Ion-Exchange Materials for Membrane Capacitive Deionization

The authors acknowledge the UK’s Engineering and Physical Sciences Research Council (EPSRC) under Grant EP/L01548X/1 for funding R.M.’s doctoral studies through the University of Manchester’s Graphene NOWNANO CDT account. The research reported in this publication was supported by funding from KAUST.


INTRODUCTION
The availability of potable water is an ever-expanding global challenge, with factors such as population growth and climate change driving the depletion of freshwater sources to unprecedented levels. 1 Consequently, the field of salt water desalination has received enormous research attention, in attempts to convert the abundance (>70% of Earth's surface) of natural saline waters into freshwater. Reverse osmosis is generally considered as the benchmark technology in terms of salt rejection and energy efficiency. 2 However, recent droughts and the subsequent rationing of water supplies in large coastal (or near coastal) cities such as Cape Town and Saõ Paulo suggest that desalination is not always an affordable means for large-scale freshwater production. 3 Capacitive deionization (CDI) is an emerging technique for the removal of solvated ions from aqueous solutions, gaining increasing application across fields such as desalination, water softening, wastewater treatment, and removal of heavy metal ions. 4 Conventional "flow-between" CDI removes ions from a feedwater stream by application of a small voltage (<1. 2 V) between two oppositely charged porous electrodes. Ions are stored capacitively in electrical double layers (EDLs) within the electrode pores and are flushed out upon zeroing or reversing the voltage, regenerating the electrodes. Remediation by CDI provides some key advantages compared to existing water treatment processes. The charge storage mechanism described is analogous to a supercapacitor, enabling energy storage capability while simultaneously achieving desalination during CDI. 5 Another crucial benefit of CDI is the ability to operate under ambient conditions (low pressures and room temperatures) compared to more aggressive desalination technologies such as distillation and reverse osmosis. 6 Thus, CDI has the potential to have a low energy input. CDI removes the minority component, the solute (ions), from the solvent (water); hence, desalination of brackish feed waters (low salinity) can be carried out with a high energy efficiency. 7 Electrochemical demineralization was pioneered by Blair and Murphy as early as 1960. 8 However, capacitive deionization was not formally presented until the report of Farmer et al. in 1996; they utilized carbon aerogel electrodes to remove NaCl and NaNO 3 from aqueous solutions. 9 Since then, especially in the past decade, a plethora of works have been published, generally focusing on novel electrode materials for improving and optimizing CDI performance. These include alternative carbon-based materials such as activated carbon, 10 carbon nanotubes (CNTs) 11 or nanofibers (CNFs), 12 graphene, 13 and carbide-derived carbon. 14 More recently, faradaic electrode materials have gained increasing interest, whereby a greater number of ions can be stored pseudocapacitively or by intercalation within electrodes. The disulfides of molybdenum (MoS 2 ) 15 and titanium (TiS 2 ), 16 sodium manganese oxide, 17 MXenes, 18 ferric phosphate (FePO 4 ), 19 and silver/silver chloride (Ag/AgCl) 20 have all been utilized as CDI electrodes in recent years. Modifications in CDI cell geometry and operational modes have followed these material developments, leading to developments such as flow-through CDI, 21 hybrid CDI, 17 inverted CDI (i-CDI), 22 the desalination battery, 23 and flow-electrode CDI (FCDI). 24 While excellent desalination properties have been reported for these materials and cell architectures, they are not without drawbacks. Traditional CDI with carbonaceous materials often suffers from an energy efficiency substantially lower than anticipated because of the phenomenon of co-ion expulsion. This occurs as ions of the same polarity are continuously adsorbed and desorbed from the electrode pores, reducing the amount of counterions that can be removed from the feedwater. 25 Exposure to excessive voltage can also cause parasitic reactions (e.g., anode oxidation) in the electrodes, leading to degradation and depleted performance over long-term operation. 26 Meanwhile, faradaic processes are frequently reliant on expensive and naturally non-abundant electrode materials as well as potentially irreversible redox processes occurring at the electrode surface.
To mitigate these factors, membrane capacitive deionization (MCDI) as a modification of traditional CDI is gaining increased attention over the past years. In MCDI, the presence of an anion-exchange membrane (AEM) over the anode and a cation-exchange membrane (CEM) over the cathode acts to block the transport of cations and anions, respectively. This can drastically reduce aforementioned co-ion effects and increase the rate of salt removal, besides adding a protective layer to protect against damaging faradaic reactions at the electrode surface. 27,28 MCDI was introduced by Lee et al., who desalinated thermal power plant wastewater by ion-exchange membranes (IEMs) over activated carbon cloth electrodes. 29 Following this, there have been a plethora of advancements in membranes, ion-exchange coatings, and theoretical studies over the past 15 years (Figure 1). In the context of the wider desalination field, MCDI (0.1−0.2 kWh m −3 ) has the potential to be more energy efficient than reverse osmosis (0.8−1.5 kWh m −3 ) for brackish waters with a salinity of <2 g L −1 . 7 Compared to CDI and FCDI, only MCDI is capable of operating at thermodynamic efficiencies comparable to that of reverse osmosis. 30 Ion-exchange materials and membranes are now an established and integral part of separation processes, playing a crucial role in a variety of applications such as desalination, wastewater treatment, food processing, and fuel cells. 31 The continued enhancement and engineering of ion-exchange materials are equally important to developments in system engineering, such is the reliance of current processes on the performance of ion-exchange materials of which they are comprised. This review aims to describe the principles of the MCDI technology and outline typical fabrication methods of selective IEMs for MCDI. A complete summary of developments of ion-exchange materials for MCDI is provided, including modified electrodes, polymeric membranes, and nanomaterial-incorporated ion-exchange layers. Additionally, we aim to critically compare the different ion-exchange materials for MCDI, suggesting directions that the everexpanding field should consider to achieve a greater industrial  8,9,29,31−44 ACS ES&T Water pubs.acs.org/estwater Review realization of the technology. While the focus of the review is MCDI, the use of these novel materials can be transferred to many other ion-exchange processes and will only serve to benefit various existing applications.

MEMBRANE CAPACITIVE DEIONIZATION (MCDI)
VERSUS CAPACITIVE DEIONIZATION (CDI) As mentioned above, CDI passes a feed salt water stream through a spacer channel, adjacent to two porous electrodes of opposite polarity. During the adsorption step, a cell voltage is applied and ions are stored in EDLs in porous electrodes. As the electrodes become charged, ions are stored electrostatically at the surface to maintain electroneutrality at the electrode/ solution interface. At a given volume of ions adsorbed, the electrodes reach a "saturation" point, signifying that the charge storage capacity of the electrode has been achieved. Regeneration of electrodes subsequently occurs upon zeroing or reversing the voltage, releasing the ions out into a brine stream. The capability of electrode regeneration in CDI makes it an attractive option for desalination due to low maintenance and repeatable performance over long-term desalination. 25 Alternatively, modification of CDI can be achieved by insertion of ion-exchange membranes over one (asymmetric MCDI) or both (symmetric MCDI) ( Figure 2) electrodes to enhance the properties of the system. IEMs typically comprise polymeric materials, which possess a high density of fixed charge carriers due to covalently bonded groups in the backbone. Such charged species can include quaternary ammonium cations (NH 3 + ) in AEMs or sulfonate (SO 3 − ) and phosphate (PO 3 − ) groups in CEMs. These charged species either occur naturally in the polymers or can be grafted onto the membrane via chemical reactions. 47 The presence of charged groups in the IEMs results in selective transport of ions of opposite charge (counterions) and blocks the transport of ions of the same charge (co-ions).
The enhanced performance of MCDI is mainly attributed to the reduction of the described co-ion effects. During the adsorption step in CDI, counterions are adsorbed in the micropores (pores with widths of <2 nm) while co-ions are desorbed simultaneously. The expulsion of co-ions back into the feed stream reduces the amount of salt that can be removed, as for each electron transferred between the electrodes <1 equiv of salt is removed. This will drastically decrease the ratio of salt adsorbed per unit charge, or charge efficiency (Λ) of the system as well as the amount of salt adsorbed per cycle. During adsorption in MCDI, co-ions are expelled from the micropores as counterions are adsorbed; however, the presence of a selective ion-exchange membrane prevents co-ions from exiting into the spacer channel. Instead, co-ions build up in the macropore (pores with widths of >50 μm) region of the electrodes. As co-ions accumulate in the macropores, eventually the concentration of ions in the macropores will be greater than the concentration in the spacer channel. To compensate for the excess charge buildup (electroneutrality) in the macropores, more counterions are transported across the membrane into the electrode region. 25 This means that more salt can be removed in each subsequent cycle in MCDI relative to CDI. The joint contribution of both the macropores and micropores to charge storage in MCDI rationalizes the improvement of performance compared to that of CDI. In CDI, the macropores are unable to contribute to charge storage; the concentration of co-ions in the macropores is consistently lower than that in the spacer channel during adsorption, rising to equilibrium when the electrodes become saturated with ions. 48 Consequently, the net salt removal and energy efficiency of MCDI processes can outperform CDI without ion-exchange membranes.
While the use of MCDI allows for a higher rate of salt removal and energy efficiency compared to those of CDI, another practical phenomenon that can be mitigated by employing MCDI is the degradation reactions of the capacitive carbon electrodes. The incorporation of IEMs can provide a selective barrier to protect electrode materials from unwanted faradaic reactions as a result of contact with saline water, which are often a byproduct of MCDI desalination. While it is accepted that certain faradaic processes can be beneficial to MCDI performance (e.g., faradaic and intercalation ion storage), other processes such as anodic oxidation and oxygen reduction reactions can result in water quality fluctuations and electrical energy losses. 49 In particular, anodic oxidation of the carbon electrode itself can lead to a breakdown of the internal pore structure and rapid deterioration of desalination performance. Such operational limitations are detrimental to the widespread industrial realization of (M)CDI; while less energy input would be required, compared to processes such as RO and distillation, these savings would be offset by the need for replacement materials and regular system maintenance. Consequently, the need for robust, high-performance, and affordable ion-exchange materials is crucial to ensure the longevity of electrodes for any industrial applications of MCDI.

MEMBRANE PROPERTIES AND FABRICATION METHODS
3.1. Membrane Properties and Characterization. The combination of structural and electrochemical properties of IEMs will have a discernible effect on the performance of the MCDI system. The following membrane parameters are most commonly calculated and stated for novel ion-exchange materials: ion-exchange capacity (IEC), area resistance (R A ), permselectivity, water uptake (WU), and linear swelling ratio (LSR).
IEC is a measure of the number of functional groups present in IEMs that can transport counterions. For MCDI, membranes will ideally possess a high IEC, maximizing the number of ions that can be transported across the membrane and stored in porous electrodes. IEC is usually determined by an acid−base titration method such as Mohr's method: 50 where V ab and c ab are the volume and concentration of acid or base used in the titration and m d is the dry mass of the membrane. The ability of a membrane to transport ions is another crucial electrochemical property. When a current is transported in MCDI, low membrane resistance is beneficial to reduce electrical losses, which will increase the charge/current efficiency. These resistance values are typically calculated by analysis of electrochemical impedance spectroscopy (EIS) of the IEM in a supporting electrolyte (e.g., 1 M NaCl). 51 Resistance values are calculated from an equivalent circuit model of the system and are normalized to the exposed area (A) to give the area resistance (R A ): where R MS is the resistance of the membrane immersed in the electrolyte solution and R S is the resistance of the electrolyte solution.
The permselectivity of an IEM describes the ability of the polymer to transfer current by transport of counterions only. A perfectly permselective IEM (P = 1) will completely block the transport of co-ions through the membrane. If co-ions are adsorbed, the permselectivity value will fall below the ideal value of 1. In MCDI, high-permselectivity membranes are desirable to prevent the expulsion of co-ions back into the spacer channel. This will increase salt adsorption in subsequent cycles to maintain electroneutrality and positively affect charge/current efficiency, as ions are retained in electrode pores. 48 Permselectivity can be calculated by techniques such as chronopotentiometry; the voltage response of the membrane/solution system to an imposed current is measured, giving information about preferential ion transport through the membrane. 52 Water uptake describes the amount of water absorbed by an IEM when it is in its hydrated form. The parameter is calculated as a percentage mass change when the membrane is between its dry and hydrated forms: . Different fabrication methods, which have been adopted to produce ion-exchange membranes for MCDI: (a) blending ion-exchange materials (pre-electrode fabrication), (b) coating ion-exchange materials (post-electrode fabrication), (c) solution casting followed by phase inversion, and (d) pore filling (CH 3 I, iodomethane; DIPEA, N,N-diisopropylethylamine). This figure was created by Heno Hwang, scientific illustrator at KAUST.
where m d and m h are the masses of the dry and hydrated membrane pieces, respectively. 53 Water uptake of IEMs should be sufficiently high to initially allow the uptake of ions by the membrane, however not excessive to compromise membrane permselectivity. Commercially available IEMs are generally produced with WU values in the range of 15−30%. 54 Closely related to the water uptake of the IEMs is the linear swelling ratio (LSR). It is derived from the length difference as a percentage between the dry and hydrated membrane pieces: Preferably, IEMs for MCDI should have good dimensional stability and exhibit little swelling, allowing them to maintain their structure and morphology over long-term operation.
Mechanical strength of IEMs is less pertinent to highperformance MCDI operation; however, the membrane must still be sufficiently robust to withstand assembly into the MCDI cell and high water flow rates. Mechanical or tensile testing of membranes can be conducted by elongation until break in load cells with a given force. This produces stress/ strain curves from which parameters such as Young's modulus of the membrane can be calculated. 55 3.2. Routes to Membrane Fabrication. The preparation and subsequent application of ion-exchange materials for MCDI are typically carried out by four primary methods (  (Figure 3c) (The membranes are placed adjacent to the premade electrodes in the MCDI cell to provide a selective barrier.), and (4) pore filling by incorporating ionexchange polymers into porous membranes to provide ion transport channels (Figure 3d). While ion-exchange layers for MCDI have also been prepared via alternative methods such as dip-coating or deposition methods, these four routes are most commonly adopted to prepare ion-selective layers.
The types of IEMs prepared via the methods described above will vary structurally, mechanically, and electrochemically in nature, all of which will directly impact the desalination performance of the MCDI system. However, free-standing polymer films generally have a more robust structure with a higher density of fixed charge carriers throughout the polymer backbone. 56 3.2.1. Blending of Ion-Exchange Materials (pre-electrode fabrication). An alternative means of incorporating ionexchange layers into CDI electrodes is to mix ion-exchange materials (e.g., resins and polymers) directly into the electrode premixtures, or slurries ( Figure 3a). Typical slurries for carbonaceous electrodes in CDI consist of a mixture of carbon with a large surface area for electrosorption (e.g., activated carbon) and an electrically conductive additive (e.g., carbon black) mixed with a binder [e.g., poly(vinylidene fluoride) (PVDF)] that is predissolved in an amount of solvent [e.g., Nmethyl-pyrrolidone (NMP)]. An 8:1:1 electrosorptive material:conductive additive:binder ratio has been used for fabrication of MCDI electrodes. 57 The materials and ratios used are analogous to those used in supercapacitor and battery electrode fabrications, reflecting the very similar charge storage mechanisms at work in the devices. 58 Similar to casting of polymeric membranes, the electrode mixture is usually coated onto a current collector (e.g., graphite, Ti, and Al) before annealing at 80−100°C to evaporate the solvent and form the pore structure. Again, the choice of current collector is often inspired by the field of energy storage and tends to involve inexpensive and abundant materials. Addition of ion-exchange polymers into this mixture creates a composite electrode, lining the electrode pores with an ion-selective layer. 36 Resins and polymers can be maintained in the pores via hydrogen bonding interactions between functional groups on the polymers and activated carbon (e.g., -OH, -NH 2 ). Composite electrodes prepared by blending or coating ion-exchange materials benefit from a reduced electrical resistance, which will facilitate the transport of ions into the electrodes. This is due to ionexchange layers, which are typically thinner (5−10 μm) than free-standing membrane films (<50 μm) prepared by solution casting. However, the electrode may suffer from degradation over long-term cycling due to the weakly physisorbed polymer layers on the carbon surface. Alternatively, functionalization of electrodes can be achieved by covalently grafting ion-exchange moieties such as amine and sulfonate onto the electrode surface. 59 3.2.2. Coating of Ion-Exchange Materials (post-electrode fabrication. Selective ion-exchange layers can be applied directly onto an electrode surface after the electrode has been coated onto the current collector ( Figure 3b). Studies of polymer-modified electrodes date back to the 1980s. For example, Degrand and Miller dip-coated dopamine onto vitreous carbon electrodes for oxidation processes, 60 while White et al. studied charge transport mechanisms in Nafionmodified glassy carbon electrodes. 61 Such "modified electrodes" have since been exploited for a variety of applications, including sensors, 62 supercapacitors, 63 and batteries. 64 This method is beneficial for materials that are immiscible with certain electrode mixtures, or solid particles that are insoluble in solvents such as NMP. Although this method cannot line the micropore structure as comprehensively as a blending process, a thin ion-exchange layer (a few micrometers) with low electrical resistance can still be fabricated on the outermost electrode surface. Polymer dope solutions have been prepared using the method previously outlined and coated onto activated carbon electrodes followed by heat treatment to evaporate the solvent and form the dense polymeric layer. The polymeric layers can be physisorbed onto the carbon surface, or cross-linking agents can be introduced to chemisorb the polymers. Kim et al. utilized such an approach, attaching a layer of poly(vinyl alcohol) (PVA) onto activated carbon electrodes using sulfosuccinic acid (SSA) as a cross-linking agent. 33 Uniform and careful application of the casting solution may provide good contact adhesion with the electrode surface, which is again beneficial for reduced resistance at the electrode/membrane interface. However, blade coating directly onto the surface may damage the preformed pore structure, which in turn reduces the surface area of electrodes onto which salt can be adsorbed. As an alternative to the polymeric materials approach, modification of electrodes postfabrication has proved to be an effective method for the addition of layers of metal oxides and nanomaterials. TiO 2 has been deposited by an atomic layer deposition (ALD) method, 65 and function-ACS ES&T Water pubs.acs.org/estwater Review alized graphene has been layered onto carbon fiber electrodes by a dip-coating technique. 38 These innovative electrode coating techniques have permitted a multitude of new materials to be incorporated into ion-exchange materials for MCDI, and the field continues to expand.

Solution Casting and Phase
Inversion of Polymeric Ion-Exchange Films. Solution casting and subsequent phase inversion of polymer films has been a commonly used fabrication technique for providing ion-exchange membranes for electro-membrane processes. 66 Phase inversion is a demixing method whereby a polymer in a solvent undergoes a controlled transition from a liquid state to a solid state (Figure 3c). This is done first by dissolution of a polymer in solvent, forming a dope solution. Polar aprotic solvents such as NMP, N,N-dimethylacetamide (DMAc), and N,N-dimethylformamide (DMF) can dissolve most polymers typically employed in IEMs. The homogeneity and morphology of the final membrane are highly reliant on the preparation of the dope solution. This involves steps such as overhead stirring, rolling, and degassing of the solution to remove air bubbles. The preparation of composite membranes of more than two components requires more attention to ensure the formation of homogeneous, high-quality films.
The dope solution is spread across a support using a casting knife with a specified air gap. The solvent is then evaporated either in air or in an oven, before undergoing phase inversion to produce the membrane film. The heat treatment step evaporates away a top layer of solvent to promote the formation of a dense and nonporous film, which is essential for selective ion transport and negligible water flux through IEMs. The formation of a low-porosity membrane can also be promoted by an increased polymer concentration in dope solutions. Highly porous membranes for ultrafiltration (UF) and nanofiltration (NF) are prepared from dope solutions with an initial polymer concentration in the range of 12−20 wt %, whereas dense IEMs require high polymer contents of >20 wt %. 67 Phase inversion occurs by immersion of the membrane and support into a coagulation bath, containing a large amount of nonsolvent such as water. This causes an exchange of solvent and nonsolvent; the solvent diffuses into the coagulation bath as the nonsolvent diffuses into the membrane film. At a given time, the system becomes thermodynamically unstable and demixing occurs, causing the membrane film to precipitate. Parameters such as the coagulation bath temperature, volume of nonsolvent, choice of solvent/nonsolvent mixture, and film precipitation time must all be carefully controlled for specific membrane morphologies. 68 The design and morphology of free-standing ion-exchange membranes can also be tailored depending on the materials and functionality. IEMs can be classified as homogeneous or heterogeneous in nature. Homogeneous IEMs consist of ionexchange groups chemically bonded to the polymer backbone, providing a distribution of charge carriers with a high degree of uniformity. The ion-exchange functionality can be imparted by using naturally occurring ion-exchange monomers or polymers, chemical modification of polymers in the casting solution (premembrane casting), or chemical modification of the membrane (postfabrication). Conversely, heterogeneous membranes are fabricated by blending ion-exchange domains (e.g., resin particles) into a nonfunctional polymer backbone. The ability to tailor the ion-exchange functionality of polymer membranes allows them to achieve an IEC that is higher than those of composite electrodes, permitting more ions to be transferred into the porous electrodes in MCDI. Furthermore, nanocomposite IEMs containing nanomaterials such as graphene oxide (GO), CNTs, and metal oxides embedded in a polymer matrix have been prepared. 69 The incorporation of nanomaterials can positively affect membrane properties such as mechanical strength and permselectivity, due to the presence of interfacial hydrogen bonds and selective ion transport channels between nanosheets. However, care must be taken to ensure dispersion of nanomaterials within the solvent prior to casting, resulting in a uniform film with few defects. 70 The preparation procedure and type of IEM utilized can differ in morphological and electrochemical properties, all of which will influence the performance of an MCDI system. However, the versatility and ability to tailor the properties of polymeric membranes make them an attractive choice for MCDI. 71 3.2.4. Pore Filling. Pore-filled IEMs are an innovative approach for preparing membranes with limited swelling, while maintaining good electrochemical properties. The membranes rely on the use of a mechanically robust porous substrate (e.g., polyethylene) 72 that acts as a support for the introduction of ion-exchange groups. The IEMs are prepared by adding ionexchange polymer electrolytes or by introducing monomers that undergo polymerization inside the pores (Figure 3d). Pore-filled IEMs can also be produced on a large scale by a roll-to-roll method, which is beneficial for the potential scaleup of MCDI. In a typical roll-to-roll pore-filling method, the "impregnation" of the porous support occurs by soaking of the porous substrate in the electrolyte in the presence of a crosslinking agent. Polymerization is then achieved by methods such as chemical polymerization or photopolymerization to fill the pores with the ion exchanger. To prevent leaching of the ion-exchange material, the substrate is often laminated prior to polymerization and delaminated when the reaction is complete. 73 This creates a dense and nonporous structure similar to IEMs prepared by phase inversion, with selective ion transport pathways throughout the membrane. 74 Pore-filled IEMs can achieve a water uptake (WU) and a linear swelling ratio (LSR) lower than those of membranes prepared via solution casting, while maintaining a similar IEC. 75 This is due to the mechanical restriction of ion-exchange groups within nanometer-sized pores.

MCDI OPERATIONAL MODES AND PERFORMANCE
METRICS The operational modes used in MCDI will affect which parameters will be determined experimentally. In this section, the different electrical (constant voltage vs constant current) modes and flowing (single-pass vs batch) modes will be defined, followed by the corresponding experimental parameters that are conventionally measured for each.
4.1. Constant-Voltage versus Constant-Current MCDI. Constant-voltage (CV) and constant-current (CC) modes of operation can be applied separately, or in conjunction with each other, depending on the MCDI system setup and the desired output. Constant voltage is the most commonly used mode in MCDI. This applies a fixed voltage throughout all adsorption and desorption steps, whereas the current density varies. In CV, the effluent concentration decreases slowly during adsorption and increases during desorption. As the operating voltage increases, the final concentration will decrease during the adsorption step. However, at voltages in large excess of 1.23 V, parasitic ACS ES&T Water pubs.acs.org/estwater Review reactions at the electrode surface can cause electrical energy losses. This arises due to the voltage surpassing the thermodynamic stability limit of water, after which electrolysis of water occurs to produce molecular oxygen and hydrogen. This can lead to further reactions at the electrode surface, leading to a reduced efficiency of salt removal and deposition of scale on electrodes. 76 Conversely, CC operates under a defined current density where the voltage fluctuates. Here, in principle, the effluent concentration remains at constant values for adsorption and desorption, depending on the predefined current density. Due to this, CC operation is better suited to applications in which a given final effluent concentration is required. However, the fluctuating voltage in CC must be closely monitored to ensure that cell voltages do not exceed 1.23 V. CV should be used for maximum ion removal processes, as it can deliver a higher salt adsorption rate than under the same operating conditions as the CC mode. 77 4.2. Single-Pass versus Batch-Mode MCDI. Measuring desalination capabilities in MCDI requires the monitoring of the change in ion concentration over a given time. This is typically done by monitoring the solution conductivity. Singlepass (SP) and batch-mode (BM) operation differ in the position of the conductivity probe.
In the SP method, water is fed from the reservoir through the MCDI cell and the conductivity is measured immediately upon exiting the cell. For SP measurements, the conductivity (salt concentration) initially decreases to a minimum after applying a voltage. The conductivity then increases back to the influent concentration due to the electrodes reaching saturation, expelling ions from the electrodes. A requirement for SP experiments is a large volume of feed solution. This ensures that there are no large fluctuations in salt concentration in the influent stream.
BM experiments measure the conductivity in the feed reservoir, where the volume can be much smaller than that in SP mode. In BM measurements, the feed reservoir volume must be small; otherwise, conductivity changes are too small to be measured accurately. BM experiments show a steady decrease in conductivity over time. The conductivity will stabilize at a minimum value when the electrode becomes saturated (maximum salt adsorption). 25 The difference between initial and final conductivity values can be used to calculate the amount of ions removed. BM offers the advantages of simpler operation and analysis of results, whereas SP is more representative of industrial MCDI as the feed is not continuously recycled and the effluent is collected. 78 4.3. MCDI Performance Metrics. The desalination performance of the MCDI system is determined from a variety of metrics, which vary according to the type of MCDI configuration in use and operational modes. Salt adsorption capacity (SAC) is an important parameter, representing the amount of salt removed per gram of active electrode (m e ) during adsorption: where C i and C f are the initial and final concentrations, respectively, and V s is the volume of the feed solution (liters). Ideally, SAC values should always be quoted at dynamic equilibrium (when the salt adsorption in one half of the cycle is equal to the salt desorption in the subsequent half-cycle). This is termed the maximum salt adsorption capacity (mSAC), which is the sorption capacity of an electrode fully saturated with ions. SAC is a relative value, recorded as the amount of salt adsorbed per gram of electrode. This allows it to be applied to all MCDI systems, regardless of size. This is therefore a commonly calculated metric for laboratory-scale setups, which generally use lower electrode masses. It is also widely calculated in batch-mode CDI studies, where the initial and final salt concentrations and the volume of the feed solution are easily measured. 79 Charge efficiency (Λ) is another key metric providing insight into the energy efficiency of the MCDI system. Generally, higher values of Λ signify a system with a lower energy consumption. It is defined as the ratio of salt adsorbed to charge (Q) applied to the system during the adsorption step, given by where F is the Faraday constant and M is the molar mass of the salt molecule. The charge supplied can be calculated from the time integral of the current during the adsorption step. As the charge efficiency states the amount of electric charge required to remove a number of ions, it is an indicator of total energy consumption and the occurrence of any parasitic reactions at the electrode. The charge efficiency is the metric usually used to characterize CV MCDI operation. A closely related parameter is current efficiency λ. This is used in place of charge efficiency in CC MCDI operation; this is a steady-state process in which the current and effluent salt concentration are constant over time. Due to this, it is also commonly quoted in flow-electrode (FCDI) studies.
In single-pass experiments, the effluent is collected and the conductivity measured at the exit of the MCDI cell. This is more representative of industrial MCDI operation, during which multiple stacks of parallel electrodes can be utilized. This can significantly increase the amount of salt that can be adsorbed over the duration of the experiment. In such cases, it is common to quote the salt removal efficiency (SRE) of the system: This is perhaps the most practical performance indicator for industrial translation of MCDI systems. In batch-mode operation, the conductivity is measured in the feed reservoir, meaning SRE values are low and are often not a useful performance metric. Kim et al. assembled a pilot-scale MCDI setup comprising 50 pairs of anion-and cation-selective electrodes that operated over a period of 15 days, achieving a high and consistent nitrate (NO 3 − ) removal efficiency of >90%. 80 Finally, the average salt adsorption rate (ASAR) relates the ratio of salt adsorbed to the adsorption time.
ASAR is influenced by parameters such as cell configuration, electrode features, flow rates, and feed salt concentration. Optimization of these parameters can achieve the highest salt adsorption rate of the MCDI system. 81 ACS ES&T Water pubs.acs.org/estwater Review

COMMERCIAL ION-EXCHANGE MEMBRANES FOR MCDI
Many MCDI studies have sought to utilize commercially available ion-exchange membranes, which had previously been employed for other electro-membrane processes such as electrodialysis (ED). 54 In addition to improving desalination performance, various works have used IEMs for fundamental studies, to better understand and optimize system parameters for different operational modes of MCDI. 82 Tables 1 and 2 display common commercial anion-and cation-exchange membranes studied for MCDI, respectively. 87,88 The most studied IEMs with respect to MCDI applications are Neosepta (ASTOM Co.), Selemion (Asahi Glass Co. Ltd.), and Fumasep (FuMa-Tech GmbH). Commercial (homogeneous) IEMs are typically manufactured by a paste method; this contains a monomer, a cross-linking agent, and a polymerization initiator. These polymers can then undergo a subsequent functionalization step to impart anion-or cationexchange character. The use of multiple reaction steps from monomer to final membrane structure can cause the production of commercial IEMs to be a costly process, which can somewhat inhibit their translation into industrial MCDI. Neosepta (CMX) and Selemion (CMV) CEMs are based on a polystyrene and divinylbenzene (PS/DVB) backbone that then undergoes sulfonation to impart cationexchange character. Neosepta AMX membranes also possess a PS/DVB structure that is subsequently aminated to provide anion-exchange character, while the Selemion AMV membrane is based on a polystyrene structure cross-linked with butadiene. 31 Nafion CEMs are instead prepared from perfluorinated sulfonic acid polymers and have been employed widely as the proton-exchange material in fuel cells. 99 Many commercial membranes must also be reinforced with a stable material or supporting fabric to provide dimensional stability, further increasing the material and production cost. 100 The fabrication steps, choice of reagents, and sustainability of ionexchange materials must all be carefully considered when determining the commercial viability of MCDI. In addition, supply chain cost is another aspect of MCDI that must be addressed before its widespread commercialization. For the commercialization of desalination units such as MCDI and RO, a reliable and affordable supply chain is essential should replacement parts be required. Fumasep IEMs are manufactured in Germany, while Selemion and Neosepta IEMs are both produced in Japan only. The limited global production of high-performance IEMs could lead to excessive material demand and high shipping costs. The increase in reliable production of IEMs from a larger number of companies around the world will certainly lower the cost and accelerate the growth of industrial and commercial MCDI products. This will be encouraged by the continued research and development of novel ion-exchange materials.
While the utilization of commercial IEMs provides a convenient route to ion-selective salt removal by MCDI, certain limitations of the materials have led to the development of IEMs tailored specifically for MCDI application. Commercial (homogeneous) IEMs are typically manufactured by a paste method; this contains a monomer (with groups susceptible to functionalization), a cross-linking agent, and a polymerization initiator. This is then coated onto a supporting fabric followed by functionalization. 66 Commercially available membranes are typically thicker (>100 μm) than membranes required for MCDI. This is due to their primary use in ED, where membranes must be robust and self-supporting. In MCDI, a thin selective barrier is preferable, whereby ions can be transferred a short distance across the membrane/electrode   28 In addition to this, efforts have been made to fabricate membranes with electrochemical properties suited to MCDI. Ion-exchange capacity (IEC) represents the amount of functional groups in the membrane layer that can take up corresponding counterions, usually determined by an acid− base titration approach such as Mohr's method. 101 A higher IEC of a membrane can cause a larger net removal of salt from the feed stream due to the enhanced functionality of the membrane material. Consequently, IEMs have been fabricated for MCDI with IEC values exceeding those of commercially available IEMs. 75 The area resistance (R A ) is another important membrane parameter that relates to ion conductivity across a given membrane area into the electrode pores. It is most commonly determined by analysis of Nyquist plots obtained by electrochemical impedance spectroscopy (EIS). 102 Membranes with area resistance that is lower than those of their commercial counterparts have therefore been fabricated for MCDI, reducing interfacial resistance and increasing the energy efficiency of the system. 103 Several other parameters of ion-exchange layers such as morphological, water uptake (WU), hydrophilicity, and permselectivity (selectivity to transfer specific ions) have been varied to optimize the performance of an MCDI module. 45

DEVELOPMENTS IN ION-EXCHANGE MATERIALS FOR MCDI
The subsequent sections will summarize separately and compare different ion-exchange materials that have been developed for MCDI since its conception. The chemical structures of the anion-and cation-exchange materials are depicted in Figures 4 and 5, respectively. The development of novel free-standing polymer membranes will be described, which comprises the largest percentage of MCDI studies. Ionexchange materials deposited directly onto electrodes are separated into two sections: electrodes modified by blending ion-exchange materials into electrode slurries (prefabrication) and by coating ion-exchange materials directly onto the electrode surface after electrode coating (postfabrication). The incorporation of nanomaterials such as graphene into innovative IEMs for MCDI is also discussed.
6.1. Free-Standing Ion-Exchange Membranes. Freestanding ion-exchange membranes offer myriad benefits to their application in MCDI. Polymer IEMs can be prepared by the solution casting of large area membrane films, meaning membranes can feasibly be scaled up to industrially sized modules. Self-supporting IEMs also offer stability over longterm cycling due to covalently bound functional groups fixed in the polymer matrix. Commercial CapDI (Voltea) MCDI modules have been fabricated for desalination of brackish and ground waters. 104 These systems contain multiple stacks of parallel electrode/membrane assemblies, which can be chemically cleaned for reuse. This demonstrates the practicality and the potential long-term operation of membranes in MCDI.  Early studies of ion-exchange materials for MCDI include cation-exchange membranes that were fabricated from copolymers based on sodium methacrylate (NaSS), methacrylic acid (MAA), and methyl methacrylate (MMA) (NaSS-MAA-MMA). 34 A corresponding anion-exchange membrane was also fabricated using the copolymer vinylbenzyl chlorideco-ethyl methacrylate-co-styrene (VBC-EMA-St). 105 Likewise, sulfonated CEMs were prepared from a variety of copolymers using concentrated sulfuric acid, followed by solution casting and phase inversion. 106 These studies were able to tailor material properties such as IEC, area resistance, and uptake of water by MCDI application. However, despite showing desirable membrane properties, the desalination performance was not fully tested using the MCDI performance metrics outlined previously. MCDI performance was instead quantified in terms of charge−discharge current, indicating the formation of electrical double layers at the electrode surface that would occur in (M)CDI. This may have been due to a lack of established MCDI performance metrics at the time that these works were conducted (Tables 3 and 4).
6.1.2. Anion-Exchange Membranes (AEMs). This section will outline, and compare where possible, the development of free-standing AEMs for the purpose of MCDI. A popular route by which anions are transferred through an AEM is via quaternary ammonium groups. These act as fixed cationic groups for the transport of anions through the membrane. A study that developed, characterized, and tested such a membrane in an MCDI setup was carried out by Tian et al. 107 The study prepared cross-linked and quaternized PVA membranes; the abundance of quaternary ammonium groups in the QPVA matrix imparted the membranes with a high IEC of ≤2.82 mol kg −1 . It was found that increasing the degree of cross-linking decreased the IEC and moisture content of the prepared membranes. The highest SAC achieved was 15.6 mg g −1 at an initial salt water concentration of 800 mg L −1 ; importantly, this was more than double the value of CDI using uncoated activated carbon electrodes. In addition, the measured SAC increased as the feed salt water concentration increased from 200 mg L −1 (3.4 mg g −1 ) to 800 mg L −1 (15.6 mg g −1 ). This was an interesting insight into how fundamental parameters (applied voltage and feed salt concentration) directly affected obtained values of SAC.
Jeong et al. employed a similar route for preparing AEMs based on copolymers of PVDF and vinylbenzyl chloride (PVDF-g-VBC), followed by amination with trimethylamine (TMA) and solution casting. 108 The AEMs were employed in a symmetric MCDI cell using a feed salt solution of 250 mg L −1 . The highest IEC of the membranes produced via this route was a modest 1.31 mequiv g −1 , which is lower than that of the commercial AMX membrane. However, the R A of membranes (2.0 Ω cm 2 ) was lower than the reported value of the AMX membrane. Desalination performance in this case was measured by salt removal efficiency. The study stated a maximum removal rate of 79%, 51% higher than during CDI operation. However, limited MCDI performance data make it more difficult to compare to similar literature.
An example of a pore-filled AEM designed for MCDI was prepared by ul Haq et al., based on the introduction of an ionexchange monomer solution into a microporous polyethylene (PE) supporting membrane. 75 The PE membrane was immersed in a monomer solution of chloromethylstyrene (CMS) to block the pores, followed by radical polymerization at 80°C. A dense film layer with a suppressed pore structure was successfully produced; the IEMs exhibited very low water uptake (5%) and LSR (2%) due to the restriction of the polymer in the membrane pores. Despite this, the membrane retained excellent IEC (3.0 mequiv g −1 ) and low area resistance (0.32 Ω cm 2 ) due to the abundance of charge carriers in the membrane pores. The pore-filled AEM was employed alongside the commercially available CMX membrane using a 10 mM NaCl solution, achieving superior SAC (16.1 mg g −1 ) and charge efficiency (98.3%) compared to MCDI using AMX/CMX membranes (SAC, 14.5 mg g −1 ; charge efficiency, 94.8%). This study showed that AEMs with enhanced dimensional and electrochemical properties could be fabricated via innovative methods, which translated into improved MCDI performance.
The group of Chang et al. sought to delve into the effect of AEM properties on MCDI performance. AEMs were fabricated on the basis of poly(phenylene oxide) with various degrees of quaternization and employed in an asymmetric, batch-mode MCDI cell, covering only the anode. 109 The membranes with different degrees of quaternization exhibited varying IEC, WU, and area resistance properties. The optimum a-MCDI performance was obtained using a membrane with a degree  of quaternization of 60%, with an IEC of 1.9 mmol g −1 , an area resistance of 2.6 Ω cm 2 , and a water uptake of 22.5%. This produced a maximum SAC of 7.4 mg g −1 and a charge efficiency of 55%, with devices also maintaining stability over 50 cycles. This study concluded that high IEC, low resistance, and low water uptake were pertinent to high-performance MCDI. However, permselectivity values were not determined for the AEMs, which are important to consider to limit co-ion expulsion. Importantly, it highlighted the importance of the AEM in an MCDI module; SAC values obtained using solely the lab-synthesized AEM (7.4 mg g −1 ) were comparable to that of a symmetric MCDI that also included a commercial CMX membrane (7.2 mg g −1 ). It was hypothesized that the inclusion of an AEM was more pertinent than the CEM, due to the reduction of parasitic oxidation reactions occurring at the anode. The protection of the anode with an AEM negated the anode from contacting the bulk solution, while simultaneously transferring anions, giving excellent salt adsorption capacity and cycling stability. Although the charge efficiency is lower using AEM only (due to co-ion effects at the cathode), it gives the possibility of maintaining excellent desalination performance while reducing the operational cost by omission of a CEM.

Cation-Exchange Membranes (CEMs).
Research has been carried out to develop similar novel CEMs with improved properties compared to those of commercial membranes. Qiu et al. prepared a CEM (PE-CSPS) by filling porous polyethylene membranes with a cross-linked sulfonated polystyrene to provide cation-exchange character. 72 The CEM was employed both in single-pass asymmetric MCDI (PE-CSPS only) and symmetric MCDI (PE-CSPS/AMX) and compared to CDI with no membranes. Again, the symmetric MCDI cell (22 mg g −1 ) exhibited a 2-fold increase in SAC compared to CDI (10 mg g −1 ) for an initial salt concentration of 500 mg L −1 . The SRE and current efficiency also reached 42% and 88%, respectively, for the symmetric cell. The asymmetric cell also showed a slight improvement (13 mg g −1 ) compared to CDI and to MCDI using commercial CMX membrane (12 mg g −1 ). The improved desalination characteristics were attributed to the lower area resistance of the membranes (0.33 Ω cm 2 ) compared to the CMX membrane. These studies showed that improved properties in novel membranes can directly improve the MCDI performance relative to commercially available IEMs. Kang et al. prepared CEMs based on sulfonate-grafted PVDF (PVDF-g-PSVBS), adopting their previous method to prepare AEMs from various copolymers. 110 In this case, the reported maximum salt removal rate (SRE) was 41.6% (compared to 79% for the aminated PVDF-g-VBC AEM). This was due to the excessive water uptake (+60%) and the relatively low IEC of the membrane (1.14 mequiv g −1 ), which would sacrifice the selectivity properties of the membrane. This highlights the potential detrimental effects to MCDI performance of enhanced membrane hydrophilicity. To improve these properties, the same group prepared new membranes based on the modification of polyketone films with ion exchanger NaSS (PKs-g-NaSS). 111 The NaSS was grafted onto the polyketone backbone via γ-irradiation. The membranes gave a comparable IEC (1.10 mequiv g −1 ) but a lower WU (34.6%) compared to those of previously prepared PVDF-g-PSVBS CEMs, resulting in an increased SRE of 87.6% at the same feed salt concentration (250 mg L −1 ) and operating conditions (1.5 V adsorption/−1.5 V desorption). Another approach employed by the same group was the fabrication of heterogeneous CEMs based on blending commercial cation-exchange resins within an ethylene vinyl acetate (EVA) supporting matrix. 112 The study reported favorable mechanical properties and comparable resistance to commercial IEMs. However, a similar problem was encountered for the previously prepared PVDFg-PSVBS CEM: the high mass loading (70%) of hydrophilic resin particles caused a very high water uptake (≤59%) of the heterogeneous membrane. This meant a maximum salt removal rate of 56.4% was achieved when employing the membrane in the same MCDI configuration under identical conditions. These studies outline a range of useful fabrication routes and material processing aspects for IEM production. While the highest-performance CEM was produced via radiation grafting, a blending procedure with resin particles is a more economically viable option for large-scale production. All of these factors must be considered during MCDI process scale-up.
Following their preparation of a pore-filled polyethylene AEM, the group of Cha et al. also fabricated CEMs based on sulfonated poly(ether ether ketone) (sPEEK). 41 The polymers were dissolved in DMF (10 wt %) followed by membrane casting and solvent evaporation at 70°C for 24 h. The sPEEK-80 CEM (80% degree of sulfonation) was characterized and employed alongside an AMX membrane in an MCDI cell. The electrochemical properties were slightly inferior to the porefilled AEM; the highest-performance membrane possessed an IEC of 2.0 mequiv g −1 , a WU of 28%, and an R A of 0.85 Ω cm 2 . The lower IEC and the increased swelling of this membrane arose due to the lack of restriction of ion-exchange groups. Nevertheless, the sPEEK-80 membrane gave an increase in SAC of 14% compared to the use of the commercial CMX membrane.
Another innovative solution to CEMs was implemented by Jain et al.; membranes based on sulfonated pentablock copolymers (sPBC) were created with varying degrees of ion-exchange capacity (1.0, 1.5, and 2.0 mequiv g −1 ) by variation of the casting solvent composition (10−60 wt %). 45 It was found that the membrane microstructure was influenced by the polarity of the casting solvent. As expected, the SAC and charge efficiency of the MCDI devices increased when using membranes with a high IEC and a low area resistance. Contrary to previous reports, 109 the study found that a high water uptake is beneficial to the CEM to minimize ionic resistance, as long as a high density of fixed charge carriers (IEC) is maintained in the membrane backbone. Although it was shown that highly hydrophilic CEMs can improve MCDI performance, the contradictory findings in reports to date highlight the need for further fundamental studies into membranes for MCDI. This can give definitive evidence into how membrane structural and electrochemical properties impact desalination performance.
Polymeric materials have attracted plentiful research attention as ion-exchange membranes for MCDI, because of their versatility, tunable properties, and ease of processing. Ceramics (e.g., fluorite and perovskite) make up another family of materials that have shown promise for electrochemical separations in batteries and supercapacitors. 113 An initial use of ceramic materials in electrochemical water desalination was the use of a sodium superionic conductor (NASICON, Na 3 Zr 2 Si 2 PO 12 ) as a solid electrolyte in a "seawater battery". In this case, NASICON acted as a solid electrolyte between two carbon electrodes immersed in ACS ES&T Water pubs.acs.org/estwater Review seawater and an organic electrolyte. The seawater battery can be viewed as a faradaic analogue of MCDI, desalinating by reversible redox reactions at electrode surfaces, in the presence of an ion-selective separator. The use of redox processes enables the desalination of seawater concentrations of NaCl in the presence of a redox electrolyte, given that a highly permselective separator can be found to screen the redox ions from the seawater source.
Since the inception of the seawater battery, various innovative designs that have successfully utilized ceramic materials as cation-exchange membranes have been studied. Lee et al. developed a NASICON membrane using a solid-state reaction process, for electrochemical desalination in the presence of a separate channel containing a redox electrolyte (NaI). 114 This unique cell design permitted the simultaneous desalination of NaCl and reduction of iodide ions at separate electrodes in the seawater battery, increasing the desalination capacity of the system. The NASICON cation-exchange membranes were shown to combine high Na-ion conductivity and high permselectivity, to mitigate iodide crossover into the desalination compartment. The use of the NASICON membrane and the contribution of the redox electrolyte (NaI) enabled a high desalination capacity of 87.2 mg g −1 , even for solutions at seawater concentrations of NaCl (600 mM). An alternative seawater battery configuration was later developed by Kim et al., consisting of two separate system compartments whereby salination and desalination occur during charging and discharging, respectively. 115 Again, the system performance was heavily influenced by the implementation of a NASICON membrane. The membrane separated the sodium metal anode from the seawater compartment, while enabling continuous Na + ion transport between the seawater and the metal anode. The compartmentalized system was able to deliver 84% total dissolved solid (TDS) removal from seawater, owing largely to the incorporation of a highly permselective and ion conductive NASICON separator. These studies highlighted the unique properties of ceramic membranes and that their use in seawater batteries can permit desalination of salt water sources beyond those of conventional MCDI.
6.1.4. Membranes for Removal of Alternative Salts (other than NaCl). A notable advantage of MCDI is that by utilization of an electric field for purification, ion removal is not limited solely to Na + and Cl − , but to any charged, solvated species in solution. Siekierka et al. extended the focus of MCDI toward desalination of salt solutions other than sodium chloride (NaCl), such as lithium chloride (LiCl) and potassium chloride (KCl). 116 This work prepared AEMs to remove Cl − from three monovalent salt solutions, based on the modification of polyvinyl chloride (PVC) films by ethylenediamine in combination with a lithium-selective sorbent. This configuration varied the salt removal of the systems depending on the cation, returning SAC values of 34.2, 8.6, and 9.8 mg g −1 for LiCl, KCl, and NaCl, respectively. The high SAC obtained for LiCl removal showed that the use of a selective sorbent in combination with an anion-exchange membrane can be effective for the targeted removal of certain species. Aiming to further improve the desalination performance, the same group fabricated a novel AEM for lithium extraction via hybrid capacitive deionization (HCDI). 117 The MCDI cell consisted of an AC electrode covered by an AEM [PVDF film functionalized by ethylenediamine (EDA)], combined with a lithium−manganese−titanium oxide (LMTO) cathode, similar to those used in batteries and supercapacitors. The use of the lithium-selective LMTO cathode aimed to increase the SAC in HCDI, without the use of a conventional CEM over the cathode. It was determined that the system was able to reach an SAC of >30 mg g −1 and a current efficiency of ∼90%, although the SAC value was seen to decay over three repeat cycles. These studies demonstrated that MCDI could be tailored to selectively remove specific solvated ions.
Expanding on earlier studies of pore-filled membranes, Kim et al. prepared pore-filled CEMs for removal of multivalent salts (e.g., MgCl 2 ). The CEMs were prepared by thermal polymerization of either styrene or glycidyl methacrylate (GMA) monomers in a porous substrate, cross-linking agents (e.g., DVB), followed by a sulfonation reaction to impart cation-exchange properties. 44 The highest-performance porefilled membrane gave a salt removal efficiency of >80% for both sodium and magnesium ions, representing a 16% increase compared to that of the commercial CMX membrane. This was attributed to the low resistance (1.14 Ω cm 2 ) and high selectivity (selectivity coefficient of 3.2 of Mg 2+ /Na + ) to multivalent cations. In addition, the CEMs performed well in a large-scale MCDI setup consisting of 25 stacks of electrodes. This study demonstrated the versatility of both pore-filled CEMs and MCDI for the desalination of multivalent salt solutions, which would be the case for MCDI on an industrial scale.
6.2. Ion-Exchange Coatings for MCDI (pre-fabrication of electrodes). Free-standing polymeric membranes have been successfully fabricated and outlined, providing versatility (removal of various salts) and tunable electrochemical properties. However, ion-exchange layers can also be included by modification of carbon particles in the electrode slurries. For the sake of clarity, this section will outline all ion-exchange materials introduced prior to electrode casting (prefabrication), which are blended with carbon electrode slurries. This effectively forms a composite electrode after coating; as a result, ion-exchange material properties such as IEC, area resistance, and permselectivity can rarely be determined as in the free-standing polymeric membranes.
This approach was carried out in 2012 by Nie et al. 35 The group used an innovative electrophoretic deposition method, depositing a slurry of CNTs and poly(acrylic acid) (PAA) onto a graphite current collector by application of a DC voltage. The PAA acted as a cation-exchange layer in the composite electrode, and the performance was compared to that of pure CNT and a commercial CEM. The highest salt removal efficiency of 83% was observed using the CNT-PAA composite, achieving an increase of 12% compared to that of the CNT-CEM configuration. The CNT-PAA composite also showed no detectable decline in performance over 30 repeated charge−discharge cycles. The success of the composite was attributed to the low interfacial resistance and high cation permselectivity of the PAA layer; however, as previously mentioned, these values were not calculated explicitly in this study. Nevertheless, the study was a pioneering work for the production of ion-selective layers by electrochemical modification of electrode mixtures.
In 2014, Liu et al. expanded on this study by using ionexchange polymers to cover both the anode and the cathode. Polyethylenimine (PEI, cation-exchange polymer) and dimethyl diallyl ammonium chloride (DMDAAC, anion-exchange polymer) were mixed with CNT electrode mixtures and coated onto graphite substrates, to produce an anode and cathode, respectively. 37 An optimum mass of 150 mg of ion-exchange polymer in the CNT slurries was found to produce the maximum desalination efficiency; above this mass, the polymers acted to block the electrode pores and reduce the salt storage capacity. By employing both a covered anode and a cathode, a desalination efficiency of 93% was achieved. This represented an increase of 19% compared to that of a cell incorporating commercial anion-and cation-exchange membranes and 9% relative to that of the CNT-PAA composite prepared in their previous work. A highest charge efficiency of 70% was also achieved, indicating that the ion-exchange polymers act as a permselective membrane to minimize co-ion expulsion. These works serve as an excellent demonstration of the value of a low-resistance and permselective IEM by simply incorporating functional materials into electrode slurries. Nafion is a commercially available, sulfonated fluoropolymer, which has been utilized frequently as a cation-exchange membrane for fuel cell applications. 118 Cai et al. prepared a Nafion cation-exchange layer on cathodes for asymmetric MCDI, by mixing the Nafion solution into the activated carbon slurry. 119 Due to the use of asymmetric MCDI with an uncovered anode, any improvement in performance could be attributed solely to the inclusion of the Nafion layer. The highest-performing Nafion-AC MCDI system produced an SAC of 10.8 mg g −1 and a charge efficiency of 45%. This was a considerable improvement relative to using uncovered electrodes, which gave an SAC and a charge efficiency of 6.9 mg g −1 and 24%, respectively. The Nafion-AC composite was also shown to be stable over 100 electrochemical cycles; however, the cycling stability of the MCDI system was tested over only four cycles. This study was further confirmation that desalination improvement can be achieved by a simple blending procedure of activated carbon with an ion-exchange polymer.
This method was adapted later by Fritz et al., who coated similar polyelectrolyte layers by mixing with activated carbon for inverted membrane capacitive deionization (i-MCDI). 120 In i-MCDI, additional surface charges on electrodes are utilized to store ions passively during adsorption. This gives the technique potential to reduce energy consumption, as adsorption occurs at 0 V and desorption at a negative voltage (active phase). The ion-exchange materials utilized in this study were PSS (cation exchange) and PDMDAAC (anion exchange). Through utilization of this method, SAC values of 5.2 mg g −1 were achieved alongside a low energy input and process exergy loss, comparable to that of traditional MCDI. This was attributed to the ion-selective surface charges imparted to the electrodes by the polyelectrolyte layers; the polyelectrolyte-based IEMs were shown to uniquely modify MCDI systems.
Evans et al. aimed to further improve the electrochemical characteristics of a carbon electrode by introducing conductive polymer polyaniline (PANI). 121 The polymerization of aniline was carried out in situ alongside the carbon slurry, producing a functional PANI coating on the porous carbon electrode. The study found that the presence of the PANI layer increased the specific capacitance compared to that of the unmodified carbon electrode. High SAC values were obtained for salt solutions (1500−1700 ppm) of LiCl (12.7 mg g −1 ), NaCl (14.1 mg g −1 ), KCl (18.9 mg g −1 ), MgCl 2 (13.3 mg g −1 ), and CaCl 2 (20.8 mg g −1 ). Promisingly, more than 90% of the charging capacity of the composite electrode was retained after 300 cycles, indicating that the PANI layer remained affixed in the electrode pores. This gave insight into how ion-exchange layers can maintain high desalination performance over longterm MCDI operation.
6.3. Ion-Exchange Coatings for MCDI (post-fabrication of electrodes). An alternative route to achieve composite electrodes is by coating of the ion-exchange material directly onto the preformed electrode. Thus, the electrode and the ion-exchange material are coated in separate fabrication steps. Utilization of this approach ensures that a robust electrode structure is fabricated, prior to coating with an ion-exchange material. An early study in this field coated electrodes by packing the MCDI flow chamber with ionexchange resin granules. 36 This configuration, named R-MCDI (92%), improved the salt removal efficiency by >30% relative to MCDI using commercial IEMs (60%). This was attributed to a reduction in ohmic resistance from R-MCDI (3.9 Ω) compared to that of MCDI (9.6 Ω), as determined by electrochemical impedance spectroscopy. This was an important study that characterized the electrical resistance of the coated layer; the lower resistance and better contact adhesion of the resin layer were shown to promote ion transport into the electrodes and improve salt removal efficiency.
Kim et al. formed composite electrodes by coating aminated polysulfone (APSf) and sulfonated PVA (sPVA) polymers directly onto the anode and cathode, respectively. 122 The polymers were applied to the electrodes with a casting knife and, after solvent evaporation, had a layer thickness on a few micrometers. The desalination performance of the coated electrodes was tested in salt solutions of NaCl, CaSO 4 , and MgCl 2 , as well as a mixture of the salts. The salt removal efficiency was >90% for all of the salt solutions for initial concentrations of 100 mg L −1 , and SEM images showed good anchoring of the polymers with no sign of delamination from the electrode. This study showed that ion-exchange layers coated directly onto electrodes are a viable alternative to freestanding polymeric IEMs for MCDI; in addition, the ability to prepare very thin layers (<10 μm) make them an attractive option to reduce electrical losses and hence increase the charge efficiency of the MCDI system.
Water hardness is another pressing environmental issue that can have various detrimental effects. The presence of Ca 2+ and Mg 2+ , arising due to the contact of water with various minerals, is damaging to various industrial and domestic appliances due to the buildup of scale. Removing hardness has therefore become a pertinent issue to which MCDI can contribute. The group of Yoon et al. prepared alternative ion-exchange layers to remove hardness from water (Ca 2+ ). This was done by casting a sodium alginate solution onto AC electrodes, followed by ion exchange in a 2 M CaCl 2 solution to form a Ca alginate gel-like layer. 40 The thickness of the Ca alginate layer was comparable to those of commercial IEMs when fully swollen (140−160 μm); however, the layer dramatically increased the hydrophilicity at the electrode surface (contact angle of water in air of 27°). The lowest area resistance of the Ca alginate membrane (0.63 Ω cm 2 ) was far lower than that of commercial CMX membrane (6.27 Ω cm 2 ), resulting in a high charge efficiency (95%) and SAC (15.6 mg g −1 ) for removal of 10 mM CaCl 2 when employed with AMX in the MCDI system. This demonstrated that effective, ion-selective membranes could be fabricated by direct coating onto carbon electrodes, for removal of various ions beyond the traditional NaCl solutions.
ACS ES&T Water pubs.acs.org/estwater Review This was reinforced by Zuo et al., who targeted the selective removal of sulfate ions from water by coating a slurry of quaternized PVA (QPVA)/sulfate-selective ion-exchange resin onto the anode. 123 In mixtures of NaCl and Na 2 SO 4 , the AEM resin layer was shown to preferentially transport SO 4 2− over Cl − (selectivity coefficient of 2.57) into the anode, because of the greater effect of the applied electric field on the divalent SO 4 2− ion. Furthermore, the MCDI system with QPVA/resin AEM displayed excellent stability over 50 cycles, suggesting it could be a feasible industrial technology for removal of SO 4 2− from wastewaters.
6.4. Nanomaterial-Incorporated Ion-Exchange Layers. Nanomaterials (NMs) have recently emerged as candidates to augment the properties of IEMs for electromembrane processes. The incorporation of NMs can affect all of the mechanical, structural, wetting, and electrochemical properties of the ion-exchange layer. 69 For MCDI specifically, nanomaterials have been employed in stand-alone layers and nanocomposite membranes and electrodes. NMs were incorporated into MCDI architectures by Lee and Choi, by decorating carbon electrodes with sulfonated polystyrene and TiO 2 nanoparticles. 124 An optimum TiO 2 particle concentration of 10 wt % was found to dramatically reduce the resistance of the ion-exchange layer, however reducing selectivity due to the formation of pores. Overall, the use of a TiO 2 -incorporated composite improved the salt removal efficiency of MCDI by 30% compared to that of the uncoated electrode, indicating the improvement that NMs could have on desalination performance.
The group of Qian et al. utilized an innovative approach to assemble an ultrathin layer of sulfonated graphene oxide (sGO) as a stand-alone CEM for MCDI. Sulfonic groups were grafted onto graphene oxide nanosheets followed by "dipcoating" onto carbon nanofiber (CNF) electrodes by immersion into the sGO water solution. 38 The hydrophilic sGO sheets with their abundance of negatively charged sulfonic acid groups demonstrated good self-assembling capacity, and EDX maps showed a uniform distribution of functional groups over the electrode surface. The sGO/CNF composite showed a 40% increase in capacitance compared to that of CNF electrodes, which arose due to the thin (<5 μm) and low-resistance sGO layer that facilitated the transport of sodium ions into the electrodes. An asymmetric system with an uncovered anode achieved a maximum SAC of 9.5 mg g −1 and a charge efficiency of 43%, compared to an SAC of 5.01 mg g −1 and a charge efficiency of 21% using pristine CNF electrodes. The system also displayed excellent stability over 60 cycles, indicating the positive influence that nanomaterials can have when employed as a stand-alone membrane in MCDI.
An alternative means of incorporation of NM into MCDI membranes is by combination with a polymer in a nanocomposite membrane. Zhang et al. prepared a ternary composite AEM comprising reduced graphene oxide (rGO) functionalized with polyaniline (PANI), within a PVDF supporting matrix. 39 The inclusion of the rGO/PANI composite had the desired effect of reducing the membrane resistance and increasing the salt removal efficiency, because of the conductive nature of the rGO/PANI filler. The study also found that the IEC of the membranes could be optimized depending on rGO/PANI mass loading. It was stated that the filler material added more sites for ion exchange in the PVDF matrix. Due to these varied properties, the salt removal efficiency of MCDI could also be altered according to the membrane composition. A removal efficiency of >90% and an electrosorption capacity of 1.56 mg g −1 were achieved using the highest-performance AEM. This study further highlighted that NM incorporation can influence the electrochemical properties and subsequent desalination performance of ionexchange membranes for MCDI.
Metal oxides such as TiO 2 , which can be prepared in nanoparticulate form, are another low-cost and well-studied nanomaterial that can tune the properties of composite membranes and/or electrodes. A recent innovative technique sought to produce a coated electrode by means of atomic layer deposition (ALD) of TiO 2 on multiwalled CNT membranes and/or electrodes. 65 This work utilized an approach of "pulsing" TiCl 4 and H 2 O precursors in an ALD reactor. This sophisticated technique allowed control of the mass loading of TiO 2 according to the number of pulse cycles. Optimum performance was observed for the composite with a 20 wt % mass loading of TiO 2 , and this resulted in an increase in hydrophilicity, evidenced by a decrease in the water contact angle (69°) relative to that of the pristine CNT (114°). The increased hydrophilicity allowed the aqueous salt solution to access a greater pore volume in the electrode. The excellent wettability, in combination with good capacitive behavior, was The structure and name of the abbreviated material are outlined previously in Figure 4. Unless stated otherwise, a commercial CEM was used to cover the cathode. Scalability (1, no potential for scalability; 5, high potential for scalability), ease of preparation (1, high difficulty of preparation; 5, low difficulty of preparation), and sustainability of materials (1, low sustainability; 5, high sustainability) have been rated on a scale of 1−5.
ACS ES&T Water pubs.acs.org/estwater Review ascribed to the higher SAC of the composite (3.33 mg g −1 ) compared to that of the unmodified CNT electrode (1.35 mg g −1 ) at 1.2 V. This work demonstrated the versatility of methods by which NMs can be integrated as an ion-exchange layer; however, the cost and environmental considerations (e.g., nanomaterial leaching off electrode surface) of such a deposition method would likely hinder the scaling of the technology.

COMPARISONS OF ION-EXCHANGE MATERIALS
Regardless of the ion-exchange material and preparation method, there are important characteristics that must be considered when assessing the application of ion-exchange materials for MCDI. We have selected six crucial factors that we believe will greatly impact the large-scale realization of MCDI as a means of water purification. In addition to the factor of the increase in the performance of MCDI relative to CDI (R-factor), the stability, cost, scalability, preparation, and sustainability of the ion-exchange material must all be carefully considered when assessing the quality of the membrane layer. These characteristics for anion-exchange materials are summarized in Table 5.
The wide varieties of feed salt concentrations, operational modes, and performance metrics used in MCDI render crossstudy comparison notoriously difficult. However, upon calculation of a performance increase factor for each study (R-factor), this allows the reader to isolate the effect of the IEM on the desalination performance. The R-factor was defined as the ratio of the increase in performance (in terms of salt removal efficiency or salt adsorption capacity) of the MCDI system studied, relative to the CDI results with no ionexchange material. This parameter was briefly introduced in a study of asymmetric MCDI using AEMs only; 109 however, the use of this factor has not become widespread in the field of MCDI. A high R-factor is desirable, indicating that the inclusion of the ion-exchange material results in a larger amount of salt being removed from the feed stream. The highest R-factor of 3.7 was calculated for both DMDAAC (with PEI on cathode) and rGO/PANI/PVDF anion-exchange materials. For the DMDAAC/PEI configuration, this also benefited from a simple blend coating of the ion-exchange polymer onto the carbon electrode and a low material cost. While these studies showed a high R-factor, the stability of the MCDI systems was not sufficiently explored. No repeat cycles were displayed; hence, information about the long-term performance of the membrane could not be determined. The stability of the membrane over repeat cycles is imperative for use in a commercial application, and conductivity profiles over repeated cycles should always be displayed in MCDI studies.
The anion-exchange membrane based on QPPO (Table 5 and Figure 4) was one that exhibited excellent stability over 50 repeated adsorption/desorption cycles (1 h duration). It also achieved an R-factor of 3.1 with an uncovered cathode; thus, all performance improvement could be ascribed to the AEM. Importantly, the material is scalable, as large area films of the membrane can be produced by solution casting methods. Alternative anion-exchange materials that were determined to have good potential for scalability were the pore-filled (PE-CMS) membranes and electrodes coated with ion-exchange polymers APSf (anode) and sPVA (cathode). Large area films of porous PE are commercially available and mean the IEMs can be produced on a large scale by scaling up the pore-filling polymerization reaction. Likewise, APSf and sPVA polymers can be blade coated onto a large area of the electrode and pilot-scale studies have already been successfully conducted on polymer-coated electrodes.
To determine the ease of preparation (material processing) of the ion-exchange material, factors such as the number of fabrication steps, reagents, and reaction times and conditions were all considered. The materials with the simplest preparation steps were as-purchased ion-exchange polymers (DMDAAC, PEI, and PSS) blended directly with carbon slurries and coated as composite electrodes. These were prepared by a one-pot coating technique with no additional modification steps required. Ion-exchange materials (APSf, sPVA, Ca alginate, and PAA) that were coated directly onto the electrode (postfabrication) were also rated as having a high ease of preparation. These were rated at a lower value due to the electrode and ion-exchange layer being prepared in separate fabrication steps. Materials with a low ease of preparation included the PVDF-g-VBC copolymer and the rGO/PANI/PVDF composite anion-exchange membranes. This arises because of the preparation of 2+ reagents using multiple steps, over a long duration. Rigorous preparation steps increase the cost and hinder the feasibility of the industrial application of the membranes for MCDI.
A final and somewhat overlooked variable that we considered was the sustainability of the materials and processes used to fabricate ion-exchange materials. MCDI is rightly praised as a sustainable process, with energy requirements that are lower than those of competing desalination technologies such as distillation. However, more efforts must be made to The structure and name of the abbreviated material are outlined previously in Figure 5. Scalability (1, no potential for scalability; 5, high potential for scalability), ease of preparation (1, high difficulty of preparation; 5, low difficulty of preparation), and sustainability of materials (1, low sustainability; 5, high sustainability) have been rated on a scale of 1−5.
ACS ES&T Water pubs.acs.org/estwater Review ensure that material and process components of MCDI such as polymers, chemical modifiers, solvents, and electrodes are as sustainable as possible. For example, the replacement of a toxic NMP solvent in electrode slurries with an alternative green solvent could promote industries to invest in MCDI as a safe and environmentally friendly technology. Studies with highly rated sustainability included the PANI coating on activated carbon, which used carbon electrodes derived from automobile tire waste. The preparation of carbon/ion-exchange polymer (PEI and DMDAAC) mixtures was undertaken in aqueous solutions and therefore offered good sustainability, negating the need for damaging organic solvents. In addition to safety and environmental benefits, the preparation of electrode mixtures in water permits evaporation to occur at ambient temperatures (40°C) compared to organic solvents such as NMP (80−100°C) after electrode casting. These factors indicate that the blending of water-soluble ion-exchange polymers with carbon is a more sustainable fabrication route than solution casting of polymers (PVDF-g-VBC, QPPO, and APSf), which typically require polymer dissolution in organic solvents (e.g., DMF, DMAc, and NMP). In the next decade, the sustainability of the MCDI architectures and processes should be prioritized, if MCDI is to progress into an established desalination technology. Characteristics of fabricated cation-exchange materials are summarized in Table 6. Similar trends were observed to the AEMs, with highly scalable polymer materials (PE-CSPS, sPEEK, and PVA-SSA) produced via solution casting or pore filling of porous membranes. Further innovative fabrication methods were used to prepare CEMs, such as PVA coating onto electrodes using SSA as a functional cross-linking agent. This was an important study that demonstrated excellent potential for scalability and high ease of preparation. The CEM was applied by direct coating onto the carbon electrode and the use of cross-linker (SSA) meant that the ion-exchange layer was chemically affixed to the electrode. This use of cross-linker can ensure the chemical stability of the coated membrane, giving confidence that the membrane layer can remain intact when processing large volumes of water and over repeated desalination cycles.
The use of sGO as a cation-selective coating was a particular study that displayed great promise for the use of nanomaterials as ion-exchange materials for MCDI. The use of sGO membranes almost doubled the SAC compared to that of CDI, showed excellent stability (60 cycles), and could be assembled on the electrodes by a simple dip-coating technique. In terms of ease of preparation and scalability, a thin, standalone nanomaterial layer with abundant functional groups such as this may be preferable compared to nanomaterial/polymer composite (rGO/PANI/PVDF) membranes for MCDI purposes.
Two cation-exchange materials that displayed excellent sustainability aspects were the use of PAA and Ca alginate membranes for hardness removal. Both PAA and alginate layers are biodegradable, and all starting materials were prepared in aqueous solutions; this drastically reduces the energy consumption of the preparation process. In addition, the Ca alginate layer had by far the lowest material cost of all ion-exchange materials. The use of low-cost, functional and naturally derived ion-exchange materials may also shape the progression of MCDI technology in coming years.
A further interesting observation gathered by comparison of Tables 3 and 4 is the difference in R-factor between the anion-and cation-exchange materials. Where data were available from the literature, the average R-factor was higher for the anionexchange (2.8) than cation-exchange (1.8) materials. This indicated that the AEM has a more pronounced effect on the desalination performance than CEMs. This could be due to the aforementioned discussion of the role of AEM in limiting degrading oxidation reactions occurring at the anode, which would reduce the desalination performance over time. The discrepancy could also be rationalized by the typical intrinsic membrane properties. For example, the pore-filled AEM (PE-CMS) was able to achieve a superior IEC (3.0 mequiv g −1 ) compared to that of the corresponding pore-filled CEM (PE-CSPS; IEC = 1.0 mequiv g −1 ) for a similar area resistance. Such properties would increase the R-factor for AEMs as more salt is removed relative to CEMs. Previous studies have shown that membrane chemistry, thickness, and conductivity can all contribute to salt removal and energy expenditure in MCDI. 103 Further fundamental studies such as this one are required, to understand fully the separate contributions of the AEM and CEM to MCDI performance.

CONCLUSIONS, FUTURE OUTLOOK, AND NEXT
STEPS This final section will outline the key areas that we believe should be the focus of MCDI in the immediate future. This is based on the findings of the review, current trends within the field, and certain gaps in the literature that need to be addressed for MCDI to continue to expand as a means of water purification.
8.1. Industrial Potential of MCDI Ion-Exchange Materials. After evaluation of the literature, in terms of a large-scale application of MCDI, there are certain types of ionexchange materials and fabrication routes that appear to be favorable. Free-standing polymeric membranes, produced by solution casting/phase inversion or pore filling, have the ability to be scaled up to large area films. To process large volumes of saline water, electrode and membrane interfaces must be large to maximize the contact area for salt removal, and IEMs are an already established component of desalination techniques such as RO and ED. The fabrication methods are trusted and reproducible, and membranes can be removed, reused, modified, and cleaned if necessary, which is beneficial to long-term operation. In addition, the thickness of polymeric membranes tailored to MCDI has been reduced by ≤80% compared to commercially available IEMs, which is a technological improvement to reduce ionic resistance and energy consumption. Despite these benefits, there is an urgent need to reduce the cost of fabricating polymeric membranes. As one can see in Tables 3and 4, the material cost of freestanding polymer membranes (e.g., PVDF-g-VBC and PE-CMS) is among the highest of all ion-exchange materials, because of the use of multiple reagents and several fabrication and modification steps. The reliance of dope solution preparation in organic solvents is also an area that must be addressed moving forward. There are now a plethora of "green" solvents and mixtures that should be explored as alternatives during membrane fabrication. If these issues can be remedied, free-standing IEMs are very promising candidates for ion-exchange materials in industrial MCDI.
Composite electrodes have been discussed at length in this review. These are less established commercially but are an equally promising route to ion-exchange layers in MCDI. Ionexchange polymers (APSf and sPVA) coated directly onto the ACS ES&T Water pubs.acs.org/estwater Review electrode surface act as an effective IEM, and pilot-scale studies have proven that this can be done reliably and in large quantities. Coating onto the electrode surface also eliminates the need for phase inversion to precipitate the membrane and provides a better contact adhesion to the electrode surface than a free-standing membrane. This is beneficial for reducing resistive losses and uptake of water by the electrode. Considering scale-up, blade coating onto carbon electrodes must be done carefully, to avoid mechanical damage to the porous structure and breakdown of the MCDI device. Fouling studies must also be carried out on the ion-exchange layer, and appropriate cleaning procedures must be established so that polymer-coated electrodes can maintain long-term operation. Similarly, the blending of ion-exchange polymers into carbon slurries is a very attractive option for an ion-exchange layer in terms of ease of preparation and overall material cost. However, reports utilizing these methods have not sufficiently studied the performance over multiple hours and cycles. For the potential scale-up of the technology, it must be confirmed that the physisorbed polymers remain coated onto the electrode after multiple cycles; otherwise, the desalination performance of the system would rapidly diminish. The use of functional cross-linkers could be a way to improve ionexchange capacity and chemically anchor the polymers onto the carbon electrodes. As discussed, nanomaterial (sGO) coatings have also shown great potential as ion-exchange coatings, providing thin and highly selective layers that facilitate ion transport. However, as with blended ion-exchange polymers, further research on the stability of these layers over time must be conducted. Producing and coating these materials on a large scale and at a low cost should also be prioritized if they are to be used in an industrial setting. 8.2. Pilot-Scale Applications of MCDI. Despite MCDI attracting plentiful research interest for more than a decade, pilot-scale studies are still quite scarce. However, studies on a large scale are essential to ensure that lab-scale desalination performance can be maintained for larger systems that process larger and more concentrated volumes of saline water. Ultimately, these factors will determine the feasibility of using MCDI for purification on an industrial scale. A 2019 study has recently assembled a pilot-scale system to purify municipal wastewater via electrode/membrane composites. 80 The system consists of 50 parallel pairs of activated carbon electrodes (dimensions of 10 cm × 10 cm) that are oppositely coated with anion-and cation-exchange polymers (Siontech). The membrane layers were approximately 20 μm and showed a uniform distribution over the electrodes. Real wastewater effluent was pumped through the module; the study found that the system displayed good removal efficiency for various anions (NO 3 − , NO 2 − , SO 4 2− , and Cl − ) and cations (Na + , K + , Ca 2+ , Mg 2+ , and NH 4 + ). Reproducible performance was observed even after continuous operation for 15 days, which was partly credited to the smooth and homogeneous ionexchange layer that prevented the accumulation of fouling agents on the membrane. Although the study did not fully characterize the membrane layer, the work clearly showed a stable system with a high rate of contaminant removal. The same pilot-scale system was utilized in a separate study to remove bromide ions from recycled domestic wastewater. 46 The presence of bromide in water supplies can lead to the formation of bromate ions, which are highly regulated due to their toxicity. Promisingly, a final bromide concentration of <0.1 mg L −1 was achieved after MCDI cycling. This value was comparable to those achieved using brackish water reverse osmosis, while exhibiting a lower energy consumption. These studies highlighted the versatility of MCDI to remove various ions and that polymer-coated electrodes can purify real wastewater effluent on a close to industrial scale. Pilot-scale studies should be a key focus of the field in the coming years to confirm the practicality of MCDI on an industrial scale.
8.3. Continuous Operation of MCDI Using Flowing Carbon Electrodes. Of equal importance to the production of large-scale MCDI modules is the ability for systems to achieve a high desalination performance in a continuous mode of operation. The uninterrupted and stable operation over periods of hours and even days will be another deciding factor for the practical application of MCDI. A route that has been employed to achieve continuous desalination is the use of flowing suspension electrodes (FCDI) in combination with ion-exchange membrane separators. Gendel et al. modified the principle of MCDI by incorporating activated carbon flow electrodes adjacent to the feedwater channel. 125 Indeed, the selection of ion-exchange material is crucial to provide a selective barrier separating the flow channels and to mitigate cross-contamination. The study explicitly compared the performance of batch-mode (cyclic) and continuous operation. For batch-mode operation, an unprecedented SAC of 260 mg g −1 was achieved for a NaCl concentration of 15 g L −1 . In addition, the continuous MCDI configuration achieved a high desalination rate of 99% for feed NaCl solutions of 1 g L −1 . These results demonstrated that FCDI was a feasible option for continuous desalination of salt concentrations approaching that of seawater. The practical application of continuous FCDI was expanded upon by Porada et al., who utilized flowing carbon electrodes for water desalination as well as continuous energy harvesting. 126 Importantly, the study was able to investigate how characteristics such as flow rates and mass loading of carbon electrodes affected the salt removal and efficiency of the system. The utilization of flowing carbon electrodes alongside custom-made cylindrical ion-exchange membranes allowed for continuous energy generation via concentration gradients in water salinity and gas phase CO 2 . These innovative studies highlight the importance of continuous operation for industrial processes. The ongoing optimization of electrodes, membranes, and system design will help continuously operated FCDI to access further applications in the future.
8.4. Benchmarking MCDI Technology. While MCDI is not a mature technology, it is also not in its infancy, having been conceptualized in 2006 and the number of studies increasing every year since. Despite this focus, and the commercialization of some MCDI modules, MCDI is currently not in a position to compete with RO as a stand-alone desalination technique. The materials research in MCDI is now substantial, and efforts should be made in the next few years to explicitly compare MCDI with competing desalination techniques, in terms of salt removal and energy consumption. The energy efficiency of MCDI compared with other techniques has been a contentious issue, with disparate findings across studies. Patel et al. published an article stating that ED outperforms MCDI in terms of energy efficiency; however, this was based solely on a theoretical model. 127 Likewise, a theoretical study conducted by Qin et al. concluded that the energy consumption of CDI was far greater than RO for the same water recovery and salt rejection. 128 However, an opposing account was recently put forward by Porada et al., ACS ES&T Water pubs.acs.org/estwater Review whereby a joint experimental/theoretical study was conducted to compare MCDI and RO. 129 It was established that for a fair comparison between MCDI and competing technologies, consistent performance metrics must be defined across different technologies. The study found that by operating MCDI in an intermittent flow mode, a high water recovery and salt removal could be achieved with a lower energy consumption than RO for salinities of 2 g L −1 . Also, pilotscale studies using polymer-coated electrodes have found that MCDI can consume lower energies than second-pass brackish water RO. No matter the findings, studies such as these are essential in the next few years to determine the viability of MCDI compared to existing technologies. 8.5. Hybrid MCDI Technologies. As previously mentioned, stand-alone MCDI modules have been commercialized by companies such as Siontech 130 and Voltea, 131 with the ability to process varying volumes of saline water. However, the lack of a widespread commercialization of MCDI suggests that further research is required before it can be considered as an alternative to RO in the market. One route in which MCDI could access the current desalination market and beyond is by combination of MCDI with an existing process. Hybrid MCDI technologies have already shown great promise and can incorporate the unique advantages of MCDI into established techniques and devices. A recent study integrated MCDI into an RO system; after first-pass RO to desalinate the high-salinity feedwater, an MCDI/reverse electrodialysis (RED) system was installed as a replacement to second-pass brackish water reverse osmosis (low salinity). 132 The water processed by the hybrid system was found to meet World Health Organization water quality regulations, and the substitution of RO with the MCDI-RED for brackish water desalination reduced the energy consumption by almost 40%. This study helped to justify the labeling of MCDI as an energy efficient alternative to existing desalination techniques, especially for low-salinity regimes. Similarly, a combined MCDI/ion-exchange (MCDI-IE) system was also able to selectively recover nitrogen from wastewaters by preferential removal of multivalent cations (Ca 2+ and Mg 2+ ) over ammonium (NH 4 + ) during the MCDI pretreatment step. 133 Another important feature of MCDI that must be fully exploited is the low voltage required for salt removal. This low electrical energy input has the potential to be powered by renewable energy systems. Tan et al. assembled a 1 kW pilotscale plant that used photovoltaic cells and battery storage to harness the energy required to power an MCDI device. 42 The system was capable of electrode charging currents of 100 A, processing volumes of water of ≤5 m 3 /day, and could operate for 24 h without connection to the grid. Importantly, conditions of the hybrid system were optimized to increase the system performance time. This work has recently been expanded upon by Ramalingan et al.; this study assembled a novel, self-sustaining module whereby electrochemical desalination was powered by a visible light solar cell. 134 This unique design utilized separate iodide and ferricyanide redox reactions in the solar and desalination compartment, respectively. This self-sustaining device negates the need for electrical energy input as in traditional desalination, while permitting continuous desalination of the feedwater stream. The development of such "photodesalination" technologies could provide future clean water solutions to hot and arid regions in particular, where sunlight is plentiful and natural freshwater is scarce. MCDI has proven to be a somewhat enabling technology for electrochemical desalination via redox reactions. Chen at al. recently assembled a device to achieve desalination via viologen redox flow reactions. 135 Viologens make up a family of redox-active organic molecules that can readily form aqueous electrolytes, allowing for combination of an MCDI/ battery technology to continuously desalinate a flowing saline stream. The separation of flow channels by efficient IEMs contributed to the high performance of the system, achieving a salt removal efficiency of >95% for an initial salt concentration of 6000 ppm. The continued enhancement of materials, redox electrolytes, and system design for electrochemical desalination will see it play a prominent role in future applications, although the toxicity of such redox-active components needs to be considered before practical applications can come to fruition. The principle that MCDI can be integrated into and influence such a variety of systems, however, is extremely promising, and hybrid systems should continue to be explored to harness the full potential of MCDI across various fields.
8.6. Innovative Applications of MCDI. An alternative route for MCDI to access current markets is by application to purification beyond desalination of brackish waters of various salts. A particular remediation that MCDI could target is that of removal of heavy metal ions from waste waters. Elements such as arsenic (As), lead (Pb), chromium (Cr), and cadmium (Cd) can be toxic or carcinogenic in nature and contaminate water supplies both naturally and artificially. These ions are typically found at low concentrations, making them ideally suited for an electrosorptive removal process such as MCDI. 136 Indeed, such a remediation is preferable to existing methods such as chemical precipitation, which can produce large volumes of toxic sludge. Beyond heavy metal ions, potentially toxic species such as nitrite (NO 2 − ) should also be targeted. Another unique field in which MCDI has shown promise is the capture of greenhouse gases such as CO 2 . Legrand et al. utilized a solvent free approach to capture CO 2 via MCDI, by production of carbonate and bicarbonate ions after the reaction of CO 2 with water. 137 Encouragingly, this was achieved under ambient conditions without the use of additional chemicals, suggesting that MCDI could be a green and efficient route for the capture and reuse of harmful atmospheric gases. MCDI has already shown great promise for alternative sustainable processes, and finding further niche applications of the technology will serve only to increase the commercial potential of MCDI. 8.7. MCDI with Natural Saline Streams. An area that has plentiful room to be explored in the near future is the use of MCDI for purification of natural waters. A majority of MCDI studies currently use lab-prepared solutions consisting of a single salt solution. Studies that either utilize or replicate natural streams must be carried out. The presence of contaminants such as organic foulants could hinder the uptake of the desired salt into the membrane and electrode and decrease the desalination performance. 138 Studies such as these will confirm if any pretreatment or cleaning steps are required either before or during MCDI operation, to remove other potentially damaging solvated species.
8.8. Fouling Studies. As a natural continuation to using natural saline streams, fouling studies should also be carried out to assess any changes to the ion-exchange material and performance over long-term operation. Until recently, fouling and cleaning of MCDI systems has been a somewhat overlooked topic, but one that will have enormous implications on the adoption of MCDI for remediation purposes. Studies ACS ES&T Water pubs.acs.org/estwater Review that focus explicitly on membrane fouling will be fundamental for the maintenance and cleaning of any commercial MCDI devices. Hassanvand et al. studied the effect of foulants alginic acid and humic acid on the performance of both CDI and MCDI. 139 It was found that the incorporation of an ionexchange membrane mitigated the fouling effects on the carbon electrodes to a high degree, but that traditional alkaline solutions may be ineffective for cleaning due to the breakdown of PVDF binder in the carbon electrode. Chen et al. carried out another important study that investigated the long-term performance of MCDI in the presence of organic foulants. In addition to humic acid, sodium alginate foulant was also used as a representative of organic matter. 140 Over the course of 15 days of MCDI operation, it was found that the salt removal decreased by 5.3 and 3.3 mg and the energy consumption increased by 57% and 26% in the presence of humic acid and alginate, respectively. Significantly, both of these studies suggested that a pretreatment step may be necessary for the sustainable and optimal performance of MCDI. This issue was addressed in a subsequent study by Liang et al. via the fabrication of an integrated ultrafiltration-capacitive deionization (UCDI) device. 141 This innovative process removed humic and alginic acid via size exclusion/electrocatalytic oxidation, followed by desalination by MCDI. The device was stated to provide high foulant and salt removal for both synthetic solutions and real wastewater effluent, making it a promising advance for organic removal while maintaining high desalination performance. Unique solutions to organic fouling and alternative mitigation measures must continue to be rigorously explored for MCDI to be more widely commercialized. 8.9. General Calls for Action: Reporting of MCDI Results. Finally, after examination of the literature, improvements that could be made regarding the reporting of MCDI results remain. In particular, greater consistency between studies will allow for easier comparison of different ionexchange materials. The following suggestions are made, which we believe will aid in the reporting and understanding of the MCDI literature.
8.9.1. Performance Improvement Factor (R-factor). In this review, we have calculated, where possible, the performance improvement factor (R-factor, in terms of salt removal) when using MCDI compared to CDI without membranes. Such a factor should be calculated and stated explicitly for all MCDI studies. This could be done for all common factors such as SRE, SAC, and charge and current efficiencies. For ionexchange materials, this is crucial to observe the effect that the inclusion of a membrane has on the overall performance of the system. New studies would easily be able to compare their values to those of the other literature, and the proposed approach is a simple way to quantitatively observe the improvement of an ion-exchange material.
8.9.2. Feed Salt Concentration. The selection of feed salt concentration for desalination is a parameter that varies to a large extent across separate studies. However, the concentration of the salt solution can affect calculated performance metrics such as SAC. This makes comparison between studies an especially difficult task. To further aid comparison between works, efforts should be made to select some "standard" feed salt concentration for MCDI studies. For example, 500 mg L −1 would be an appropriate standard to use for low-salinity MCDI desalination, 10000 mg L −1 for a medium-salinity regime, and 30000 mg L −1 for a high-salinity regime, approaching seawater concentrations. Using consistent feed salt water concentrations is one of the easiest ways to ensure fair comparisons between works and a collaborative field between different research groups.
8.9.3. Cycling Stability. The long-term performance of an ion-exchange material is a factor that is often overlooked when evaluating the MCDI performance. However, the maintenance of performance over repeated desalination cycles demonstrates that the ion-exchange material remains intact and is feasible for industrial-like operation. The cycling stability of an MCDI system should be classified as being as important as performance metrics such as SAC, SRE, and charge, current, and energy efficiencies. First, a benchmark should be set in terms of either the number of cycles or the hours of operation over which a system must operate to exhibit cycling stability. For a single-pass system with shorter adsorption/desorption cycles, ideally the system would maintain performance over at least 50 cycles. For batch-mode systems that often have longer cycles, at least 30 h of consistent operation could signify a system with good stability. While such guidelines may not be applicable for every MCDI system, it is imperative that repeated desalination cycles be displayed in all studies. The desalination results obtained for one single cycle are likely to be far different from those of a system that has been operating over hours or days. The cycling stability should be emphasized and tested rigorously whenever a new ion-exchange material is developed for MCDI.