suggest that fouling hindered ion movement through the CEM. We expect that in situ EIS monitoring can be more informative if the setup more closely mimics the setup used for offline ionic conductivity measurement (e.g., planar electrodes in direct contact with the membrane^95,96).

From fast CP, we can extract two informative parameters: an initial voltage drop (ohmic potential drop caused by membrane and electrolytes without concentration polarization)^51,52,93,94 and transition times (indicating a change in ion transport mechanisms).^93,97,98 For each membrane used in each reactor, we observed a weak increasing trend in initial voltage drop over time for fast CP at 2 mA/cm² but no temporal trend for fast CP at 10 mA/cm² (Figure S2-26). These results suggest that during the time scales for which CEMs were used in this study (5–8 days), membrane degradation and fouling were insufficient to significantly affect initial potential drops during fast CP.⁵² For most of the fast CP data collected during our experiments, no clear transition times could be identified (Figures S2-27 through S2-31), which could be related to the heterogeneity of the CEM, the current densities chosen, or the electrode configuration (Section S2.2.2). While we used a two-electrode configuration for simplicity and cost reduction, conducting CP at lower but still overlimiting current densities using a four-electrode configuration⁹³ could improve the identification of transition times and support the development of more informative CP-based monitoring techniques.

3.2.3.3. Omniphobic Membrane: Water Flux, TAN Flux, and Product Conductivity

Intact omniphobic membranes facilitate diffusion of NH₃ and potentially other small, uncharged, gaseous species (e.g., water vapor) from Chamber 2 to 3. However, wetted omniphobic membranes allow liquid water transfer and diffusion of dissolved charged species between Chambers 2 and 3, decreasing product purity.^99,100 During long-term operation, water flux for Reactor B increased steadily, suggesting increasing omniphobic membrane wetting (Figure S2-32); water flux for Reactor A was generally lower, explaining this reactor’s greater product purity (Figures 2B and S2-10). Therefore, product (Chamber 3) volume and water flux could be informative monitoring parameters, but further investigation of the relationship between these metrics and product quality is needed to identify threshold values indicating omniphobic membrane issues.

The decreasing flux of desired species over time indicates fouling of omniphobic and hydrophobic membranes.^55,101,102 TAN recovery flux across the omniphobic membrane remained stable at 27.5 ± 3.8 g/m²/h for Reactor A and 28.1 ± 0.86 g/m²/h for Reactor B at 48 h onward (Figure S2-33, Table S2-4). The TAN recovery rate also remained stable throughout the entire experiment at 0.117 ± 0.0041 g/h (Reactor A) and 0.118 ± 0.0035 g/h (Reactor B), with little effect from omniphobic membrane replacement. Consistent TAN recovery flux and rate suggested that omniphobic membrane fouling did not interfere with TAN transport from Chamber 2 to Chamber 3, but increasing membrane wetting could have compensated for fouling-induced hindered transport.

Distillate conductivity has been proposed as an indicator of wetting for hydrophobic membranes in membrane distillation.^54,99,103 We observed a reproducible decrease in Chamber 3 conductivity for both reactors from fresh H₂SO₄ solution to final recovered batches of (NH₄)₂SO₄ solution (Figure S2-34), which was mainly driven by TAN chemistry: when NH₃ is absorbed in Chamber 3, a proton is consumed to generate NH₄⁺, which has an ionic conductivity almost 5 times lower than that of protons.¹⁰⁴ Water transfer across the omniphobic membrane also diluted the product and contributed to decreasing conductivity; however, other charged species transfer (e.g., K⁺, Na⁺) would have increased conductivity. Based on these results, conductivity is not sensitive enough to be used as an indicator of product purity decreasing below a 95% threshold (i.e., below ACS reagent grade). However, the conductivities of simulated products with varying purities (Figure S2-35, Table S2-5) illustrated a near-linear relationship between conductivity and purity when neglecting water transfer across the omniphobic membrane. Further study is needed to understand the competing impacts of water and ion transfer on conductivity and to evaluate the effectiveness of conductivity as an indicator of more dramatic changes in purity.

3.2.4. Cation Exchange and Omniphobic Membrane Failure and Characterization

Because CEM failure (based on Chamber 1 effluent pH) occurred multiple times throughout our long-term experiments, CEM lifetime and failure mechanisms warrant further discussion. The duration for which CEMs are installed in reactors is a common way of reporting lifetime throughout literature despite variations in operating conditions (e.g., current density, influent flow rate).^52,105−108 Therefore, we described the membrane lifetime with duration of installation and several other metrics related to operation time and treatment achieved (Table S2-6). Because product purity decreased throughout the long-term experiment, we characterized the omniphobic membranes used for 33 and 31 days in Reactors A and B, respectively, to identify fouling that may have contributed to diminishing membrane performance.

SEM-EDS and ATR-FTIR indicated both organic and inorganic fouling on CEMs after use in ECS (Figures S2-36 through S2-38, Table S2-7 in Section S2.2.4), while only inorganic scaling was detected for omniphobic membranes (Figures S2-39 and S2-40, Table S2-8 in Section S2.2.4). Carboxylic acids and amides were detected on many CEMs, demonstrating the presence of organic foulants, possibly including amino acids, organic acids, creatinine, caffeine, pharmaceuticals, and various carbohydrates present in urine.^4,85 Scaling mainly consisted of calcium and magnesium carbonates and phosphates for both CEMs and omniphobic membranes. Variations in composition within and across membranes (especially CEMs) indicated heterogeneous membrane fouling and reflected differences in the composition of the Chamber 2 solution (Figure S2-41). These results demonstrate the need to mitigate organic and inorganic fouling to maintain membrane functionality and ECS performance stability. Potential mitigation strategies include using alternative membrane materials,^109,110 modifying reactor architectures to protect membranes from harsh conditions,^16,111−116 implementing tailored cleaning protocols,^{58,103,108,117−123} and incorporating urine pre-treatment.

3.3. Impacts of Urine Collection Factors on ECS Performance

Understanding the impacts of upstream urine collection and treatment practices on ECS operation can inform the infrastructure required for implementing ECS. We considered three influent urine composition factors in 48 h continuous experiments: dilution with flush water, precipitation of divalent cations, and extent of hydrolysis. Flush water dilution relates to urine collection methods; divalent precipitation and hydrolysis relate to storage time and conditions. Despite significant differences in the composition of influent urine based on these three factors, ECS achieved more than 69% removal and 65% recovery for all conditions (Figures 4 and S2-42). Further interrogating the relationship between ECS performance and influent conditions can inform the design and operation of complete urine collection and treatment systems.

Figure 4.

Total ammonia nitrogen (TAN) removal and recovery efficiency and energy consumption for electrochemical stripping (ECS) treating urine with different degrees of dilution, divalent precipitation, and hydrolysis. Black vertical lines separate the conditions tested for each of the three factors. A version of this plot with an expanded right y-axis is shown in Figure S2-42. The results of statistical analysis are shown in Tables S2-9 and S2-10.

We studied three dilution factors: no dilution (i.e., unmodified urine); dilution in the case of an ultralow-flush, urine-separating toilet (i.e., 2 times dilution); and dilution in the case of a standard urinal (i.e., 20 times dilution). A positive correlation between removal/recovery efficiencies and urine dilution was observed, but not all differences between conditions were statistically significant (Figures 4, S2-42, Tables S2-9 and S2-10). At greater dilutions, more complete treatment was achieved because HRT and current density remained constant at lower TAN concentrations. However, the additional treatment entailed greater energy consumption due to the lower conductivity of diluted influent urine, which caused a greater full-cell voltage. The standard-flush influent condition (1,950 ± 290 MJ/kg N recovered) resulted in about 30 times greater energy consumption than for unmodified urine (64.7 ± 1.1 MJ/kg N), and the ultralow-flush condition (126 ± 2.7 MJ/kg N) required almost twice as much energy as unmodified urine. The drastic increase in energy consumption and declining current efficiency (Figure S2-43) with increasing urine dilution highlight the importance of limiting flush water for energy-efficient ECS and tailoring operating conditions to influent composition and treatment goals.

Urine hydrolysis reduces the concentration of divalent cations (Mg²⁺ and Ca²⁺) due to the precipitation of struvite, calcium phosphate (e.g., hydroxyapatite), and calcium carbonate at increasing pH values.^68,69 To reflect the varied precipitation of divalent cations, we conducted experiments with complete precipitation (i.e., unmodified urine), 50% precipitation (i.e., hydrolyzed urine amended with 2.35 mM Ca²⁺ and 2.05 mM Mg²⁺),⁴ and no precipitation (i.e., hydrolyzed urine amended with 4.7 mM Ca²⁺ and 4.1 mM Mg²⁺)⁴ relative to fresh urine concentrations. We observed similar removal efficiencies (79–81%), removal fluxes (6.5–7.7 mg/m²/s), and energy consumption for removal (54–61 MJ/kg N) for all conditions (Figures 4, S2-42, and S2-44, Tables S2-9 and S2-10). The lack of significant differences between conditions and the observed maximum current efficiency at 50% amended influent (Figure S2-43) suggested a trade-off between phenomena impacting TAN removal. Divalent cations compete with NH₄⁺ transport across the CEM,¹²⁴ but higher divalent cation concentrations in the influent and Chamber 1 can cause additional osmotic water transfer that increases NH₄⁺ transfer,¹²⁵ resulting in a net neutral effect on TAN removal. For energy consumption, CEM resistance, voltage, and energy demand likely increased with increasing scaling at greater divalent cation concentrations (Figure S2-45); however, solution conductivity also increased with higher divalent cation concentrations, reducing the voltage and energy demand. Similar removal performance enabled similar recovery performance across different divalent cation concentrations. While we did not observe statistically significant impacts, divalent cations might have greater effects during longer operation due to accumulating CEM scaling, especially based on the increase in full-cell voltage observed over 48 h (Figures S2-45 and S2-46). These results motivate storing urine until complete hydrolysis and divalent cation precipitation before ECS treatment for optimal TAN recovery and further studying the long-term impacts of divalent cations.

Urine hydrolysis causes several additional compositional changes that increase conductivity and alkalinity, including rising pH and converting urea to TAN and bicarbonate.^69,126 Thus, we studied conditions reflecting different extents of hydrolysis, including completely hydrolyzed (i.e., unmodified urine), 50% hydrolyzed (i.e., 50% (v/v) fresh urine plus 50% (v/v) hydrolyzed urine), and 25% hydrolyzed (i.e., 75% (v/v) fresh urine plus 25% (v/v) hydrolyzed urine). Based on initially measured urea and TAN concentrations in partially hydrolyzed urine (Figure S2-47), the extents of hydrolysis (i.e., conversion of urea to ammonia) were 96.4 ± 23% for 50% hydrolyzed and 87.0 ± 5.7% for 25% hydrolyzed. Removal efficiencies decreased with decreasing extent of hydrolysis, and the 25% hydrolyzed case was statistically significantly different from the other two conditions. However, recovery efficiencies were similar for different extents of hydrolysis. Lower influent TAN concentrations at lower extents of hydrolysis caused greater competition between NH₄⁺ and other ions (e.g., Na⁺, K⁺) for transfer across the CEM.¹²⁴ Energy for removal and recovery increased as the extent of hydrolysis decreased, but differences across the conditions were not statistically significant. Lower conductivity and lower influent TAN concentrations in less hydrolyzed urine likely caused protons to be more dominant charge carriers, decreasing TAN current efficiency (Figure S2-43) and increasing energy demand. These findings further motivate the storage of urine until complete hydrolysis before ECS treatment to reduce energy consumption and maximize TAN recovery.

Treating partially hydrolyzed urine in ECS enables investigation of the fate of urea in the process. We focus the discussion on the 25% hydrolyzed case because the influent composition remained stable for 48 h, unlike the 50% hydrolyzed case (Figure S2-47). Most urea exited ECS in the effluent or persisted in Chamber 1 (49.5 and 10.0%, respectively), while about 7.1% transferred to Chamber 2; urea was not detected in Chamber 3 (Figure S2-48). About 33.4% of urea mass was lost (Figures S2-48 and S2-49), suggesting urea transformation within ECS. Extreme ECS pH conditions (1.50 ± 0.19 in Chamber 1 and 13.3 ± 0.18 in Chamber 2) make enzymatic urea hydrolysis unlikely because urease is most stable and active from pH 5.5 to 9.^126−128 Chemical urea transformation could occur electrochemically in Chamber 1 and due to high pH in Chamber 2.¹²⁹ Depending on electrode materials, pH, and electrical potentials, electrochemical urea conversion can yield many different nitrogen products, such as N₂, CNO^–, NO₃^–, NO₂^–, N₂O₂^–, N₂O, NO₂, and TAN.^130−133 We observed a TAN recovery efficiency exceeding the TAN removal efficiency (p = 0.0008) (Figure 4) and a steady-state average TAN mass balance of 119%, which suggests that urea was transformed to TAN (Figure S2-49A). Accounting for both TAN and urea as nitrogen species, the steady-state total mass balance was about 87.8%, which supports that urea conversion to ammonia within ECS could account for the excess recovered TAN (Figures S2-49 and S2-50). For potential urea oxidation products, NO₂^– was below detection, while NO₃^– was detected and improved the closure of the steady-state nitrogen mass balance to 89.6% (including TAN, urea, and NO₃^–). Notably, NO₃^– was also observed in the treated effluent for conditions representing different flush volumes and extents of divalent precipitation, suggesting that TAN oxidation could also be a source of effluent NO₃^– (Figure S2-51). Future experiments with gas chromatography (N₂, N₂O) and spectrophotometric methods (chloramines) could improve nitrogen mass balance closure.²³

3.4. Preliminary Cost Assessment

To connect lab-scale investigations to practical research gaps for electrochemical processes,^15,26 we completed a preliminary cost evaluation based on the lab-scale ECS reactor and performance during long-term urine treatment. At this scale (3.1 L and 170 cm² for reactor without ancillary equipment) and under these operating conditions, ECS treated about half of the volume of urine excreted per person each day.⁴⁵ The process could be scaled up to serve a household (5–10× higher flow rate) and even larger buildings (100–1,000× higher flow rate). At these different scales, costs—especially capital costs—and their relative contributions could vary due to different economies of scale associated with reactor components and O&M inputs. Nevertheless, a cost estimate at the lab scale rooted in experimental results can inform scale-up investigations and design decisions.

Over a 20-year analysis time frame, ECS produced ammonium sulfate (as a 21 g/L N solution) at a cost of $34.40/kg N, which is about 16 times the North American commercial price of solid ammonium sulfate of $2.18/kg N (Figure 5A).¹³⁴ Despite the need for substantial cost reductions, ECS could help enhance global food security in the face of fertilizer price volatility¹³⁴ caused by volatile energy prices and geopolitical disruptions affecting supply chains.^135−138 Since the early 2000s, the price of ammonium sulfate has risen at rates exceeding inflation (e.g., by 2.8 times from 2001 to 2014), and even more dramatic price changes (e.g., a 2.7-time increase from January 2021 to April 2022) have occurred in recent years (Figure S2-53).^139−141 ECS implementation in regions where fertilizers are more expensive or less accessible (e.g., sub-Saharan Africa) may be more economically viable and would improve equity in fertilizer access.^39,142,143 Product diversification toward higher-value products with more applications (e.g., NH₃ vs. (NH₄)₂SO₄) and targeting multiple profitable uses for each chemical (e.g., (NH₄)₂SO₄ as a reagent is about 4–5 times more valuable than as a fertilizer)¹⁴⁴ could further improve process economics.

Figure 5.

Life-cycle cost of total ammonia nitrogen (TAN) Recovery for electrochemical stripping (ECS) treating urine. Capital (CapEx) and operating (OpEx) costs are accounted for in panel (A), while panel (B) shows an expanded version of only operating costs. A comparison of operating costs of ECS to operating costs of other TAN removal and recovery technologies (where NDN is nitrification-denitrification) is shown in panel (C). Detailed breakdowns of capital and operating costs of ECS are shown in Figure S2-52.

The breakdown of the ECS costs provides insights into specific cost reduction strategies. Capital costs accounted for half of the ECS life-cycle cost. Reactor housing and the anode each accounted for about 46–48% of capital costs, while the cathode, CEM, and omniphobic membrane each accounted for about 2% of capital costs (Figure S2-52, Table S1-6). Reactor housing would likely become much less expensive with scaled-up production. Exploring lower-cost but equally efficient anode materials (e.g., manganese oxybromide,¹⁴⁵ cobalt manganese binary oxide,¹⁴⁶ iron oxide-titanium dioxide¹⁴⁷) could reduce capital costs, but a trade-off between capital cost and energy consumption may exist. Omniphobic membrane replacement accounted for the largest fraction (52%) of operating costs, followed by anode replacement (26%), CEM replacement (10%), and electricity (8.8%) (Figure 5B). Lifetimes of all components, including membranes and electrodes, are uncertain because different criteria for replacement can be defined and studies of component lifetime are underrepresented compared to studies of other performance metrics, such as activity and selectivity.¹⁴⁸ Thus, the contribution of electrode and membrane replacement to operating costs may differ from this study’s estimate. Multifaceted efforts are needed to enhance the economic viability of ECS, including advances in low-cost, durable materials; energy-efficient reactor configuration and operation; and recovery of high-value products.

Due to the uncertainty of several parameters that influence costs, we conducted a sensitivity analysis to determine the life-cycle cost for ECS with individual variables set to their cost-minimizing values (Table S2-11). When changing only one variable at a time (as opposed to concurrent sensitivity analyses of multiple variables), the most influential components were the costs of the anode, omniphobic membrane, and reactor housing. These variables influenced overall cost most because reactor housing and the anode dominated capital costs, and omniphobic membrane replacement dominated operating costs, providing additional motivation for material and reactor engineering targeting cost reduction.

Compared to four other TAN removal and recovery processes, ECS operating costs were about 4 times greater than ammonia stripping costs, 8 times greater than costs associated with the most established wastewater TAN removal process (nitrification-denitrification or NDN), and 10–15 times greater than costs of anammox and ion exchange (Figure 5C). We excluded capital costs in these comparisons because estimates varied in the scale and scope of systems considered (Table S2-12). Reported ECS operating costs are likely underestimates because we did not include labor and chemical costs associated with membrane and/or electrode cleaning (nascent and highly variable estimations exist), nor with post-treatment (e.g., for balancing pH in the ECS anode effluent). Other benefits of ECS, such as reduced chemical transport needs compared to conventional ammonia stripping and a circular nitrogen economy facilitating more equitable fertilizer access with lower environmental impacts, make process engineering to reduce capital costs, extend component lifetimes, and accelerate implementation imperative.

4. Conclusions and Implications

We demonstrated that ECS can recover TAN from several waste streams (e.g., FSTP effluent, reverse osmosis concentrate, anaerobically treated wastewater) and can treat urine continuously and consistently for over a month. Future work should target reduced energy demand, reactor configurations that readily facilitate O&M tasks like membrane replacement, and investigation of membrane durability, mechanisms of degradation and failure, and cleaning for mitigating organic and inorganic fouling. Simple, inexpensive, and rapid online process monitoring remains a challenge, motivating further investigation of in situ electrochemical impedance spectroscopy methods that provide an indicator of CEM fouling and degradation that can be measured while ECS is operating. Moreover, multipronged approaches to improve the economic viability of ECS on a life-cycle basis should include reducing reactor component costs (especially for the anode); extending component lifetimes and reducing energy consumption through material, reactor, and process engineering; and targeting recovery of higher-value products specific to an implementation context.¹⁸ Our study provides evidence for urine treatment as a foothold for electrochemical nutrient recovery from wastewater while demonstrating proof-of-concept for broad applicability to many wastewaters. Further research guided by our findings can accelerate the deployment of ECS at increasing scales and at more sites as the process moves along the S-curve of technology adoption toward maturity.¹⁴⁹ Ultimately, findings from this study support the growth of a circular nitrogen economy that fosters progress toward sanitation provision, efficient chemical production, and protection of water resources against pollution.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsenvironau.3c00058.

• Tables describing influent streams, experimental conditions, assumptions for cost estimate, and results of membrane characterization and cost estimate; equations for efficiencies, energy demand, fluxes, product purity, current efficiency, and analysis of electrochemical impedance spectroscopy data; and figures showing reactor setup, temporal trends in performance metrics and composition, membrane characterization results, and cost estimate results (PDF)

Author Contributions

CRediT: Anna Kogler conceptualization, data curation, formal analysis, investigation, methodology, project administration, validation, visualization, writing-original draft, writing-review & editing; Neha Sharma data curation, formal analysis, investigation, methodology, software, validation, visualization, writing-original draft, writing-review & editing; Diana Tiburcio data curation, formal analysis, investigation, visualization, writing-original draft; Meili Gong data curation, formal analysis, investigation, methodology, validation, writing-review & editing; Dean Mackenzie Miller formal analysis, methodology, writing-review & editing; Kindle Shea Williams methodology, visualization, writing-review & editing; Xi Chen methodology, writing-review & editing; William A. Tarpeh conceptualization, funding acquisition, project administration, resources, supervision, writing-review & editing.

The authors declare no competing financial interest.

Special Issue

Published as part of ACS Environmental Au virtual special issue “2023 Rising Stars in Environmental Research”.