Removal of As(III) by Electrically Conducting Ultrafiltration Membranes
- Björn Otto
- 2 days ago
- 31 min read
Updated: 8 hours ago
Shengcun Maa, Fan Yanga, Xin Chena,b,c, Chia Miang Khora, Bongyeon Junga, Arpita Iddyaa, Gaurav Santa,b,c,d,e, David Jassbya,c,*
Abstract
As(III) species are the predominant form of arsenic found in groundwater. However, nanofiltration (NF) and reverse osmosis (RO) membranes are often unable to effectively reject As(III). In this study, we fabricate highly conducting ultrafiltration (UF) membranes for effective As(III) rejection. These membranes consist of a hydrophilic nickel-carbon nanotubes layer deposited on a UF support, and used as cathodes. Applying cathodic potentials significantly increased As(III) rejection in synthetic/real tap water, a result of locally elevated pH that is brought upon through water electrolysis at the membrane/water interface. The elevated pH conditions convert H3ASO3 to H2AsO3⁻/HAsO3²⁻ that are rejected by the negatively charged membranes. In addition, it was found that Mg(OH)2 that precipitates on the membrane can further trap arsenic. Importantly, almost all As(III) passing through the membranes is oxidized to As(V) by hydrogen peroxide produced on the cathode, which significantly decreased its overall toxicity and mobility. Although the high pH along the membrane surface led to mineral scaling, this scale could be partially removed by backwashing the membrane. To the best of our knowledge, this is the first report of effective As(III) removal using low-pressure membranes, with As(III) rejection higher than that achieved by NF and RO, and high water permeance.
1. Introduction
Both chronic and acute exposure to arsenic can be highly hazardous (Jang et al., 2016). Elevated concentration (i.e., >10 μg/L) of arsenic in groundwater have been reported in many countries, including the USA, India, Japan, Germany, and China, with over-extraction of groundwater, brought about by water shortages, increasing the risk of arsenic exposure (Ayotte et al., 2017; Singh et al., 2015; Smith et al., 2018). Arsenic occurs in various oxidation states, including arsine (As3 ), elemental arsenic (As0), arsenite (As(III)), and arsenate (As(V)) (Singh et al., 2015), and the toxicity of arsenic varies with arsenic speciation. As(III) is the most toxic species, and is thermodynamically favorable in reduced redox environment (Singh et al., 2015); As(V) is the less toxic species, and is the predominant species in oxygenated environments (Jang et al., 2016). It is noted that both As(III) and As(V) may occur simultaneously in groundwater due to their slow redox kinetics (Singh et al., 2015).
Various engineering methods have been employed to remove arsenic from drinking water. Because As(III) is more mobile and harder to treat, the oxidation of As(III) to As(V) is often used to achieve arsenic removal targets, when As(III) is the predominant arsenic species (Singh et al., 2015). Oxidation techniques able to convert As(III) to As(V), including chemical oxidation (Dodd et al., 2006), photo-oxidation (Ryu et al., 2013), bio-oxidation (Saalfield and Bostick, 2009), and electrochemical oxidation (Kim et al., 2014), have been widely investigated. However, these oxidative methods require additional solid/liquid separation steps such as sand filtration (Leupin and Hug, 2005) and coagulation/sedimentation (Lee et al., 2003). Flocculation and coagulation are the most widely employed techniques in drinking water plants to directly remove arsenic, the effectiveness of which depends on the types of coagulant, arsenic species, and various operational conditions (e.g., pH) (Singh et al., 2015). It is noted that flocculation, and in particular, aluminum hydroxide-based flocculation, couldn’t effectively remove As (III) (Hering et al., 1997), and As(III) removal by ferric chloride is less efficient compared to As(V) (Hering et al., 1997; Lee et al., 2003). Also, flocculation and coagulation produce large volumes of waste sludge with elevated concentrations of arsenic. More importantly, it is impractical for those who get drinking water from private wells to use a coagulation-based system. Sorption-based treatment of arsenic is sometimes used by small-scale or point of use (POU) treatment systems (Jang et al., 2008). However, the effectiveness of sorption is dependent on the solution pH (Jang et al., 2008; Manna et al., 2003) and phosphate concentrations (Chowdhury and Yanful, 2010), and As(III) removal is generally not as good as As(V) in the neutral pH range (Singh et al., 2015), which limits this approach’s utility.
Many owners of private wells, and indeed, many households in general, utilize membrane units to ensure the safety of their drinking water (George et al., 2006). Nanofiltration (NF) membranes are able to reject up to 99% of As(V), but only achieve 5% to 33% removal of As(III) (EPA, 2000; Seidel et al., 2001; Yoon et al., 2009). It has been reported that RO systems can remove between 5%-84% of As(III) in groundwater (Akin et al., 2011; Ning, 2002), but many household RO systems in the USA remove less than 50% of arsenic when As(III) is the dominant arsenic species (Walker et al., 2008). Importantly, a study that collected samples from 65 wells across the USA found that As(III) is the dominant (>90%) arsenic species in approximately 45% of these wells, and As(III) contributed to more than 50% of the total arsenic in more than 50% of these wells (Sorg et al., 2014), indicating that a large number of people who get drinking water from wells may experience exposure to arsenic contamination. The challenge in removing As(III) with tight membranes is due to a fact that As(III) mainly presents as H3AsO3, as the pKa of H3AsO3 is 9.23, in natural waters-a neutral and highly mobile molecule. NF and RO membranes both rely on charge and size to effectively remove small molecules, with both offering relatively poor rejection of small, uncharged species such as H3AsO3.
Electrically conducting membranes (ECMs) are versatile membranes that are able to induce electrochemical reactions directly at the membrane/water interface (Zhu and Jassby, 2019). These electrochemical reactions can enhance the rejection of pollutants by modifying the local pH along the membrane surfaces through water electrolysis (Jung et al., 2020), as well as triggering redox transformations of contaminants (Duan et al., 2017; Le et al., 2019). ECMs can be fabricated with organic and inorganic substances (e.g., conductive polymers, carbon nanotubes (CNTs), and metal or metal oxides) (Almassi et al., 2018; Ronen et al., 2016). CNT-based ECMs have been fabricated using pressure deposition (Duan et al., 2016) or spray coating approaches (Dudchenko et al., 2017; Tang et al., 2017). In addition, previous work has demonstrated that ECMs are capable of resisting multiple forms of fouling, including biofouling (Ronen et al., 2015), organic fouling (Dudchenko et al., 2014), and mineral scaling (Duan et al., 2014). Recently, nickel-coated CNT (Ni-CNT) based ECMs were employed to catalyze water splitting and enhance the local pH near the membrane surface (Hou et al., 2018; Iddya et al., 2020).
In this study, we fabricated highly electrically conducting UF membranes for effective As(III) rejection. These UF membranes consisted of a hydrophilic Ni-CNT layer deposited on a hydrophilic polysulfone support. We demonstrate that the application of cathodic potentials to the membrane surface drives the hydrogen evolution reaction (HER), catalyzed by the Ni coating, that leads to a local pH increase, shifts H3AsO3 to its ionic form, and enables electrostatic repulsion of this species by the membrane. As(III) species that do pass through the membrane are transformed to As(V) species through electrochemically-generated H2O2 (a result of oxygen reduction on the membrane/cathode), which reduces the mobility and toxicity of any arsenic that is not rejected. Mg/Ca hydroxides and/or carbonates precipitated on the membrane surface (a result of the locally elevated pH conditions), but membrane performance could be restored with a backwashing step. To the best of our knowledge, this is the first study to report high As(III) rejection with porous membranes having high water permeance without using oxidative pre-treatment.
2. Materials & methods
2.1. Chemicals and instruments
Sodium dodecylbenzenesulfonate (DDBS, technical grade), nickel chloride hexahydrate (ACS reagent grade), sodium (meta)arsenite (>90%), and sodium arsenate dibasic heptahydrate (ACS reagent grade, >98%) were purchased from Sigma Aldrich (St. Louis, MO). Boric acid (ACS reagent grade, Fisher) and nickel sulfate heptahydrate (98%, Alfa Aesar) were purchased from Fisher Scientific and used as received. Tap water was obtained at UCLA and used in experiments without further treatment. Cations (e.g., Na, Mg, Ca, and K) in the tap water were measured by inductively coupled plasma-optical emission spectrometry (ICP-OES) (ICPE-9000, Shimadzu), while anions (e.g., F⁻ , Cl⁻ , Br⁻ , NO3⁻ , SO4²⁻ , and HPO4²⁻ ) were measured by ion chromatography (IC) (Dionex™ Integrion™ HPIC™ System) equipped with an AS-19 column. The concentrations of the main ions in the tap water can be found in Table 1. Synthetic solutions were prepared using deionized (DI) water and ACS certified salts (NaCl, Na2SO4, CaSO4, and MgSO4), purchased from Fisher Scientific. The exact composition of each synthetic solution is listed in Table S1. Total arsenic was analyzed by inductively coupled plasma mass spectrometry (ICP-MS) (PerkinElmer NexION 2000). A combination of high-performance liquid chromatography (HPLC) (Altus, PerkinElmer) and ICP-MS was used to differentiate As(III) from As(V) (Londesborough et al., 1999). Hydrogen peroxide was measured with a N,N‑diethyl-p-phenylenediamine (DPD) colorimetric method with CHEMets® Visual Kit (Catalog No.: K5502), with a detection limit of 25 µg/L. The conductivity and pH of water samples was measure using a conductivity (Thermo Scientific™ Orion Star™ A222) and pH meter (Thermo Scientific™ Orion Star™ A221 pH Portable Meter).
2.2. Membrane preparation and characterization
UF membranes coated with a 6 µm-thick CNT layer were fabricated via a pressure-deposition method, the details of which were described previously (Dudchenko et al., 2014). In brief, 45 mL of 1 g/L multi-wall CNTs functionalized with 7.0% carboxylic groups (Cheaptubes Inc., Brattleboro, VT) were pressure deposited on PS-35 UF membranes (Solecta Membranes Inc., Oceanside, CA) at a pressure of 60 psi (Dud-chenko et al., 2014). Ni was electrodeposited on the CNT UF membranes using a similar method to that reported in our previous research (Hou et al., 2018; Iddya et al., 2020). Briefly, the CNT UF membrane was taped to a stainless steel mesh (Hou et al., 2018; Iddya et al., 2020), that was used as a cathode, while a nickel 200 plate (McMaster-Carr, Atlanta, GA, USA) was used as an anode in an electro-deposition solution containing 23 g/L NiSO4, 3.2 g/L NiCl2, and 31 g/L H3BO3 (Hou et al., 2018; Iddya et al., 2020); a constant current of 20 mA was supplied by an external power source (model: KA3005P, KORAD brand, Shenzhen, China) for 3 h to drive nickel deposition on the CNT surface. The initial pH of electro-deposition solution was 5.79 ±0.01,and the final pH of electro-deposition solution was 5.97 ±0.12 after a 3-h electrodeposition.
A four-point conductivity probe (Veeco, Plainview, NY) was used to measure membrane sheet resistance. Contact angle measurements were conducted using a contact angle goniometer (Model: 250,rame´-hart instrument co., Succasunna, NJ). A scanning electron microscopy (SEM, ZEISS, Supra 40VP SEM; Oberkochen, DE) equipped with an energy-dispersive X-ray (EDAX) detector was used to characterize the membrane’s surface morphology and elemental composition. Linear sweep voltammetry (LSV), cyclic voltammetry (CV) and open circuit potential (OCP) measurements were conducted with a potentiostat (CH Instruments 6005E; Austin, TX), with a Ag/AgCl electrode (Fisher brand) used as a reference. The pore size of the Ni-CNT membrane was determined using SEM micrographs with ImageJ software. The OCP measured between a working electrode (WE) and reference electrode (RE) was 0.06 V, measured in a beaker with tap water, in which the distance between the WE and counter electrode (CE) was 4 mm (similar to the distance in the membrane module). To determine the relative potential (vs. a Ag/AgCl RE), the potential difference between the WE (i. e., the membrane) and a CE (a platinum-coated titanium sheet) was measured using a multimeter, while the voltage difference between the WE and the WE was read using a potentiostat (CH Instruments 6005E; Austin, TX). A summary of the different cell potentials and the corresponding relative cathodic potentials vs. a Ag/AgCl reference are pro vided in Table 2.
2.3. Reactor and operations
A modified crossflow filtration module was employed in all experiments, with details of this module described elsewhere (de Lannoy et al., 2013; Duan et al., 2017). Briefly, the cell contained built-in titanium electrodes that supply an electrical potential to the membrane surface (and do not come into contact with the feed stream), which has an effective area of 40 cm², and to a platinum-coated titanium counter electrode located 4 mm above the membrane surface (de Lannoy et al., 2013; Duan et al., 2017). All membranes used in this study were initially compressed at 100 psi (i.e., 689 kPa) with a diaphragm pump (Hydra-Cell, MN) for at least 12 h before use. A gear pump (Cole-Parmer, IL, USA) with a flow rate of 11.3 ±0.8 mL/s (corresponding to a cross-flow velocity of 7.0 ±0.5 cm/s) was used to conduct the filtration experiments. In membrane performance tests, real tap water was used, while synthetic tap water was used when investigating the mechanistic aspects of the observed phenomena. Arsenic concentrations in groundwater vary greatly from one location to another, with a concentration ranging from 0.5 to 5000 µg/L (Singh et al., 2015). In our experiments, 750 µg/L As(III) were spiked into both real tap water and synthetic water solution; this concentration has been previously used in other studies investigating the removal of As(III) from water (Akin et al., 2011). Due to differences in membrane permeance between the CNT and Ni-CNT membranes, to achieve comparable fluxes in these membranes 0.9 ± 0.1 psi (i.e., 6.2 ±0.7 kPa) and 4.1 ±2.9 psi (i.e., 28.2 ±20. 0 kPa) were used for CNT (8.9 ±1.0 L/(m2•h)) and Ni-CNT (9.1 ±0.4 L/(m2•h)) membranes, respectively. For As(III) rejection in real tap water, experiments were conducted for 2h before samples at OCP (i.e., 0 V applied) were collected; then, 3,5 and 7 V cell potentials were sequentially applied to the membrane/counter current collectors with membrane as the cathode via an external power supply (model: KA3005P, KORAD brand, Shenzhen, China). A peristaltic pump (model: RF100, Greylor, FL, USA) was used for membrane backwashing at a pressure of 8 psi (i.e., 55.1 kPa). In scaling experiments, the system was run for 100 min with 7 V cell potential applied (membrane as cathode), followed by a 5-minute crossflow flushing and a 5-minute backwashing with DI water. All experiments were conducted in triplicate.
2.4. Potential profile on the membrane surface
It was reported that the overall resistance of a CNT thin film has a linear relationship with its length (Blighe et al., 2007). Therefore, a linear potential profile on the membrane surface can be calculated using the current and resistance of the membrane/electrode soaked in water. The current was measured with a multimeter. The resistance of the membrane soaked in solution was calculated using an equivalent circuit model, where the membrane and solution are recognized as resistors in parallel (Zhu and Jassby, 2019). With this model, the total resistance was calculated using Kirchhoff’s Law:
Where Rtotal, Rmembrane, and Rsolution represent the total system resistance, membrane resistance, and solution resistance, respectively. Because the solutions used in this work (real and synthetic tap water) had a significantly lower conductivity (~0.04 S/m) than the CNT membranes (~2280 S/m), let alone Ni-CNT membranes (~29,976 S/m), it is reasonable to assume that Rtotal is close to Rmembrane(i.e., 1/Rsolution is ignored in Equation Resistance).
The potential drop on the membrane was calculated with ohm’s law
V=I ∗ Rtotal
Where V is voltage drop (V), and I is the current (A)
2.5. Models to predict pH on the membrane surface
The pH on the membrane surface was simulated with COMSOL Multiphysics© using adaptive time-stepping, equilateral triangle mesh elements (mesh opening of 173.21 µm²), and periodic boundary conditions. The membrane was modeled as a planar electrode that generates OH ions through the water electrolysis reaction:
The OH⁻ generation rate was calculated from the cell current using Faraday’s equation:
where jOH is the flux of OH⁻ , z is the charge of OH , i is the current, and F is Faraday’s constant. We assumed the distribution of OH⁻ is only governed by Fick’s first and second laws:
where DOH is the diffusion coefficient of OH⁻ (taken as 5.27e-5 cm²/s. (Haynes, 2014)), and COH is the concentration of OH⁻ . Consequently, the pH at the membrane surface can be calculated as the function of time (see Figure S2). Note that, for laminar flow conditions, the crossflow rate adjacent to the membrane surface is small,(Leal, 1992) and therefore, OH⁻ transport through convection was not considered. In addition, due to the large potential drop (Bard and Faulkner, 2001) at the mem brane/solution interface (i.e., owning to the formation of the electrical double layer, EDL), ion migration was not considered for simplicity. It is noted that oxygen reduction is occurring along the membrane. However, both oxygen reduction reactions (Equations S1 and S2) will not impact the Faradaic efficiency used to simulate the pH because the products of these oxygen reduction reactions are still OH⁻ .
2.6. Mineral speciation under different pH and cation concentrations
Geochemists’ Workbench Student (14.0) was used to calculate magnesium species distribution versus pH and calcium concentrations; an open carbonate system was assumed.
3. Results and discussion
3.1. Membrane characterization
The electro-deposition of nickel imbued the black CNT UF membranes with a metallic sheen (Figure S1). The contact angle of a plain PS-35 membrane was reported as 49.2 ± 0.9◦ in our previous study (Dudchenko et al., 2014), while the deposition of CNTs increased the contact angle to 113.7 ± 6.0◦ (Fig. 1A), due to the hydrophobicity of CNTs (Dudchenko et al., 2014). However, the electro-deposition of nickel decreased the contact angle to 68.2 ±1.6◦(Fig. 1B), which corresponds well with previous reports (Iddya et al., 2020). The CNTs formed a porous network with pores ranging in diameter between 10 and 100 nm (Dudchenko et al., 2014), and our previous research suggests that it is the PS-35 UF support, rather than the CNT network, that governs the composite material’s rejection (Dudchenko et al., 2014), which has a mean pore size of 17 nm (Aher et al., 2017). The Ni-CNT network has a mean pore size of 11 nm, measured from the SEM micrograph (Fig. 1B). However, nickel didn’t form a dense film, and as a result, the CNT network was still visible after the electro-deposition step (Fig. 1B). Although nickel reduced the membrane’s water permeance from 8 L/(m2•h)/psi (i.e.,1.2 L/(m2•h)/kPa) to 2L/(m2•h)/psi (i.e., 0.3 L/(m2•h)/kPa)), nickel deposition dramatically decreased the electrical resistivity of the membranes from 73.1±7.1 ohms/square to 5.6±5.5 ohms/square. This decreased electrical resistance causes a significantly reduced potential drop across the membrane compared to CNT UF membranes (Fig. 1D), which was calculated based on current density measurements acquired during the real tap water test (Section 3.2). The detailed discussions are described in SI. We note that the potential drop differences between Ni-CNT UF and CNT UF membranes increased with increasing potential (Fig. 1D) due to the larger current densities at higher potentials. In addition, the lower resistance of Ni-CNT UF membranes compared to CNT UF membranes resulted in a higher current density, as measured by LSV in tap water (Fig. 1E). CV curves revealed that Ni-CNT UF membranes experienced a lower water electrolysis onset potential and higher currents compared to the CNT UF membranes, likely because of the catalytic properties of the deposited nickel towards the HER (Fig. 1F) (Hou et al., 2018).
3.2. As(III) removal performance and mechanism
To test As(III) removal by CNT UF and Ni-CNT UF membranes in real water samples, 750 µg/L of As(III) (sodium (meta) arsenite) was spiked into Los Angeles tap water. It is noted that tap water in Los Angeles is a blend composed of groundwater and surface water. In addition, our tap water has bicarbonate concentrations (shown in Table 1) that fall within the range associated with groundwater (Saha et al., 2019). Therefore, it is reasonable to use this tap to represent groundwater. To investigate the removal mechanisms, As(III) removal and arsenic speciation (i.e., As (V) and As(III)) were evaluated in different synthetic salt solutions (e.g., Na2SO4, NaCl, and a mixture of Na2SO4, MgSO4, and CaSO4) that had similar conductivity (404.9±3.6 μS/cm) to real tap water (358.0±0.9 μS/cm). In tap water tests, experiments were run for 2 h before samples were collected at OCP to investigate As(III) removal of the uncharged membranes. Then, 3, 5, and 7 V cell potentials (membrane as cathode) were sequentially applied, with each potential being applied for 30 min. Under OCP conditions, As(III) removal by the CNT UF membrane was 8.9% ±1.7% (Fig. 2A), suggesting that sorption could remove a limited amount of As(III). This corresponds with previous CNT adsorption studies, which showed that CNTs have a limited adsorption capacity towards As(III) (i.e., 13.5 µg/g) (Ali, 2018).
Negative potentials applied to the CNT UF membranes significantly increased As(III) rejection, with rejection increasing with an increase of the applied potential (Fig. 2A). With a low operating pressure (i.e., 0.9 ±0.1 psi (i.e., 6.2 kPa ±0.7 kPa), 8.9 ±1.0 L/(m2•h)), CNT UF membranes achieved As(III) removal as high as 72.6% ±5.1% when 7 V was applied; lower potentials yielded lower rejection (Fig. 2A), but these rejection values were still significantly higher than As(III) rejection reported for by the majority of high-pressure NF or RO membranes (Fig. 2B and Table 3). The flux of the CNT UF membranes slightly increased with an increase in potential (i.e., when the applied cell potential was >5 V in Fig. 2C), likely because the applied potential increased the local pH, which further deprotonated carboxyl groups on the CNTs and made the membrane more hydrophilic. Over time, higher potentials led to flux decline (93.4% ±10.3% of original flux) due to scaling caused by the high local pH. It is noted that the pH of the feed is supposed to be constant because all the permeate was returned to the feed. However, the feed pH slightly dropped from 7.69 to 7.23 over the course of the experiment, likely due to the precipitation of Mg(OH)2 on the membrane surface.
Ni-CNT UF membranes further enhanced As(III) rejection in tap water to 93.3% ±3.9% when a cell potential of 7 V was applied, with a flux of 9.1±0.4 L/(m2•h) (operating pressure: 4.1±2.9 psi (i.e., 28.2±20.0 kPa)) (Fig. 2A), which is close to the minimum removal (i.e., 98.7% for 750 µg/L As(III)) needed to achieve the MCL (10 µg/L) as set by the EPA. It is worth noting that we tested our membranes under highly challenging conditions (i.e.,>90% of total arsenic was in As(III) form). These conditions only occur in 29 of 65 wells investigated across the USA (Sorg et al., 2014), while As(V) is the predominant arsenic species in 31 of 65 wells (Sorg et al., 2014). Considering that As(V) is much easier to be removed than As(III) by membranes (Singh et al., 2015), that As concentrations ranged between 19 – 89 µg/L in the US well water survey, and that we could remove 93.3%±3.9% of As(III), it is reasonable to expect that the MCL could be achieved by our membranes in most of cases (Sorg et al., 2014). The flux through the Ni-CNT UF membrane declined with increasing cathodic potentials (Fig. 2C). This is likely a result of mineral scaling that forms on the membrane surface – a result of the locally elevated pH conditions.
The low As(III) rejection achieved by conventional UF/NF/RO membrane processes is due to the uncharged form of As(III) (H3AsO3) in most natural systems (the first pKa of H3AsO3 is 9.23(Sadiq et al., 1983)) (Ma et al., 2020). However, the application of cathodic potentials in creases the local pH on the CNT UF and Ni-CNT UF membrane surface, as evident by the elevated pH of the permeate (pH of 10.9±0.1 and 11.8± 0.1 for the CNT UF and Ni-CNT UF, respectively). This pH is higher than the first pKa of As(III) (i.e., 9.23), converting H3AsO3 to its charged form, H2AsO3 .This ionic form of As(III) (i.e., H2AsO3 ) is rejected by the negatively charged membranes due to powerful electrostatic repulsive forces induced by the applied potential (Dudchenko et al., 2014). Importantly, the pH at the membrane/water interface is likely higher than these values, but is impossible to be measured directly. Based on our simulations (Section 2.5), the pH at the membrane surface was close to 13 (Figure S2). This pH is near/above the second pKa of H3AsO3 (i.e., 12.13), potentially leading to the formation of HAsO3²⁻,whose increased negative charge would enable even more efficient electrostatic repulsion and higher rejection.
In addition to transforming As(III) to its charged form, the elevated pH conditions along the membrane surface led to the precipitation of metal hydroxides and carbonates (e.g., Mg(OH)2, CaCO3), evident by the flux decline at higher applied potentials (Fig. 2C), which was particularly apparent when the Ni-CNT UF membranes were tested. However, metal hydroxides can play an important role in arsenic removal. Fig. 2D shows As(III) rejection by CNT UF membranes under different synthetic salt solutions having similar conductivity to tap water. The presence of Mg and Ca in the feed significantly enhanced the rejection of As(III) (Fig. 2D). We speculate that this is because precipitates formed by Mg/ Ca (e.g., CaCO3 and/or Mg(OH)2) can adsorb As(III). The Geochemist Workbench software package predicted that Mg(OH)2 was the dominant mineral formed on the surface of the membrane (i.e., when pH was over 10.5, and the concentrations of Mg²⁺and Ca²⁺were 10 ppm and 50 ppm, respectively) (Fig. 2E). Therefore, we hypothesize that Mg(OH)2 is the main species responsible for arsenic adsorption. Previous studies have reported that Mg(OH)2 is an effective sorbent for both As(V) and As (III) at high pH (i.e., pH =11) (McNeill and Edwards, 1997). A previous study has also shown that nanostructured MgO has a high removal capacity for arsenic (Liu et al., 2011). EDX analysis of a scaled membrane surface confirmed that the accumulated scale was primarily composed of Mg-containing minerals (Figure S3B&D). It is possible that competing ions (e.g., phosphate) could influence the adsorption of arsenic (Pincus et al., 2020). However, the concentration of phosphate was low in the tap water (Table 1). In addition, As(III) rejection in the presence of NaCl was slightly lower than that in the presence of Na2SO4 (Fig. 2D), which is possibly because reactive chlorine species (RCS) formed on the anode compete with the water reduction reaction on the cathode (Kim et al., 2014), which led to a lower pH along the membrane that reduced As(III) ion formation.
The conversion of As(III) to As(V) is desirable because As(V) is less toxic and mobile than As(III) (Sharma and Sohn, 2009). As(III) may be directly oxidized on the anode or oxidized by RCS, or other oxidizing agents, formed on the anode or cathode (Kim et al., 2014). In the presence of chloride ions (i.e., in the NaCl solution), As(V) concentrations in the feed increased from 18%±4% to 30%±3% from 30 min to1h (Fig. 2F), while As(V) levels only increased slightly from 14%±1% to 17%±3% when the feed was Na2SO4 (Fig.2F). This indicates that direct oxidation of As(III) by the platinum-coated titanium anode was slow, and RCS produced from chloride oxidation were largely responsible for the oxidation of As(III) in the bulk feed (Kim et al., 2014).
Interestingly, nearly 100% of arsenic in both Na2SO4 and NaCl feed solutions was oxidized to As(V) in the permeate (Fig. 2F), indicating that As(III) was primarily oxidized to As(V) as it passed through the cathodic membrane rather than at the anode or in the bulk feed (e.g., by RCS). This is counter-intuitive as reduction reactions are typical on the cathodic membranes. We speculate that hydrogen peroxide, produced through the electroreduction of oxygen, was responsible for the oxidation of As(III) to As(V) (Qian et al., 2015); we measured approximately 50 µg/L H2O2 in the permeate. Oxygen, the precursor of hydrogen peroxide, could come from the air or the anode, where water splitting reactions produce oxygen, and oxygen can diffuse to the cathode (Kim et al., 2014). Oxidation of As(III) by electrochemically generated H2O2 on a cathode was first observed and studied by Qian et al. in 2015, who noted that high pH, the presence of humic acid, HCO3⁻ , Ca²⁺, and Mg²⁺ could enhance the oxidation process (Qian et al., 2015). Since these conditions are commonly found in drinking water sources, this mechanism can explain the conversion of As(III) to As(V) as the compounds move through the membrane/cathode. Unfortunately, it is impossible to rule out the possibility that oxidization of As(III) to As(V) occurs during preparation of ICP-MS samples (i.e., stabilizing samples with 5% nitric acid). The presence of oxygen is inevitable because the oxygen evolution reaction occurs on the anode, and this gas may diffuse to the cathode. Therefore, experiments in zero-oxygen conditions are difficult to accomplish. Also, our results indicated that As(V) wasn’t reduced to As (III) by our cathodic membranes, likely because of As(V)’s electrochemical inactivity (Huiliang et al., 1988).
It is noted that nickel layer on the top of CNTs is a mixture of Ni(0) and Ni(OH)2 (Hou et al., 2018). Ni(OH)2 is reported to be stable in the presence of H2O2 (Yan et al., 2012), and Ni(0) is also highly resistant to H2O2 in neutral or alkaline conditions (Wilbraham et al., 2018).Importantly, there is a highly reductive and alkaline environment around the Ni-CNT UF membranes, as we are applying cathodic potentials, suggesting that the leaching of Ni²⁺ is highly unlikely. Also, a recent study found that little Ni dissolved at pH 12 after 50 h operating with a Ni activated carbon cathode (Kim et al., 2018). Therefore, it is reasonable to conclude that Ni-CNT UF membranes are resistant to trace H2O2 formed in the cathode.
3.3. Scaling and cleaning
While electroactive membranes have been demonstrated to be able to prevent multiple forms of fouling, including biofouling (Ronen et al., 2015), and organic fouling (Dudchenko et al., 2014), the application of high cathodic potentials, such as were used here to achieve good rejection, inevitably lead to the formation of mineral scaling (by Mg(OH)2 and CaCO3) on the membrane surface (Fig. 3). Sequential testing of a Ni-CNT UF membrane to remove As(III) from tap water showed a 36%±15% decline in flux after 300 min (Fig. 3A). A 5-minute backwash and a 5-minute crossflow with DI water could recover the membrane f lux for three sequential cycles (Fig. 3A). However, the flux decline in each of these sequential cycles was more rapid than the first round, with flux declining by approximately 46%±3% (average decline of 2nd, 3rd, and 4th run) within 100 min, indicating that the cleaning process led to incomplete scale removal. More rigorous chemical cleaning (e.g., a periodic weak acid rinse under cathodic protection(Pierozynski et al., 2009)) would likely be needed to completely recover the membrane surface. It is critical to note that the backwash water is toxic as it contains arsenic adsorbed by magnesium hydroxide. It has been reported that such materials (arsenic sorbed onto Mg(OH)2 ) can be used as additives in cement manufacturing, where they can be safely sequestered in concrete products (Tresintsi et al., 2014).
3.4. Economic analysis
The treatment costs of arsenic in drinking water are highly dependent on the dominant As species and the technology used, and span a very wide range - between near-zero (e.g., laterite absorption) to as high as $70/m³ (Shan et al., 2019). Several highly affordable engineering methods (e.g., coagulation and flocculation) are available to effectively remove As(V) under normal drinking water conditions. However, there are few options to cheaply remove As(III) directly (i.e., without pre-oxidization steps), particularly using point-of-use applications. For example, even household RO, considered by many as the gold-standard of point-of-use systems, can only achieve 50% As(III) removal (Walker et al., 2008), with other treatment technologies, such as NF, achieving even lower removal. In contrast, our electroactive UF membranes are highly effective at removing As(III). As a result, it is not straight-forward to compare the treatment costs of water dominated by As(III) contamination.
In our treatment system, the additional cost of electricity plays an important role in the overall operating cost of the process, with increasing energy associated with higher voltages; these higher voltages also enable higher rejection of As(III) (Fig. 2A). In other words, the energy required to achieve acceptable permeate As(III) concentrations (i.e., 10 µg/L) increases linearly with increasing As(III) concentrations in the feed stream (Fig. 4); in this figure, the required energy to reduce the permeate concentration to below 10 µg/L was calculated based on the experimental data presented in Fig. 2A. To account for the average lower flux caused by the membrane fouling, we assumed that the average flux is 60% of the initial flux, which is the average flux during the 4 experimental runs. In addition, based on a recent EPA study, contaminated well water in the USA contains anywhere between 0.3–68 µg/L As(III) (Sorg et al., 2014). Based on these initial concentrations and an average flux (i.e., 60% of the initial flux), the energy required to treat this water ranges between 1.94 kWh/m³ for the case where the initial concentration is just above the maximum allowable limit(11 µg/L),to 12.00kWh/m³ if the As(III) concentration is 68 µg/L. Based on the average cost of household electricity in the USA ($0.167/kWh for the USA West Coast), the additional electrical costs range between $0.32/m³ and $2.00/m³. In addition, this cost is highly sensitive to flux (Duan et al., 2017) (e.g., the additional electrical costs could drop to between $0.09/m³ (i.e., 0.57 kWh/m³) and $0.59/m³ (i.e., 3.56 kWh/m³) if the initial flux increased to 27L/(m²•h), which is a reasonable flux for a UF process).
Few studies report the total cost (i.e., the sum of capita and operating costs) of a household UF system; a single study reports that the total cost of a household UF treatment system is approximately $0.10/m³ of product water (Praneeth et al., 2014), but the system contains a hand-pump and is not particularly relevant when pressurized water lines are available (as is the case in the vast majority of the developed world). It is reported that the cost of household RO system is as high as $3.4/m³ of product water (Wang et al., 2011), while the cost of industrial-scale RO-treated water ranges from 0.38 to 0.52 $/m³ produced water (Jung et al., 2020). If we assume that the cost of household UF scales similarly to household RO (the ratio of household RO cost to industrial RO costs ($3.4/m³ of 3 to $0.45/m³ ), we can estimate the cost of a household UF system from the cost of an industrial UF drinking water treatment plant (i.e., $0.28/m³)(Duan et al., 2017). Therefore, the total cost of a household UF system is estimated at $2.1/m³ Considering an additional energy-related operating cost of $0.35/m³ (i.e., 2.07 kWh/m³ ), calculated assuming an initial As(III) concentration of 39.50 µg/L (i.e., the mid-point between 11 µg/L and 68 µg/L), and additional material cost (i.e., from the addition of CNTs and nickel) of approximately $0.03/m³ (assumes membrane could be used for 1 year - see detailed calculations in the SI), the total cost (i.e., operating and capital cost) of using household Ni-CNT UF membranes to treat arsenic contaminated water is $2.48/m³.It is noted that the majority of this cost is from the capital costs of a household UF membrane system, This cost will be significantly lower for industrial UF systems. While this number is speculative, it does provide a framework for the solution of As(III) removal at the household level. It is critical to note that the solution proposed here is, to the best of our knowledge, the only technological solution capable of removing As(III) at a household level without pre-oxidizing As(III). The oxidation of As(III) to As(V), needed to enable other treatment technologies, adds another layer of complexity and cost to the process.
4. Conclusion
This study demonstrates an effective, membrane-based, low pressure As(III) removal technology. OH⁻ ions, produced locally on the membrane surface, transform neutral H3AsO3 to its ionic forms (H2AsO3⁻/H2AsO3⁻) that are effectively rejected by negatively charged Ni-CNT UF or CNT UF membranes. In addition, Mg(OH)2 precipitates on the membrane surface further adsorb As(III), which increases removal. Importantly, nearly all As(III) passing through the membrane is oxidized to As(V) (transformed by hydrogen peroxide produced on the cathodic membranes), which decreases arsenic’s mobility and toxicity. Mineral scaling formed on the membrane could be partially removed by physical cleaning (backwash and cross-flushing), but more rigorous chemical cleaning (e.g., an acid rinse) would likely be needed periodically to completely recover the membrane surface.
Declaration of Competing Interest
There are no conflicts of interests to declare.
Acknowledgements
The study was financially supported by the Department of Energy under award number FE0030456. Many thanks go to Prof. Eric Hoek at UCLA and our colleagues in the lab (Yiming Liu, Yiming Su, Unnati Rao, Xiaobo Zhu, and Jingbo Wang). The authors acknowledge the use of ICP- MS core facility within the UC Center for Environmental Implications of Nanotechnology in CNSI at UCLA. Also, we would like to thank the graduate scholarship for Shengcun Ma from CA-NV AWWA Scholarship Program.
References
Aher, A., Cai, Y., Majumder, M., Bhattacharyya, D., 2017. Synthesis of graphene oxide membranes and their behavior in water and isopropanol. Carbon 116, 145–153.
Akin, I., Arslan, G., Tor, A., Cengeloglu, Y., Ersoz, M., 2011. Removal of arsenate [As (V)] and arsenite [As (III)] from water by SWHR and BW-30 reverse osmosis. Desalination 281, 88–92.
Ali, I., 2018. Microwave assisted economic synthesis of multi walled carbon nanotubes for arsenic species removal in water: Batch and column operations. Journal of Molecular Liquids 271, 677–685.
Almassi, S., Li, Z., Xu, W., Pu, C., Zeng, T., Chaplin, B.P, 2018. Simultaneous Adsorption and Electrochemical Reduction of N-Nitrosodimethylamine Using Carbon-Ti4O7 Composite Reactive Electrochemical Membranes. Environmental science & technology 53 (2), 928–937.
Ayotte, J.D., Medalie, L., Qi, S.L., Backer, L.C., Nolan, B.T, 2017. Estimating the High- Arsenic Domestic-Well Population in the Conterminous United States. Environ Sci Technol 51 (21), 12443–12454.
Bard, A.J., Faulkner, L.R, 2001. Fundamentals and applications: Electrochemical methods. Electrochemical Methods 2, 13–15.
Blighe, F.M., Hernandez, Y.R., Blau, W.J., Coleman, J.N, 2007. Observation of Percolation-like Scaling – Far from the Percolation Threshold – in High Volume Fraction, High Conductivity Polymer-Nanotube Composite Films. Advanced Materials 19 (24), 4443–4447.
Chowdhury, S.R., Yanful, E.K, 2010. Arsenic and chromium removal by mixed magnetite–maghemite nanoparticles and the effect of phosphate on removal. Journal of Environmental Management 91 (11), 2238–2247.
de Lannoy, C.F., Jassby, D., Gloe, K., Gordon, A.D., Wiesner, M.R, 2013. Aquatic biofouling prevention by electrically charged nanocomposite polymer thin film membranes. Environ. Sci. Technol. 47 (6), 2760–2768.
Dodd, M.C., Vu, N.D., Ammann, A., Le, V.C., Kissner, R., Pham, H.V., Cao, T.H., Berg, M., Von Gunten, U., 2006. Kinetics and mechanistic aspects of As (III) oxidation by aqueous chlorine, chloramines, and ozone: relevance to drinking water treatment. Environmental science & technology 40 (10), 3285–3292.
Duan, W., Chen, G., Chen, C., Sanghvi, R., Iddya, A., Walker, S., Liu, H., Ronen, A., Jassby, D., 2017. Electrochemical removal of hexavalent chromium using electrically conducting carbon nanotube/polymer composite ultrafiltration membranes. Journal of Membrane Science 531, 160–171.
Duan, W., Dudchenko, A., Mende, E., Flyer, C., Zhu, X., Jassby, D., 2014. Electrochemical mineral scale prevention and removal on electrically conducting carbon nanotube–polyamide reverse osmosis membranes. Environmental Science: Processes & Impacts 16 (6), 1300–1308.
Duan, W., Ronen, A., de Leon, J.V., Dudchenko, A., Yao, S., Corbala-Delgado, J., Yan, A., Matsumoto, M., Jassby, D., 2016. Treating Anaerobic Sequencing Batch Reactor Effluent with Electrically Conducting Ultrafiltration and Nanofiltration Membranes for Fouling Control. Journal of Membrane Science 504, 104–112.
Dudchenko, A.V., Chen, C., Cardenas, A., Rolf, J., Jassby, D., 2017. Frequency-dependent stability of CNT Joule heaters in ionizable media and desalination processes. Nature Nanotechnology 12 (6), 557–563.
Dudchenko, A.V., Rolf, J., Russell, K., Duan, W., Jassby, D., 2014. Organic fouling inhibition on electrically conducting carbon nanotube–polyvinyl alcohol composite ultrafiltration membranes. Journal of Membrane Science 468, 1–10.
EPA, U., 1999. Technologies and costs for removal of arsenic from drinking water. EPA, 815-P-01-001.
George, C.M., Smith, A.H., Kalman, D.A., Steinmaus, C.M, 2006. Reverse osmosis filter use and high arsenic levels in private well water. Archives of environmental & occupational health 61 (4), 171–175.
Haynes, W.M., 2014. CRC handbook of chemistry and physics. CRC press.
Hering, J.G., Chen, P.-Y., Wilkie, J.A., Elimelech, M., 1997. Arsenic removal from drinking water during coagulation. Journal of Environmental Engineering 123 (8), 800–807.
Hou, D., Iddya, A., Chen, X., Wang, M., Zhang, W., Ding, Y., Jassby, D., Ren, Z.J., 2018. Nickel-Based Membrane Electrodes Enable High-Rate Electrochemical Ammonia Recovery. Environ. Sci. Technol. 52 (15), 8930–8938.
Huiliang, H., Jagner, D., Renman, L., 1988. Flow potentiometric and constant-current stripping analysis for arsenic(V) without prior chemical reduction to arsenic(III). Analytica Chimica Acta 207, 37–46.
Iddya, A., Hou, D., Khor, C.M., Ren, Z., Tester, J., Posmanik, R., Gross, A., Jassby, D., 2020. Efficient ammonia recovery from wastewater using electrically conducting gas stripping membranes. Environmental Science: Nano 7 (6), 1759–1771.
Jang, M., Chen, W., Cannon, F.S, 2008. Preloading hydrous ferric oxide into granular activated carbon for arsenic removal. Environmental science & technology 42 (9), 3369–3374.
Jang, Y., Somanna, Y., Kim, H., 2016. Source, distribution, toxicity and remediation of arsenic in the environment–a review. International Journal of Applied Environmental Sciences 11 (2), 559–581.
Jung, B., Kim, C.Y., Jiao, S., Rao, U., Dudchenko, A.V., Tester, J., Jassby, D., 2020. Enhancing boron rejection on electrically conducting reverse osmosis membranes through local electrochemical pH modification. Desalination 476, 114212.
Praneeth, K., Bhargava Suresh, K., Tardio, J., Sridhar, S., 2014. Design of novel ultrafiltration systems based on robust polyphenylsulfone hollow fiber membranes for treatment of contaminated surface water. Chemical Engineering Journal 248, 297–306.
Kang, M., Kawasaki, M., Tamada, S., Kamei, T., Magara, Y., 2000. Effect of pH on the removal of arsenic and antimony using reverse osmosis membranes. Desalination 131 (1-3), 293–298.
Kim, J., Kwon, D., Kim, K., Hoffmann, M.R, 2014. Electrochemical production of hydrogen coupled with the oxidation of arsenite. Environ Sci Technol 48 (3), 2059–2066.
Kim, K.Y., Yang, W., Logan, B.E, 2018. Regenerable Nickel-Functionalized Activated Carbon Cathodes Enhanced by Metal Adsorption to Improve Hydrogen Production in Microbial Electrolysis Cells. Environ Sci Technol 52 (12), 7131–7137.
Le, T.X.H., Haflich, H., Shah, A.D., Chaplin, B.P, 2019. Energy-Efficient Electrochemical Oxidation of Perfluoroalkyl Substances Using a Ti4O7 Reactive Electrochemical Membrane Anode. Environmental Science & Technology Letters 6 (8), 504–510.
Leal, L.G., 1992. Laminar flow and convective transport processes. Elsevier. Lee, Y., Um, I.H., Yoon, J., 2003. Arsenic(III) oxidation by iron(VI) (ferrate) and subsequent removal of arsenic(V) by iron(III) coagulation. Environ Sci Technol 37 (24), 5750–5756.
Leupin, O.X., Hug, S.J, 2005. Oxidation and removal of arsenic (III) from aerated groundwater by filtration through sand and zero-valent iron. Water Res 39 (9), 1729–1740.
Liu, Y., Li, Q., Gao, S., Shang, J.K, 2011. Exceptional As (III) sorption capacity by highly porous magnesium oxide nanoflakes made from hydrothermal synthesis. Journal of the American Ceramic Society 94 (1), 217–223.
Londesborough, S., Mattusch, J., Wennrich, R., 1999. Separation of organic and inorganic arsenic species by HPLC-ICP-MS. Fresenius’ journal of analytical chemistry 363 (5-6), 577–581.
Ma, S., Poon, S., Mulchandani, A., Jassby, D., 2020. The Evolution of Metal Size and Partitioning Throughout the Wastewater Treatment Train. Journal of Hazardous Materials 402, 123761.
Manna, B.R., Dey, S., Debnath, S., Ghosh, U.C, 2003. Removal of arsenic from groundwater using crystalline hydrous ferric oxide (CHFO). Water Quality Research Journal 38 (1), 193–210.
McNeill, L.S., Edwards, M., 1997. Arsenic removal during precipitative softening. Journal of Environmental Engineering 123 (5), 453–460. Ning, R.Y., 2002. Arsenic removal by reverse osmosis. Desalination 143 (3), 237–241.
Pierozynski, B., Jankowski, J., Sokolski, W., 2009. Application of nickel-coated carbon f ibre material in cathodic protection of underground-buried steel structures. Corrosion Science 51 (11), 2605–2609.
Pincus, L., Rudel, H., Petrovic, P., Gupta, S., Westerhoff, P., Muhich, C., Zimmerman, J.B, 2020. Exploring the Mechanisms of Selectivity for Environmentally Significant Oxo- anion Removal During Water Treatment: a Review of Common Competing Oxo- anions and Tools for Quantifying Selective Adsorption. Environ. Sci. Technol 54 (16), 9769–9790.
Qian, A., Yuan, S., Zhang, P., Tong, M., 2015. A new mechanism in electrochemical process for arsenic oxidation: production of H2O2 from anodic O2 reduction on the cathode under automatically developed alkaline conditions. Environ. Sci. Technol 49 (9), 5689–5696.
Ronen, A., Duan, W., Wheeldon, I., Walker, S., Jassby, D., 2015. Microbial Attachment Inhibition through Low-Voltage Electrochemical Reactions on Electrically Conducting Membranes. Environ. Sci. Technol 49 (21), 12741–12750.
Ronen, A., Walker, S.L., Jassby, D., 2016. Electroconductive and electroresponsive membranes for water treatment. Reviews in Chemical Engineering 32 (5), 533–551.
Ryu, J., Monllor-Satoca, D., Kim, D.-h., Yeo, J., Choi, W., 2013. Photooxidation of arsenite under 254 nm irradiation with a quantum yield higher than unity. Environmental science & technology 47 (16), 9381–9387.
Saalfield, S.L., Bostick, B.C, 2009. Changes in iron, sulfur, and arsenic speciation associated with bacterial sulfate reduction in ferrihydrite-rich systems. Environmental Science & Technology 43 (23), 8787–8793.
Sadiq, M., Zaidi, T.H., Mian, A.A, 1983. Environmental behavior of arsenic in soils. Theoretical. Water, Air, & Soil Pollution 20 (4), 369–377.
Saha, S., Reza, A.S., Roy, M.K, 2019. Hydrochemical evaluation of groundwater quality of the Tista floodplain, Rangpur, Bangladesh. Applied Water Science 9 (8), 1–12.
Seidel, A., Waypa, J.J., Elimelech, M., 2001. Role of charge (Donnan) exclusion in removal of arsenic from water by a negatively charged porous nanofiltration membrane. Environmental Engineering Science 18 (2), 105–113.
Shan, Y., Mehta, P., Perera, D. and Varela, Y. (2019) Cost and Efficiency of Arsenic Removal from Groundwater: A Review, United Nations University-Institute for Water, Environment and Health, 11-13.
Sharma, V.K., Sohn, M., 2009. Aquatic arsenic: toxicity, speciation, transformations, and remediation. Environ Int 35 (4), 743–759.
Singh, R., Singh, S., Parihar, P., Singh, V.P., Prasad, S.M, 2015. Arsenic contamination, consequences and remediation techniques: A review. Ecotoxicology and Environmental Safety 112, 247–270.
Smith, R., Knight, R., Fendorf, S., 2018. Overpumping leads to California groundwater arsenic threat. Nature communications 9 (1), 1–6.
Sorg, T.J., Chen, A.S., Wang, L., 2014. Arsenic species in drinking water wells in the USA with high arsenic concentrations. Water Res 48, 156–169.
Tang, L., Iddya, A., Zhu, X., Dudchenko, A.V., Duan, W., Turchi, C., Vanneste, J., Cath, T. Y., Jassby, D., 2017. Enhanced flux and electrochemical cleaning of silicate scaling on carbon nanotube-coated membrane distillation membranes treating geothermal brines. ACS applied materials & interfaces 9 (44), 38594–38605.
Tresintsi, S., Simeonidis, K., Katsikini, M., Paloura, E.C., Bantsis, G., Mitrakas, M., 2014. A novel approach for arsenic adsorbents regeneration using MgO. J Hazard Mater 265, 217–225.
Vrijenhoek, E.M., Waypa, J.J, 2000. Arsenic removal from drinking water by a “loose” nanofiltration membrane. Desalination 130 (3), 265–277.
Walker, M., Seiler, R.L., Meinert, M., 2008. Effectiveness of household reverse-osmosis systems in a Western US region with high arsenic in groundwater. Science of the Total Environment 389 (2-3), 245–252.
Wang, L., Chen, A.S., Sorg, T.J., Supply, W., 2011. Costs of arsenic removal technologies for small water systems: US EPA arsenic removal technology demonstration program, 92. United States Environmental Protection Agency, Cincinnati, pp. 40–74.
Wilbraham, R.J., Boxall, C., Taylor, R.J, 2018. Photocatalytically driven dissolution of macroscopic nickel surfaces. Corrosion Science 131, 137–146.
Yan, Q., Wang, Z., Zhang, J., Peng, H., Chen, X., Hou, H., Liu, C., 2012. Nickel hydroxide modified silicon nanowires electrode for hydrogen peroxide sensor applications. Electrochimica Acta 61, 148–153.
Yoon, J., Amy, G., Chung, J., Sohn, J., Yoon, Y., 2009. Removal of toxic ions (chromate, arsenate, and perchlorate) using reverse osmosis, nanofiltration, and ultrafiltration membranes. Chemosphere 77 (2), 228–235.
Zhu, X., Jassby, D., 2019. Electroactive Membranes for Water Treatment: Enhanced Treatment Functionalities, Energy Considerations, and Future Challenges. Accounts of chemical research 52 (5), 1177–1186.
