Electrochemical Methods for Water Purification, Ion Separations, and Energy Conversion

Abstract

Agricultural development, extensive industrialization, and rapid growth of the global population have inadvertently been accompanied by environmental pollution. Water pollution is exacerbated by the decreasing ability of traditional treatment methods to comply with tightening environmental standards. This review provides a comprehensive description of the principles and applications of electrochemical methods for water purification, ion separations, and energy conversion. Electrochemical methods have attractive features such as compact size, chemical selectivity, broad applicability, and reduced generation of secondary waste. Perhaps the greatest advantage of electrochemical methods, however, is that they remove contaminants directly from the water, while other technologies extract the water from the contaminants, which enables efficient removal of trace pollutants. The review begins with an overview of conventional electrochemical methods, which drive chemical or physical transformations via Faradaic reactions at electrodes, and proceeds to a detailed examination of the two primary mechanisms by which contaminants are separated in nondestructive electrochemical processes, namely electrokinetics and electrosorption. In these sections, special attention is given to emerging methods, such as shock electrodialysis and Faradaic electrosorption. Given the importance of generating clean, renewable energy, which may sometimes be combined with water purification, the review also discusses inverse methods of electrochemical energy conversion based on reverse electrosorption, electrowetting, and electrokinetic phenomena. The review concludes with a discussion of technology comparisons, remaining challenges, and potential innovations for the field such as process intensification and technoeconomic optimization.

1. Introduction

1.1. Conventional Methods of Water Purification

It is estimated that four billion people live in localities which are, for at least one month of the year, under conditions of severe water scarcity.^1,2 One increasingly common method used to secure supplies of potable water is desalination, and so the development of desalination systems that are energy and infrastructure efficient is a critical technological challenge.³ In the most general sense, desalination is a process that is used to remove ions, colloidal particles, chemical compounds, and organic matter, referred to hereafter by the single term contaminants, from saline water. Existing methods for desalination can be broadly categorized into physical methods and chemical methods.³

Physical methods include distillation,⁴⁻⁶ freezing (or freeze–thaw) desalination,^7,8 (liquid-phase) solvent extraction,⁹ membrane processes,¹⁰⁻¹² solar desalination,¹³⁻¹⁵ and wave-powered desalination.¹⁶ Distillation, a process which appears to have been used by early experimentalists of the classical era such as Aristotle,¹⁷ involves separation of water from contaminants across the interface between a gas and a liquid by selective boiling and condensation. Modern implementations of this method include multi-stage flash distillation, multiple effect distillation, vapor compression, and humidification dehumidification, all of which are in essence a sequence of countercurrent heat exchangers.^5,18−24 In a similar way to distillation, freezing desalination also uses a phase change (freezing and melting) to separate water from contaminants.^25,26 Solvent extraction is used to separate contaminants based on their relative solubilities in two immiscible liquids, normally water (polar) and an organic solvent (nonpolar), where transport is driven by gradients in the chemical potential of the contaminants.²⁷ Membrane processes are diverse in that the kind of (semipermeable) membrane used can be tuned based on the target contaminant from which the water is to be removed. These processes include microfiltration, ultrafiltration, nanofiltration, reverse osmosis (RO), and forward osmosis (FO), and they are distinct primarily in the pore sizes of the corresponding membrane.^11,28−38 During operation, water is driven across a membrane by an input of mechanical work (or by a gradient in osmotic pressure in the case of FO) to retain the contaminants in a concentrated brine. Like distillation, solar desalination is said to have been employed by humans for thousands of years, originally by Greek mariners and Persian alchemists.^39,40 The basis of this technology is not distinct from distillation or membrane processes; it is simply a means to generate the energy that these methods require: that is, heat for distillation or electricity for membrane processes. Wave-powered desalination is similar in principle to solar desalination methods in that it generates electricity (by the motion of submerged buoys in this case) to run a desalination process based on RO.¹⁶ Although these technologies are mature and are cost-effective for the desalination of seawater (as well as other concentrated solutions), they are inherently inefficient and energy intensive when used to treat dilute feeds or to selectively remove target contaminants from a concentrated feed, as explained in the following section. It is in these situations that selectively removing trace amounts of a desired species is preferable to indiscriminately concentrating all dissolved species in a brine.

Chemical methods, which tend to be selective in molecular separations, can be classified into two major types: the first, discussed in this section, involves no electrochemistry (i.e., the established chemical methods), and the second, introduced in sections 1.3 and 2, is based on electrokinetic and electrochemical phenomena. Established chemical methods include precipitation,⁴¹ coagulation flocculation,⁴² adsorption,^43,44 ultraviolet, ozone, and chlorine disinfection,⁴⁵ aeration,⁴⁶ and ion exchange.^47,48 Precipitation involves the creation of a solid (the precipitate) from a solution using a chemical referred to as the precipitant.^49,50 Similarly, coagulation flocculation involves the addition of compounds (typically metallic salts) that promote the clumping of fines into larger floc, which can be readily separated from water by sedimentation, filtration, or flotation.^51,52 Adsorption is a physicochemical phenomenon that is used to remove contaminants by binding them to the surfaces of an adsorbent material.^44,53 Disinfection technologies in general are used to kill bacteria, viruses, and other disease-causing pathogens present in water. The most common disinfection treatments are based on ultraviolet radiation, chlorination, and ozonation, all of which inactivate the waterborne pathogen by disrupting its cellular functions.⁴⁵ Aeration of water is achieved by passing air through the liquid and is typically used to remove iron or organic matter, dispel certain dissolved gases, or oxidize dissolved or suspended compounds.^54,55 Ion exchange represents a broad class of processes where ions are exchanged between an electrolyte solution and a solid ion exchanger, such as polymeric resin, chelating agents, zeolites, clay, and montmorillonite.^{47,48,56−62} Ion exchange is in general a reversible process, where the ion exchanger is regenerated using a wash solution. Altogether, these traditional chemical methods are based on either chemical reactivity (precipitation, adsorption, and chelation), affinity for charged or functionalized surfaces (coagulation flocculation and ion exchange), or susceptibility to oxidative degradation (disinfection and aeration) of the contaminants.

In practice, a water treatment process often combines several of the methods described above to improve the quality of water and make it suitable for a specific end use. Figure 1 shows the process diagram of a representative municipal water treatment facility, in which the water treated contains high levels of hardness and iron.⁶³ As described in ref (63), raw water is taken from wells and sent to an aerator, where contact with air removes volatile solutes (e.g., H₂S, CO₂, CH₄) and odorous substances (e.g., CH₃SH, bacterial metabolites). Contact with oxygen further promotes iron removal by oxidizing soluble Fe(II) to insoluble Fe(III). After aeration, lime is added (as CaO or Ca(OH)₂) to raise the pH and cause the precipitation of Ca²⁺ and Mg²⁺. Precipitates of these hardness ions settle from the water in the primary basin, and much of the remaining solid material is suspended and requires the addition of coagulants (e.g., Fe(III), Al₂(SO₄)₃) to settle. Activated silica or synthetic polyelectrolytes (e.g., poly(sodium styrenesulfonate), poly(acrylic acid)) may also be added to induce coagulation or flocculation. Settling of colloidal particles occurs in the secondary basin after the addition of carbon dioxide to lower the pH. Sludge from both basins is then pumped to a lagoon, and the water is finally chlorinated, filtered, and pumped to the water mains.

Figure 1.

Process diagram of a representative municipal water treatment facility. This process combines several of the methods described in section 1.1 to improve the quality of water and make it suitable for domestic use. Adapted with permission from ref (63). Copyright 2001 CRC.

1.2. Limitations of Conventional Methods

Almost all of the methods introduced in section 1.1 have seen commercial success for a range of applications across numerous industries. Each of these methods, however, has applications and operating constraints outside of which the use of an alternative process would be more practical. For example, thermal distillation has historically been the dominant means of seawater desalination, but the most prevalent technology used today in large-scale desalination plants is RO because of its high energy efficiency and small footprint.^33,64 This technology works by pumping the feed at pressures above the osmotic pressure of the solution through a membrane permeable only to water molecules (3–5 MPa for seawater).¹² RO has been optimized over several decades of development for the desalination of concentrated feeds like seawater,^38,65 and modern seawater RO (SWRO) plants currently require under 3 kW h m^–3 of energy when including all pre- and post-treatment steps.⁶⁶⁻⁶⁹ While RO is the best available solution for city-scale seawater desalination, RO systems demonstrate poor scaling of energy demand with decreasing feed concentration, as demonstrated in Figure 2.⁷⁰ For example, brackish water RO (BWRO) plants require nearly the same energy input as SWRO plants (1–3 kW h m^–3), despite the fact that brackish water is less concentrated than seawater by about an order of magnitude.⁷¹ According to the van’t Hoff equation, osmotic pressure is linearly proportional to concentration for dilute solutions, which suggests that the energy demand for BWRO should be about an order of magnitude lower than for SWRO. Friction losses in RO systems, however, do not scale with salt concentration, but rather with the amount of water transported across the membrane. This intrinsic feature of RO explains the relatively high losses and poor energy efficiency of BWRO compared to SWRO.⁷² Because desalination of brackish water is a promising solution for water scarcity, and because removal of trace contaminants from dilute feeds is an important capability, technologies whose energy demand scales with feed concentration would be more desirable than RO for these applications.

Figure 2.

Estimates of the volumetric energy consumed by RO and a generic electrochemical process based on the analysis in section 6.1, specifically eqs 17 and 18. These estimates assume that the feed is desalinated to a final concentration of 1 μM; here, γ = 0.5 (water recovery, defined as the fraction of the feed recovered as permeate) and 10 L h^–1 m^–2 (productivity). Energies are compared to the thermodynamic limit, represented by the dashed curve, and are reported in units of kW h m^–3 of diluate.

Another notable drawback of RO plants is that they require large capital expenditures and mature infrastructure, which limits their utility for small-scale applications or in remote locations.⁶⁶ Moreover, it is difficult to downscale RO systems because high-pressure pumps and resilient plumbing are required at any scale to pressurize the feed in excess of the osmotic pressure. Facile downscaling to smaller, less-expensive plants that can be introduced into small residential areas and communities would help water treatment technologies further penetrate the market. Inexpensive, small-scale plants may in fact be the most appropriate solution for developing and off-grid locations, where water scarcity is severe and commonplace, and where infrastructure may be underdeveloped or nonexistent. Small-scale portable desalination units are also in demand by industrial facilities, by mobile military units and vessels, in recreational spaces, and in the travel industry.^73,74 To meet these growing and diverse needs for purified water, the technological focus should extend beyond RO by including scalable systems with lower energy demands and more flexible infrastructure requirements for treatment of brackish water and dilute feeds.

Many of the drawbacks of RO, particularly when used to treat dilute feeds, can be overcome by using the physicochemical methods introduced in section 1.1. These methods, however, have their own limitations and often require nonreusable chemicals. Solvent extraction can be both efficient and cost-effective in separating hazardous contaminants from benign feeds, but this process requires large volumes of organic extractants and sometimes toxic solvents, and the entrainment of phases yields low-quality effluents.^75,76 Precipitation is another simple and cost-effective process that is commonly used to remove toxic heavy metals from water, but it produces large amounts of sediment and sludge that is often difficult and expensive to dispose of. Precipitation is also ineffective at removing ions that are present at low concentration, and its utility may be limited when the water is contaminated with multiple metals.^76,77 Coagulation flocculation is often employed after precipitation to remove solid particulates from water, and this method could also be used to capture larger particles and inactivate biological agents.^78,79 The applicability of this method is limited, however, because it requires nonreusable inorganic coagulants which are usually toxic.^76,80 Ion exchange, on the other hand, offers a wider range of simple and well-established commercial products, many of which can be regenerated for repeated use.^76,77,81 The performance of ion exchange resins and chelating agents, however, is sensitive to variations in pH, and some of these agents react with dissolved metal ions to form soluble metal complexes that lead to secondary pollution.^76,77,82

1.3. Emerging Electrochemical Methods

A variety of innovative techniques based either on electrokinetics or electrosorption have been proposed for water purification and ion separations, and these techniques have given rise to emerging electrochemical methods.⁸³⁻⁹¹ The recent discovery of deionization shock waves in microchannels^92,93 and porous media,^87,94,95 for instance, inspired a new area of research in electrokinetic methods for deionization. Parallel developments in materials science have uncovered a wealth of novel electrode chemistries, where the electrosorption of ions is promoted by Faradaic reactions, to replace carbon, conventionally the material of choice in capacitive systems.^86,96 These innovations have not only enhanced deionization capacity but have also imparted molecular selectivity to the electrodes. Recent examples of Faradaic platforms for water purification are based on electrochemical reduction of target contaminants,^97,98 electrochemical switching of ion exchange,⁹⁹⁻¹⁰¹ and molecularly selective removal of ions,^86,96,102 uncharged compounds,^103,104 and biomolecules (e.g., proteins).¹⁰⁵ Many of these advances have relied on Faradaic compounds with immobilized surfaces to achieve superior electrochemical performance and chemical specificity. By modulating the binding interactions at the surfaces, the affinity of the electrodes can be tuned specifically for minority components in a feed, which in practice may be either highly valuable or seriously toxic.

Emerging electrochemical methods include electrodeionization (EDI, sometimes called hybrid ion exchange ED),^61,106−108 shock electrodialysis (shock ED),^87,109 capacitive deionization (CDI),^110,111 battery deionization (BDI),^112,113 and Faradaic electrosorption.^86,96 These technologies are unique from all of the others discussed so far in that removal of contaminants is based on their response to electric fields in solution or electrochemical reactions at electrodes. Electrochemical systems use applied electrical currents to remove contaminants from the feed by either driving separations in bulk electrolytes,^61,114 electrochemically trapping them in electric double layers (EDLs),^90,110,115 or intercalating them in solid electrodes (e.g., materials composed of two-dimensional, layered structures).^116−118 The first of these mechanisms is governed by electrokinetics, and the second and third are forms of electrosorption, as explained in Figure 3. The primary input of energy to these systems is an applied electric potential, which makes these processes scalable without the need to be operated at extreme temperatures or pressures. Energy dissipation in electrochemical systems, however, arises from three general sources:^357,1173 (i) ohmic resistance, due to hydrodynamic drag acting on moving ions in electrolytes or membranes,⁷² as well as electronic resistance in porous electrodes and current collectors; (ii) Faradaic reaction resistance at electrode–electrolyte interfaces, leading to activation overpotential; and (iii) concentration polarization, associated with limitations in ion diffusion. These losses all scale with the number of ions removed rather than the amount of solution process, as shown in Figure 2. These systems therefore tend to be more energy efficient compared to physical methods (and are molecularly selective) when used to treat brackish water and dilute feeds.

Figure 3.

Electrokinetics and electrosorption are the two main mechanisms by which contaminants are separated in any nondestructive electrochemical process. Electrokinetic processes, which are typically continuous, involve transport of charged or uncharged but dielectric^119,120 species in an electrolyte in response to an applied electric field, and so removal of contaminants relies on effective mass transfer. Electrokinetic methods like EDI and shock ED use porous materials in the feed channels to increase ion removal and improve energy efficiency when the feed is dilute. Electrosorption processes are cyclic and encompass all phenomena in which the binding of contaminants is aided by an applied electric field. In addition to effective transport, electrosorption relies on favorable reaction kinetics and thermodynamics.

Among the existing electrochemical methods, electrodialysis (ED) has been studied and used for water desalination for decades, and several ED desalination plants for treatment of brackish water are currently operational in the U.S.^{114,121−125} The past decade, however, has seen the emergence of several novel electrochemical systems for water purification with unique functionalities and working principles compared to ED, and these are EDI,¹²⁶ shock ED,⁸⁷ CDI,^115,127,128 and Faradaic electrosorption.⁸⁶ This review summarizes the development of these novel technologies to form the basis for the emerging field of electrochemical systems for water purification and ion separations. For completeness, we also briefly discuss related microfluidic technologies, which have been reviewed in detail elsewhere.^129,130 Microfluidic systems, which are driven by electrical energy, have throughputs at the scale of nanoliters and may be difficult to scale up to the volumes needed for deployment for human consumption, agriculture, or industry. Electrochemical systems instead are based on cells and stacks made of components like ion exchange membranes (IEMs), porous dielectric separators, and porous electrodes that can be produced in large areas as flat sheets or films, and thus these systems naturally have clearer pathways for scale up.

1.4. Outline of This Review

Conventional electrochemical methods, which exploit Faradaic electron transfer reactions at electrodes to drive chemical or physical transformations, have been previously reviewed, particularly in their use for removal of organic matter,¹³¹ organic compounds,^132−137 inorganic contaminants,^138−142 and microorganisms^143−145 as well as for degradation of various contaminants and micropollutants.^{123,146−149} The technologies on which many of these publications focus are well established and currently used in industry. We begin the review by introducing these established electrochemical transformation methods and the broad range of applications for which they are used (section 2). We then build on the existing literature and related reviews^{68,105,115,118,123,126,150−155} by emphasizing the principles and applications of emerging electrochemical methods for desalination, water purification, and ion separations (sections 3 and 4). As shown in Figure 4, several of these emerging methods have developed only in the last 10 years and have been reviewed either briefly or independently of other methods. To provide a foundation for our discussion, we examine recent developments in electrokinetic phenomena and electrosorption for water purification and selective ion separations. See refs (76) and (118) for tables that summarize the advantages and disadvantages of techniques used for water purification, including many of the emerging methods discussed here.

Figure 4.

Number of publications per year for various methods used for water purification or ion separations. The technologies considered here are broadly classified as either physical (top row) or electrochemical (bottom row). This review focuses on electrochemical methods, and special attention is given to the emerging methods, including shock ED, CDI, and Faradaic electrosorption. The literature search was performed using Elsevier’s Scopus database with the search field TITLE-ABS(···).

In section 3, we explain (nondestructive) methods based on electrokinetics, which may either include or exclude membranes. Two methods in the class of membrane-based systems are ED and EDI, where the membrane plays a critical role in the removal of charged species. Ions can be separated from the bulk fluid directly, however, by virtue of the electrokinetic phenomena, without much contribution from the membrane itself. Key developments have also been made in the class of “semimembraneless” systems (i.e., systems in which the feed is partitioned into fresh and brine streams across a deionization shock wave that functions as a virtual interface), which include microfluidic (or nanofluidic) concentration polarization and shock ED. These methods are both based on the phenomenon of concentration polarization, which arises due to extreme gradients in the concentration of ions in solution. In section 4, we describe electrosorption, the phenomenon that is responsible for sorption in electrochemical systems, and how this process is used for desalination and molecular separations. (These separations are often nondestructive, but in certain cases, they may involve electrochemical conversion or degradation of the contaminants.) This explanation is followed by an overview of CDI techniques as well as recent innovations in Faradaic (or redox-active) materials and their broad use in chemical and environmental processes. With these methods explained, we introduce inverse methods of energy conversion that convert gradients in salinity to energy (section 5). We conclude by discussing the energetics, thermodynamic and technological challenges, and prospects of electrochemical methods for water purification and ion separations (sections 6 and 7).

2. Electrochemical Transformations

In the general areas of water purification and wastewater treatment, a variety of electrochemical processes have been developed to remove contaminants ranging from ions to colloidal particles. For example, chemical coagulation flocculation, flotation, precipitation, and redox (reduction and oxidation) can be improved by applying electric fields.^156−158 In this section, we discuss nonelectrosorptive electrochemical methods that are well established and reviewed extensively in the literature, namely electrochemical oxidation, electrochemical reduction, electrocoagulation, electroflotation, and electrodeposition.^{98,146,156−160} These processes involve Faradaic reactions at electrodes to drive chemical or physical transformations of ionic or molecular solutes. For example, electrochemical redox reactions are used primarily when the objective is to degrade or convert nonbiodegradable organic contaminants and certain inorganic compounds (e.g., cyanides, thiocyanates, sulfides) and disinfect water. In section 3, we examine both traditional and emerging electrokinetic methods based on ED, which rely on coupled transport phenomena in electrolyte solutions. We then discuss in section 4 electrosorption systems (both capacitive and Faradaic) for selective ion removal based on various functional materials, including porous carbon, inorganic crystals, and polymers.^90,115

2.1. Electrochemical Oxidation

Electrochemical oxidation is a chemical reaction involving the loss of one or more electrons by an atom or a molecule at the anode when an electrical current is passed through the system.^146,161,162 In the context of water treatment, electrochemical oxidation generates reactive oxidizing agents called free radicals that interact with the contaminants and degrade them, as explained in Figure 5.¹²³ Superoxide (O₂^•–), hydroperoxyl (HO₂^•), hydroxyl (HO^•), and sulfate (SO₄^•–) radicals are examples of reactive agents that can degrade organic and organometallic contaminants by initiating a radical oxidation chain (see refs (123), (146), and (163−167) for lists of common reaction pathways by which these radicals are formed).^{136,168−170} Superoxide and hydroxyl are two of the most important radicals in free-radical chemistry, and they are both believed to be key species in oxidative processes.^163,164,171 While superoxide is normally a nucleophile and reducing agent,^164,172 it exists in equilibrium with the hydroperoxyl radical, which can behave as an oxidizing agent in various biological and chemical reactions.^173−177 (In general, superoxide is a weak reducing agent, but in the presence of solids or cosolvents that are less polar than water (e.g., H₂O₂), superoxide becomes reactive and can degrade halogenated aliphatic compounds,^178,179 including perfluorocarboxylic acids,^180,181 via nucleophilic attacks.¹⁸²) The hydroxyl radical oxidizes both organic and inorganic compounds with high reaction rates, such that its action occurs only in the region where it is produced.¹⁸³ The degradation process begins with formation of the reactive oxidizer, followed by initial attacks on target molecules and their breakdown into biodegradable intermediates. Subsequent attacks on these intermediates by the oxidizer can lead to their mineralization (i.e., production of water, carbon dioxide, and inorganic salts), as shown in Figure 5.^123,146 Another class of oxidizers is obtained by the oxidation of chloride ions to generate active chlorine (Cl₂), which may disproportionate to hypochloric acid (HClO) or hypochlorite (ClO^–) depending on the pH. Although these species can effectively oxidize various contaminants in real wastewaters (e.g., landfill leachates, textile effluents, olive oil wastewater, tannery wastewater),⁹⁷ this approach has the drawback of producing chlorinated organic compounds during the electrolysis, which is the main limitation of electrochemical oxidation.^97,184 These chlorinated byproducts increase the toxicity of the effluent because they tend to be much more persistent than what is initially present in the feed.^185−187 In the absence of chloride electrolytes, however, electrochemical oxidation can be reliably used for disinfection, wastewater treatment, groundwater treatment, soil remediation, wastewater sludge conditioning, and odor and taste removal.^{146,188−190}

Figure 5.

Electrochemical oxidation is an established destructive, nonelectrosorptive method that is used to degrade most organic contaminants and certain inorganic compounds. (top) The method involves formation of reactive oxidizers that interact with contaminants either at the anode surface (direct oxidation) or in the bulk (indirect, or mediated, oxidation).¹²³ Reproduced with permission from ref (191). Copyright 2017 Royal Society of Chemistry. (bottom) Complex organic compounds, such as the sulfonamide antibiotic sulfachloropyridazine, are degraded and mineralized by the attack of reactive oxidizers at key reaction sites (designated by the letters A–D). In this example, hydroxyl radicals can attack four unique sites on sulfachloropyridazine to yield different primary cyclic byproducts, all of which are eventually transformed into CO₂. Reproduced with permission from ref (192). Copyright 2012 American Chemical Society.

In the 1990s, researchers became increasingly aware that the anode material is an especially important consideration in the design and optimization of electrochemical oxidation processes.^193−196 (The cathodes are usually stainless steel plates, platinum meshes, or carbon felt electrodes.) The results obtained by several groups indeed demonstrated that the choice of anode influences the selectivity and efficiency of organic compound oxidation.^165,197 According to a model proposed by Comninellis,¹⁹⁴ anode materials are divided into active anodes (e.g., carbon, platinum, iridium oxides, ruthenium oxides) and nonactive anodes (e.g., antimony-doped tin oxide, lead dioxide, boron-doped diamond).⁹⁷ Active anodes have low oxygen evolution overpotential and are good electrocatalysts for oxygen evolution, while nonactive anodes have high oxygen evolution overpotential and are poor electrocatalysts for oxygen evolution. Anodes based on boron-doped diamond (BDD) have received considerable attention due to their chemical stability, high electrical conductivity, resistance to corrosion even in harsh environments, and wide window of electric potential.¹⁹⁸ As a result, BDD is generally viewed as one of the most effective and energy efficient anodes for mineralization of organic contaminants, although its use in practice is limited due to high manufacturing costs.¹⁹⁸ Moreover, BDD (as well as other anodes) promotes the oxidation of chloride to chlorate (ClO₃^–) and perchlorate (ClO₄), which are water-soluble disinfection byproducts that are exceedingly mobile in aqueous solutions and are highly persistent under typical water conditions.^97,146 Because disinfection byproducts in drinking water are usually regulated, their concentrations should be monitored and controlled when performing in-line electrolysis.

Other prominent anodes include dimensionally stable anodes (DSAs, also called mixed metal oxide electrodes) and substoichiometric titanium oxide anodes.^97,199,200 DSAs are fabricated by coating a substrate such as titanium with several kinds of metal oxides, including RuO₂, IrO₂, and PtO₂. These anodes exhibit high conductivity and corrosion resistance, and recent studies showed that doping DSAs with metal and nonmetal elements can further improve their performance.^201−204 Moreover, the use of nanotechnology has gained traction in the field of electrode fabrication to increase the porosity and active surface area of the anodes.^205,206 Anodes based on substoichiometric titanium oxide (Ti_nO_2n–1) also display high conductivity and corrosion resistance, and their many advantages and long service life have led to their broad use in fuel cells, lead-acid batteries, and, most recently, wastewater treatment.^207−210 Ganiyu et al. prepared a Magnéli-phase Ti₄O₇ electrode by plasma deposition and compared it to DSA and BDD anodes for the degradation of the beta-blocker propranolol²¹¹ and the analgesic paracetamol.²¹² These studies showed that the Ti₄O₇ electrode can achieve similar or better removal of organic carbon compared to DSAs and BDD. Several methods have been reported in the literature to improve the performance of substoichiometric titanium oxide electrodes, and these methods include plasma spraying,²¹³ which produces doped functional coatings, and high-temperature sintering, which produces an electrode with extensively interconnected macropores.²¹⁰

Most anodes used in electrochemical oxidation produce highly active hydroxyl radicals on their surfaces,²¹⁴ and treatment of wastewater with these materials requires adequate flow of the contaminants toward them. When the concentration of pollutants near the anode is low, the process rate is limited by mass transfer of these species to the surface of the electrode.⁹⁷ Common methods to overcome this limitation include gas sparging,²¹⁵ incorporation of turbulence promoters,⁹⁷ and use of nanoengineered materials.^216−218 Most importantly, the efficiency of electrochemical oxidation can be improved with the indirect (or mediated) oxidation method, which avoids the production of oxygen by generating precursors that are transformed to active oxidizers. Persulfate (S₂O₈^2–), percarbonate (C₂O₆), and hydrogen peroxide (H₂O₂) are examples of precursors that can be produced using BDD anodes.^219−221 These species are relatively stable at ambient conditions and generate highly active inorganic radicals that enable mediated oxidation of organic contaminants.⁹⁷ Michaud et al. experimentally tested the production of precursors using BDD and observed that persulfate is produced with high current efficiency when the electrolyte is concentrated in sulfate (SO₄^2–) and the process temperature is low.²¹⁹ The persulfate precursor is activated to generate sulfate radicals, and this step requires a transition metal catalyst or sufficient energy.²²² Activated persulfate can then oxidize organic compounds, and this approach has been used to treat groundwater and soils contaminated with biorefractory organic species.⁹⁷

In recent years, there have been significant advances in the design, synthesis, and use of nanostructured electrodes for electrochemical oxidation.²¹⁸ Nanoengineered materials exhibit new and improved properties, such as an increase in the number of active sites and an improvement in electrical conductivity, and these materials can promote heterogeneous catalysis at electrode surfaces.²¹⁸ According to Du et al., a wide range of nanostructured cathodes have been reported in literature, and they can be divided into four categories: cathodes based on carbon nanomaterials such as carbon nanotubes (CNTs) and graphene,^223,224 carbon cathodes doped with heteroatoms such as fluorine and nitrogen,^225,226 metals or metal oxides deposited on carbon,^227−229 and metal oxide cathodes.^230,231 In most studies, these electrodes are used for electrosynthesis of H₂O₂ via oxygen reduction, and the structural morphology and composition of functional groups largely affect cathode performance.^224,232,233 Du et al. also explain that there exists a large variety of nanostructured anodes, which can be divided into four categories similar to those of the nanostructured cathodes: anodes based on carbon nanomaterials such as CNTs and nanostructured BDD,^234−237 anodes doped with heteroatoms such as fluorine and boron,^238,239 metals or metal oxides deposited on carbon,^240−242 and metal or metal oxide anodes.^243−246 Most articles published on nanostructured electrodes report improvements in the kinetics of HO^• production and pollutant oxidation due to the synergistic effects of greater stability, electrical conductivity, electrochemical reactivity, and active-site exposure.²¹⁸ Even though nanoengineering has enabled major advances in improving electrode stability, more efforts are needed to demonstrate stability for long-term use, and standard protocols must be established to assess the lifetime and reliability of these systems.²¹⁸

Over the past two decades, scale up of anode systems has gained attention, where the focus has been on increasing the throughput of laboratory-scale systems while retaining performance and reliability.^247,248 At the same time, electrochemical advanced oxidation processes have been developed to improve the efficacy and applicability of conventional electrochemical oxidation.^{146,249−251} These specialized variants of electrochemical oxidation introduce Fenton’s reaction chemistry,^192,252,253 photoelectrocatalysis,²⁵⁴ sonoelectrolysis,¹⁴⁶ and aerobic or anaerobic digestion (using microbial electrochemical technologies)^255−258 to the standard process.²⁵⁹ In conventional electrochemical oxidation, the reactive oxidizers are often (but not always) produced at the anode surface; the advanced variants facilitate additional generation of oxidizers in the bulk.^146,197 Although these oxidative processes are more widely studied and used (because they usually lead to mineralization of the contaminants), treatments based on electrochemical reduction have been gaining interest because they enable partial recovery of chemicals as well as production of value-added substances.⁹⁷

2.2. Electrochemical Reduction

Electrochemical reduction, the complementary process to electrochemical oxidation, is a chemical reaction involving the gain of one or more electrons by an atom or a molecule at the cathode when an electrical current is passed through the system.¹²³ Similar to electrochemical oxidation, electrochemical reduction can occur either directly on the surface of the cathode or indirectly in the bulk by the action of a reducing agent generated at one of the electrodes.⁹⁸ This process is typically used to treat water contaminated with heavy metal ions (see section 2.5 also),^123,260 inorganic anions (e.g., bromate, perchlorate),^261,262 or halogenated organic compounds (e.g., organic volatile halides, chlorofluorocarbons, polychlorohydrocarbons, polyhalophenols)^97,98 by converting these species into more benign products. As shown in Figure 6, the mechanism of this conversion usually involves the removal of halogen atoms or the reduction of aldehydes and ketones to produces less toxic species.^97,123

Figure 6.

Electrochemical reduction is an established conversion method that is used to treat oxidized contaminants, such as (top) inorganic and (bottom) organic halides (R-X). The method involves formation of high energy electrons or reactive species that interact with contaminants either at the cathode surface (direct reduction) or in the bulk (indirect reduction). M refers to the cathode material (i.e., the catalyst, C_at); C_at-H_ads, (R-X)_adsM, and (R-H)_adsM are the hydrogen atom, organic halide, and dehalogenated organic compound (R-H), respectively, adsorbed on the cathode. Reproduced with permission from ref (98). Copyright 2020 Elsevier.

Among the major parameters that determine the efficiency of electrochemical reduction are catalyst loading, cathode potential, and water quality.⁹⁸ Generally, an increase in catalyst loading improves reduction activity, though only up to a limit beyond which activity either steadies or even decreases as the distance for electron transfer increases.²⁶³ In the case of nanosized catalysts such as palladium, the particles can aggregate at higher loading, which results in excessive local evolution of hydrogen bubbles that restrict access of the contaminants to the catalyst.²⁶⁴ As is the case for catalyst loading, a reaction will have an optimal operating cathode potential above which the abundance of hydrogen bubbles produced could inhibit adsorption of contaminants.^265,266 Electrochemical reduction is also sensitive to the quality and characteristics of the feed, which influence performance and electrode lifetime.⁹⁸ Performance typically improves at lower pH (due to increased formation of adsorbed hydrogen),^267,268 at higher ionic strength (due to smaller EDLs),⁹⁸ and in the absence of certain species (e.g., organic matter, electrocatalyst poisons, competing ions).^264,269

The choice of electrode is another critical design parameter that influences the mechanism of electrochemical reduction, as it impacts the reaction pathway, selectivity, and energy consumption.^97,98 An effective catalyst enables strong bonding on the surface of the substrate.⁹⁸ From among the many materials investigated to date, electrodes based on silver, nickel, and carbon hold prominent positions due to their high electrocatalytic activity, robustness, and inexpensiveness with respect to conversion of halogenated contaminants.^97,98 Noble metals such as palladium, platinum, and ruthenium are also effective materials for electrocatalytic hydrodehalogenation,²⁶³ especially when combined with other elements to produce bimetallic catalysts.^270−272 These bimetallic catalysts can even be modified by adding nanosized or anchored materials to further improve their electrocatalytic efficiency.^263,273,274 The main limitation of metallic catalysts, however, is their high cost, which makes the use of carbon-based materials an attractive alternative.^98,275 Some of the most effective carbon-based catalysts involve nanostructured polymer coatings that selectively adsorb halogenated compounds,^276,277 but the electrocatalytic activity of modified carbon materials in general is low.²⁷⁸ Activity can be improved by combining nanostructured carbons such as reduced graphene oxide (RGO) with metallic electrodes.²⁶⁴

In general, electrochemical reduction is an effective method not only for treatment of pollutants, such as volatile organic halides and chlorofluorocarbons, but also for their transformation into value-added products.⁹⁷ This capability can be achieved by selective removal of halogens as well as by carboxylation or carbonylation of the organic compounds. The combination of electrochemical reduction and electrosynthesis is another way by which wastewater can be treated and upgraded for synthesis of value-added organic products.^279,280 Moving forward, it will be important to assess and improve the stability and lifetime of electrocatalytic materials for use in practical applications.

2.3. Electrocoagulation and Electroflocculation

The two-step process of electrocoagulation and electroflocculation relies on the dissolution of metal anodes to induce the formation of flocs, which trap contaminants and enable their removal by settling, sedimentation, precipitation, or flotation.^{156,281−285} This method was patented in 1906 by Dieterich for treatment of sewage in London and bilge water from ships using iron and aluminum as sacrificial anodes.^286,287 As shown in Figure 7, the crucial electrochemically mediated step relies on in situ oxidation of the appropriate metal surfaces (often aluminum or iron)²⁸⁸ to produce metal ions which then form flocs that facilitate the removal of solids, organic species, and inorganic compounds.¹⁵⁸ What follows is essentially ordinary coagulation: the (typically negative) surface charge of the contaminants is neutralized, which destabilizes them and causes them to form aggregates that can be removed by settling, sedimentation, precipitation, or flotation.²⁸⁹ Electrocoagulation is regularly used in industrial applications, such as removal of heavy metals, remediation of wastewater, and treatment of produced water.^290,291,1269 To improve the performance of electrocoagulation for wastewater treatment, numerous studies have sought to integrate this technology with other processes, such as peroxidation or a more specialized biological process, both of which facilitate the removal of organic matter.²⁹²

Figure 7.

Established nondestructive, nonelectrosorptive methods of water purification. (a) Electrocoagulation and electroflocculation is a two-step process by which metal anodes are dissolved to induce formation of flocs that trap contaminants for removal by settling, sedimentation, precipitation, or flotation. (b) Electroflotation produces bubbles by water redox to transport lighter contaminants or flocs by flotation. (c) Electrodeposition is based on the electrochemical deposition of metal ions in solution onto an electrode.

Electrocoagulation has been extensively studied, often using iron electrodes, for specialty separations and wastewater treatment, particularly to remove light organic pollutants such as oils, dyes, and humic particles.²⁹³ In addition, electrocoagulation can be used with aluminum electrodes to remove heavy metal ions, including Ag⁺, Zn²⁺, Cu²⁺, Ni²⁺, and Cr⁶⁺,^294−296 as well as halide ions.²⁹⁷ Compared to classical coagulation, electrocoagulation has several advantages that lower its operating cost. For example, the cationic coagulant is generated in situ by a chemical reaction on the sacrificial electrode, and this feature limits the introduction of counterions from chemical reactants that contribute to the formation of sludge.²⁹⁸ Electrocoagulation thus requires no separation of unreacted counterions from the chemical coagulant in solution, which are usually removed to meet discharge standards.²⁹⁹ The reduced formation of sludge also lowers expenses associated with handling and disposal of this waste. Another advantage of electrocoagulation is that the electrochemical reactions that drive this process produce OH^–, which eliminates the need for external chemical agents to regulate pH.³⁰⁰ These reactions also produce gaseous H₂, which could be captured and used subsequently as a fuel.³⁰¹

In contrast to classical flocculation, which requires the input of large amounts of chemicals, electroflocculation relies on Faradaic (anodic) dissolution to dose the system with the coagulant, a property that enables finer control and easier handling of the process. Complexation of the released metal ions, such as Al³⁺ or Fe³⁺, with hydroxides produces monomeric (Al(OH)₂⁺ and Al(OH)²⁺) or polymeric species (Al₁₃(OH)₃₄) that then flocculate with organic compounds (e.g., dyes) and further promotes precipitation by either sedimentation or flotation.³⁰² Despite their advantages, however, electrocoagulation and electroflocculation exhibit operational challenges, such as electrode passivation, sludge deposition on the electrode, nonuniform dissolution of the anode, and inconsistent production of the coagulant, all of which undermine performance under long-term, continuous operation.³⁰³ The sacrificial anode is also consumed over time, which necessitates periodic replacement of the electrode, and the high concentration of residual (metal) ions requires a post-treatment step prior to discharge.³⁰⁴ But because of their operational simplicity, low cost, and versatility, electrocoagulation and electroflocculation remain active areas of research.³⁰⁵

2.4. Electroflotation

Electroflotation is often used in conjunction with electrocoagulation and electroflocculation for electrochemical–physical separations.^306−309 Electroflotation relies on the electrolytic process of water redox, in which bubbles are formed to transport lighter contaminants or flocs by flotation. The key driving Faradaic reaction for electroflotation is the electrolysis of water at both electrodes (O₂ evolves at the anode and H₂ at the cathode). The first proposed use of electroflotation is attributed to Elmore in a patent from 1905 for mining separations,³¹⁰ and this process has since been expanded to handle a range of contaminants such as oils and low-density suspended particles in mining water and groundwater, among others.³⁰⁶ The primary limitation of electroflotation is the difficulty of controlling the uniformity of bubble evolution.^158,311 Careful optimization of current densities and voltage windows is therefore crucial for effective performance.^{139,282,308,312}

Electroflotation has several features that are attractive for applications in water treatment. For example, this process can be used to recover valuable components from wastewater without the need for chemical reagents.³¹³ The reason is that electroflotation generates gaseous O₂ and H₂, which are more active than the gases used in conventional flotation (e.g., natural gas, air, N₂). Electroflotation can also produce bubbles with diameters of 1–30 μm, which leads to better dispersion, finer distribution, and longer residence times of the bubbles in solution.^314−316 This feature facilitates the flotation of fine particles in a way that is difficult to achieve by classical flotation. Finally, the energy consumption of electroflotation is in the range of 0.1–0.5 kW h m^–3, and it decreases as the electrical conductivity of the solution increases.³¹⁴

2.5. Electrodeposition

The process of electrodeposition is synonymous with electroplating and is one of the earliest applications of electrochemistry, especially in metallurgical processing.^{139,317−319} While many of these plating methods are major sources of heavy metal pollution, the same electrochemical principles have been used to treat industrial wastewater and sometimes even to recover and reuse discharged materials. As an example, copper electrodeposition is commonly used to print circuit boards and manufacture electronics. Recovery of copper from spent parts, however, has also been an application of electrochemical methods that combine leaching, ED, and electrodeposition from wastewater.^320,321

Electrodeposition involves the application of cathodic overpotentials to induce the electrochemical deposition (or reduction) of metal ions in solution onto an electrode. In other words, this process is direct electrochemical reduction of metal ions adsorbed on an electrode surface. Electrodeposition can effectively handle both copper and arsenic wastes, often with the production of pure elemental copper depending on the electrochemical parameters.^322,323 This method has also been used for secondary recovery of residual copper from low-content tailings derived from waste electrical cable.³²⁴ These applications of electrodeposition rely on the same principle of removing metal ions from aqueous solutions that is used to charge aqueous metal flow batteries, such as zinc–air,³²⁵ zinc–bromine,³²⁶ zinc–iron,³²⁷ and lithium–air batteries.³²⁸

2.6. Challenges and Limitations

Established nonelectrosorptive processes exhibit irreversible side reactions that consume significant amounts of energy and reduce current efficiency.²⁸⁷ Depending on the complexity of the electrochemical matrix, a number of byproducts can be produced, some of which passivate the electrodes and further diminish performance.^282,287,329 For example, even though there were attempts as early as the late 1800s to implement electrocoagulation and electroflotation at scale, these processes never evolved into mainstream technologies because of their prohibitively high operating costs.²⁸² Electrocoagulation and electroflotation have also lacked systematic studies aimed at scaling up the processes and optimizing their operating parameters.^282,305 These methods, however, remain promising for localized and small-scale applications, and they continue to offer interesting avenues for scientific research in interfacial science, electrochemical engineering, and reactor design.

3. Electrokinetic Separations

When current is applied across a pair of electrodes, ions and larger particles in solution are transported by electromigration and electrokinetic phenomena,^91,330−332 such as surface conduction, electroosmosis, and electrophoresis. Water purification by electrokinetics is based on the transport of contaminants in an electrolyte, and methods of this kind can be used to remove both organic^{159,333−335} and inorganic^{68,115,153,336} ions from water. Electrokinetic methods such as ED, EDI, and shock ED are continuous and involve an electric field that is perpendicular to the direction of fluid flow. These methods also include IEMs to fractionate the feed into diluate and concentrate streams. EDI and shock ED are similar in that a porous material (e.g., ion exchange resin beads, ceramics, clays, porous glass) is used to enhance mass transfer across the liquid to the solid phase. This unique feature of EDI and shock ED allows for currents beyond the ideal diffusion-limited current and makes these methods well suited to remove trace contaminants from dilute feeds. In this section, we briefly discuss the operating principles and basic physics of ED and EDI. We then focus on the emerging method known as shock ED, which is being reviewed for the first time. Our discussion of shock ED is preceded by an examination of the key developments in microfluidics and electrokinetic modeling that inspired the invention of shock ED as a method for water purification and ion separations. We conclude this section with a discussion of fouling phenomena and methods to overcome them in electrokinetic systems.

3.1. Electrodialysis

3.1.1. Basic Principles of Electrodialysis

ED is a membrane-based electrochemical method that was developed in the 1950s^114,337 for desalination of brackish water and for ion separations.^338−343 The operating principles and underlying physics of this method are summarized in Figure 8 (see refs (152 and 153) for detailed reviews of membrane phenomena in ED). The process begins by passage of the feed through a stack of parallel, nonporous cation and anion exchange membranes (CEMs and AEMs; see refs (344 and 345) for detailed reviews of IEMs and the state of their development). At the same time, direct current (DC) is applied across the stack, perpendicular to the direction of flow to separate ions in alternating channels of fresh water (diluate) and brine (concentrate). In the sections where diluate is produced, cations pass through CEMs and anions through AEMs on the opposite side, which lowers the concentration of uncharged salt in a boundary layer that grows into the channel downstream. In the sections where concentrate is produced, anions are retained by CEMs and cations by AEMs, which results in boundary layers of increasing salt concentration.^338,346,347 Once the boundary layers span the entirety of the channels, a condition termed “fully developed” forced convection,³⁴⁸ the dissolved salts have been effectively transported from diluate channels into neighboring concentrate channels for discharge.

Figure 8.

Operating principles and underlying physics of ED. (a) The feed flows through a stack of channels separated by alternating CEMs and AEMs, across which a current is applied. These channels are filled with spacer material that physically separates the membranes, promotes turbulent mixing, and inhibits ICP and the formation of laminar boundary layers near the membranes.^349,350 The concentration profiles of the ions exhibit axial growth of boundary layers of depletion and enrichment in the diluate and concentrate channels, respectively. Adapted with permission from ref (346). Copyright 1968 Elsevier. (b) Ion fluxes near a CEM and the associated convection–diffusion boundary layers, which become more strongly depleted with increasing current until the diffusion limit is reached. (c) The current–voltage relationship reaches a plateau at the limiting current, while overlimiting current, which results in an extended zone of ion depletion, is observed at higher voltages.

Depletion and enrichment of salt in unsupported electrolytes lead to self-generated diffusional electric fields that impede the transport of active ionic species through the phenomenon of ion concentration polarization (ICP). These electric fields arise to maintain electroneutrality via redistribution of the inactive (typically oppositely charged) ionic species. The additional internal voltage drop associated with variations in salt concentration relative to that of electrical conduction in a uniform bulk electrolyte (Ohm’s law) is termed the concentration overpotential. When the salt concentration tends to zero, the concentration overpotential diverges and leads to a diffusion-limited current, as long as the transport is dominated by electrodiffusion in a neutral unsupported electrolyte without the creation of any additional ions.^331,351,352 For example, the mobilities of cations and anions in an electrolyte are comparable but may differ considerably in an IEM or a nanochannel with charged surfaces.³⁵³ Salt depletion dramatically increases the electrical resistivity of the solution, which leads to significant departures from Ohm’s law as shown in Figure 8, and limits the achievable rate of ion removal by ED.^{338−341,354} The current–voltage relationship for ICP in a symmetric (z:z) binary electrolyte resembles that of an ideal diode:

1

where I is current, I_lim is the diffusion-limited current (or limiting current),⁹⁵z is charge, V is voltage, F is Faraday’s constant, R is the universal gas constant, and T is temperature. The maximum diffusion-limited current, I_lim, is approached when a large potential difference (relative to the thermal voltage, RT/F ≈ 26 mV at room temperature), V, is applied to the system and as the salt concentration in the diluate sections approaches zero.

The theory of ED was pioneered by Peers,³⁵⁵ after which it was extended for arbitrary transference numbers by Rosenberg and Tirell³⁵⁶ as well as for fully developed convection by Sonin and Probstein.³⁴⁶ A useful approximation for the limiting current is given by the Peers equation,

2

where c₀ is the bulk salt concentration, D is the effective salt diffusivity for coupled diffusion and electromigration (usually taken to be the ambipolar diffusivity in a dilute, binary electrolyte,^357,358 although concentration dependent corrections can be significant³⁵⁹), τ_α is an effective transference number for counterions that selectively cross the solution–membrane interface (τ_α = 1 for an ideal membrane or electrode),^339−341τ_β is the transference number for counterions in the solution, and δ is the effective boundary layer thickness. This thickness scales as δ/H ∼ Pe^–1/3 in the Lévêque approximation,^{338,346,348,360} where Pe = UH/D is the Péclet number, defined by the channel spacing, H, and characteristic fluid velocity, U. This theory is based on a boundary layer analysis of the steady convection–diffusion equation for an uncharged binary electrolyte,^338,348 where the flow is assumed to be fully developed and unidirectional and axial diffusion is neglected, as is usual for forced convection in straight pipes and channels.^348,361 Similar boundary layer approximations can be derived for membraneless flow batteries with forced convection over (selective) redox electrodes instead of IEMs.³⁶² Extensions for turbulent flow,³⁶³ including effects of screen spacers to promote turbulent mixing,³⁶⁴ were incorporated in a general equivalent circuit model of ED by Belfort and Guter.³⁶⁵ Theoretical models were also used to calculate the pH profile³⁶⁶ and analyze ion selectivity (e.g., K⁺ versus Ca²⁺,³⁴⁷ NO₃^– versus Cl^–367) in ED.

In practical systems, the ideal diffusion-limited current is always exceeded when the applied voltage is sufficiently high, and possible mechanisms of this “overlimiting current” have been extensively studied in membrane science.^153,368 In bulk liquid electrolytes, there are two main kinds of mechanisms responsible for overlimiting current: electrochemical and electrokinetic.³⁶⁹ Electrochemical mechanisms involve charge regulation^370−376 and self-ionization of water molecules,^72,355,377 both of which may lead to current induced membrane discharge³⁷⁸ (i.e., loss of ion selectivity) and in turn passage of co-ions that would otherwise be repelled by the membrane. Electrokinetic mechanisms are based on the Rubinstein–Zaltzman electroconvective instability, where the EDL (either in³⁷⁹ or out of equilibrium^380,381) between the depleted solution and the membrane becomes hydrodynamically unstable to electroosmosis and in turn drives bulk vortices that transport ions to the membrane faster than by diffusion.^382−387 In ED, electroconvective mixing can enable overlimiting current and improve the transport of ions,³⁸⁸ although at the expense of greater energy consumption compared to operating below the limiting current. These observations led the scientific community to develop ways to control electroconvection, including the use of topological heterogeneity, for example by patterning the surface of an IEM,^{387,389−392} and pulsed electric fields.^393−397 Phenomena resembling electroconvection also arise in metal electrodeposition, and they strongly influence the formation of patterns.^398,399 Understanding and controlling these instabilities in bulk electrolytes is a grand challenge, the solution to which could lead to novel electrochemical systems that benefit from operating in the exotic regime of overlimiting current despite the higher energy demand.

The efficiency of ED depends on device structure (e.g., spacer thickness and geometry, number of cell pairs in the stack), membrane properties (e.g., material chemistry, concentration of the fixed ionic moiety),^{364,400−404} electrode design (e.g., capacitive flow or membrane electrode, electrode redox couple),⁴⁰⁵ operating conditions (e.g., electric potential, current density, hydrodynamics, temperature), and feed composition.^114,406 On the basis of their structure, commercial IEMs are classified as either homogeneous or heterogeneous.⁴⁰⁷ A homogeneous IEM generally displays higher conductivity and permselectivity compared to a heterogeneous membrane because the latter comprises a larger insulating phase in its matrix as a result of the fabrication method.^408,409 A heterogeneous IEM, on the other hand, often has greater mechanical strength and is less expensive to manufacture.⁴¹⁰ The heterogeneous structure of such a membrane also promotes electroconvection and mass transfer by localizing the migration of ions through the conductive parts of the membrane.^{387,411−415} Because a commercial ED stack may contain 300–500 cell pairs,⁴¹⁶ the conductivity of the membrane will largely determine the overall conductivity of the stack, which in turn influences the energy consumption of the process. Another critical feature of IEMs is their resistance to the formation of deposits (e.g., organic fouling, inorganic scaling), which degrade the performance of the membranes and negatively affect the quality of the water produced.^417,418 Much research has therefore been devoted to developing IEMs that have improved antifouling properties,^419−422 as discussed in section 3.6.

3.1.2. Electrodialysis With Bipolar Membranes

One class of ED systems that remains an active area of research is based on bipolar membranes, which facilitate water dissociation without the need for chemicals.^{114,344,423−425} Bipolar membranes are fabricated by combining a CEM and an AEM, with a hydrophilic contact region separating the two IEMs. It is at this contact region where water is split into H⁺ and OH^–, which migrate into acid and base compartments, respectively, when an electric potential is applied. The primary benefits of using bipolar membranes with ED are that they catalyze the production of H⁺ and OH^– at voltages lower than what is needed for standard electrolysis at an electrode, reduce the amount of concentrate generated, and increase the recyclability of the waste by recovering ions from the feed as acids and bases.^58,426,427 The use of a bipolar membrane, however, also introduces an electrical resistance to the system that lowers the current efficiency. This decrease in efficiency is overcome by lowering layer resistance and incorporating weak ion exchange groups in the membrane.⁴²⁸ As a result, the energy required to produce H⁺ and OH^– with commercially available bipolar membranes is nearly equal to the theoretical minimum value⁴²⁸ (the theory of bipolar membranes is discussed extensively in refs (425), (428−432)).

ED systems with bipolar membranes have several applications, one of which is the recovery of acids and bases from salts produced by chemical reactions or neutralization.^433−436 Because ions must first be separated to produce these acids and bases (e.g., Li⁺ for LiOH, Na⁺ for NaOH, SO₄^2– for H₂SO₄), this process desalinates a concentrated feed in the same way as conventional ED.^{434,437−439} The first industrial process to employ ED with bipolar membranes was developed by Aqualytics to recover HF and HNO₃ from pickling baths in the steel industry.^434,440 Although commercial use of this technology remains limited due to the high cost, low permselectivity, and short lifetime of existing bipolar membranes,^114,434 novel applications are rapidly emerging, such as pH control, carbon capture, production and recovery of ammonia, and energy storage.⁴²⁵

3.1.3. Electrodialysis Driven by Chemical Energy

Recently, chemical energy was used in ED to simultaneously drive separations and generate electricity (Figure 9a).^441−444 Inspired by microbial desalination cells,^445,446 these ED systems are driven by inorganic redox couples, but they do not actually employ microbes because the use of microorganisms can limit electricity production and salt removal rates.⁴⁴⁷ ED cells driven by chemical energy can simultaneously produce clean water and generate electricity by performing a combined reaction–separation process that is thermodynamically spontaneous. Atlas et al. calculated the maximum available energy from this combined process, as shown in Figure 9b, and quantified the thermodynamic efficiency of cells driven by chemical energy, as shown in Figure 9c.⁴⁴⁷ The authors also found that, for certain chemistries, up to 25 kW h m^–3 of electricity can be produced. This technology can thus generate significant excess electricity, well above what is that needed for pre- and post-treatment of the feed, which is approximately 1 kW h m^–3 for seawater.⁶⁸

Figure 9.

Operating principles of chemical energy ED. (a) Schematic of an ED cell driven by chemical energy, where a reductant R and oxidant O are used to drive desalination and generate electricity. Adapted with permission from ref (441). Copyright 2019 Elsevier. (b) Calculations of the maximum available energy from an ED cell driven by chemical energy for various redox couples. (c) Measured thermodynamic efficiency of ED cells driven by hydrogen–oxygen and zinc–bromine redox couples. Reproduced with permission from ref (447). Copyright 2020 The Electrochemical Society.

The concept of chemical energy ED was tested with a variety of redox chemistries, including zinc–bromine,⁴⁴¹ zinc–air,¹⁵⁴ aluminum–air,⁴⁴³ hydrogen–oxygen,^447−451 and acid–base⁴⁵² couples. In particular, the hydrogen–oxygen couple is promising, as it relies on relatively inexpensive gas-phase reactants, and the product of the chemical reaction is simply water (Figure 9a); cells that use the hydrogen–oxygen chemistry are termed “desalination fuel cells.”^444,447 Other chemistries that rely on liquid-phase reactants or that produce a waste product complicate disposal of the brine.^441,443 The hydrogen–oxygen chemistry, however, exhibits relatively low thermodynamic efficiency relative to other chemistries, such as zinc–bromine, mainly due to losses at electrodes attributed to (platinum) catalyst poisoning by halide ions in the brine (Figure 9c).^447,453 Therefore, a crucial area of research is the design and development of inexpensive catalyst materials tailored to long-term operation in desalination fuel cells. Asokan et al. demonstrated the use of chloride-tolerant, iron-based catalysts for oxygen reduction in a desalination fuel cell, which opens the field of nonplatinum group metal catalysts for these systems.⁴⁵⁴

3.2. Electrodeionization

EDI, shown in Figure 10a, is typically used to generate highly pure products by processing feeds with low levels of dissolved solids (e.g., RO permeate).^{126,455−459} This method originated in the late 1950s with the intent of enabling extreme deionization of contaminated feeds by packing the channels of an ED stack with charged porous media or conductive ion exchange resin beads.^61,106 The purpose of these conductive materials is to reduce the ICP observed in ED by enhancing transport of ions via electrokinetic phenomena (see Figure 11c and Figure 13). As in ED, ions are depleted in the diluate compartments of an EDI stack and are concentrated in the adjacent (concentrate) compartments because of the permselective properties of the IEMs. To deionize the electrolyte while maintaining reasonable conductivity across the stack, however, a conductive material such as ion exchange resin is needed to lower the resistivity of the electrolyte in the diluate compartments.^61,460 As shown in Figure 10b, there is an optimal current density that is believed to coincide with the limiting current introduced in section 3.1.⁴⁶¹ Another, more subtle reason ion exchange resin is used in the diluate compartments in EDI is to regulate the pH of the product streams by exploiting the relationship between the applied electrical current and the equilibrium concentrations of H⁺ and OH^– in solution.⁴⁶²

Figure 10.

Design and operating principles of EDI. (a) As in ED, ions are depleted in the diluate compartments of an EDI stack and are concentrated in the adjacent (concentrate) compartments due to the permselective properties of the IEMs. In EDI, however, the use of a charged porous material (usually ion exchange resin) enables extreme deionization by boosting the conductivity of the electrolyte in the diluate compartments. (b) Typical current–voltage relationship in EDI, which reveals the existence of two regions with distinct resistivities (characterized by the slope of the graph). Reproduced with permission from ref (461). Copyright 2005 Springer.

Figure 11.

Role of ion-exchange resin in EDI. Comparison of (a) current–voltage and (b) resistance–voltage relationships in ED and EDI. Reproduced with permission from ref (464). Copyright 2009 Elsevier. (c) Schematic of the three pathways for charge transport in a mixed bed. (d) Specific conductance of the ion exchange resin versus that of the interstitial solution. Reproduced with permission from ref (469). Copyright 2015 Elsevier.

Figure 13.

Schematic representations of common mechanisms to sustain overlimiting current in a microchannel between a reservoir on the left and a CEM on the right. The volume average conductivity in (a) exhibits classical linear diffusion (continuous line) up to a region of depletion (dashed line), where charge carriers are transported by (b) surface conduction, (c) electroosmosis, or (d) electroconvection (visualized experimentally by Gu et al.⁵⁶⁴). The relative contributions of these effects depend on the width of the channel, such that surface conduction dominates in narrow channels and electroconvection in wide ones. Reproduced with permission from ref (527). Copyright 2011 American Physical Society.

As illustrated in Figure 8c, a typical ED cell exhibits a dramatic increase in electrical resistivity when salt is depleted from the diluate compartments. Because the concentration of ions in these compartments is smaller near the membranes than it is in the bulk, water dissociation occurs and reduces current efficiency.^463,464 The use of ion exchange resin (whether as loose beads or as a wafer^83,465,466) in EDI mitigates this effect by promoting transport in a conductive medium, which functions as a bridge for ions between membrane pairs. The current–voltage curves in Figure 11a,b show that EDI maintains increasingly higher conductivity (or lower resistivity) than ED as a function of applied voltage.¹²⁶ Pathways of charge transfer in the ion exchange bed of an EDI cell are often studied using the porous plug model^467−469 introduced by Wyllie and co-workers.⁴⁷⁰ The overall conductivity of the ion exchange bed is the sum of the conductivities of the ion exchange resin and interstitial solution. These conductivities are sustained via three different pathways, namely the resin (solid), the interstitial solution (liquid), or both (solid–liquid), as shown in Figure 11c. Alvarado et al. studied these pathways using an EDI cell as a mixed bed to treat a synthetic solution containing chromium at concentrations up to 250 mg L^–1.⁴⁶⁹ In this system, nearly 82.9% of the electric charge was transported by the combined solid–liquid pathway, whereas only 0.6% was transported by the interstitial solution alone. In total, the ion exchange resin, alone or in combination with the interstitial solution, contributed to approximately 94% of ion transport, which underscores the crucial role of ion exchange in driving charge transfer in EDI (Figure 11d).

In addition to facilitating the migration of ions, ion exchange resin enables operation beyond the limiting current, at which diffusion of ions becomes the rate-limiting step in electrochemical separations.⁴⁷¹ Operating in the regime of overlimiting current leads to water dissociation and exotic electrokinetic phenomena like the electroconvective instability,^368,472,473 as was microscopically visualized by Park and Kwak as well as Stockmeier et al.^474,475 In EDI, water dissociates due to the presence of bipolar zones formed at points of contact between resin particles and either other resin particles or IEMs.^26,465 As the ion exchange resin traps ions, H⁺ and OH^– produced by water dissociation act as charge carriers and regenerate the resin through the process of electroregeneration.¹⁰⁶ Water dissociation is also important when complete removal of weakly ionized species, such as silicon and boron, is needed to produce ultrapure water. The OH^– generated by this dissociation reacts with silicon and boron as follows:⁴⁷⁶

3a

3b

3c

Once the neutral species SiO₂ and H₃BO₃ become ionized, they are readily transported into the concentrate compartment and discharged.

Another useful function of the ion exchange resin in EDI is that, as a selective medium, it preferentially removes ions based on their affinity to the resin.⁵⁷ Selective separation of ions with similar charge and size can be achieved by controlling the mobility of these ions with a complexing agent.^477,478 For example, Taghdirian et al. used a complexing agent made of ethylenediaminetetraacetic acid (EDTA) to separate Ni²⁺ from Co²⁺.⁴⁷⁸ This complexing agent formed a strong bond with Ni²⁺, which produced a negatively charged complex whose mobility was inhibited in the bed of cation exchange resin. On the other hand, Co²⁺ remained a free cation that could readily enter the gel phase of the resin. By removing Co²⁺, the molar ratio of Ni²⁺ to Co²⁺ in the solution was increased from 3 to over 150.⁴⁷⁸

Although the microscopic mechanisms of EDI have received much less attention compared to those of ED, there have been several efforts to mathematically model^{107,459,479−482} and numerically simulate^483−486 the process of EDI. Early descriptions of EDI proposed that the removal of ions occurs in two steps.⁴⁸⁷ First, ions diffuse from the bulk to the liquid–solid (resin) interface, where counterions are exchanged with mobile ions on the resin. Second, the adsorbed counterions are transported toward and across the corresponding IEM, where they are released into the concentrate compartments. Removal of ions is controlled by the rate of diffusion from the aqueous phase to the surface of the solid.^126,487 This rate is determined by the properties of the solid surface, the thickness of the liquid layer through which ions diffuse, and the concentration gradient between the two phases. When a current is applied beyond what is needed for the electromigration of ions to the surfaces of the resin, water molecules dissociate into H⁺ and OH^–, which replace the ions that have adsorbed on the resin.¹²⁶

3.3. Electrokinetics in Nanochannels and Membranes

While nanofabricated devices are difficult to manufacture at scale, the scalable polymeric IEMs used in ED, EDI, and shock ED are essentially made of a network of nanoscale pores with high charge densities, as shown in Figure 12a.^{337,344,488,489} This nanoscale structure promotes selectivity based on the charge of an ion and on chemical interactions between the ion and the pore walls.⁴⁹⁰ IEMs are an integral component of many electrochemical methods of ion separations, and this fact underpins the development of membrane properties like conductivity and selectivity.^{337,344,491,492} There also exist exciting opportunities for enhanced separations as well as reduction in fluidic resistance using engineered nanoscale conduits like CNTs^493−496 and graphene,^497−499 but the success of these technologies in practice requires a high density of channels and a scalable and controlled fabrication process.

Figure 12.

Electrokinetic transport in charged porous media. (a) Schematic of a charged porous medium with solid (brown) and porous (blue) domains. Ion and fluid transport occur along the center axes of pores. The thickness of the EDLs and the amount of surface charge influence the electrokinetic coupling. (b) Fluid can flow due to gradients in pressure, electric field, or electrolyte concentration. Ion separations use these modes of transport to impart selectivity based on electric charge. Reproduced with permission from ref (507). Copyright 2017 American Chemical Society.

A membrane or surface in contact with an electrolyte solution either already has or will acquire a net charge, and under most circumstances, this surface charge is balanced (or screened) in the liquid by a diffuse layer of oppositely charged ions.^331,332 This concept of an “electric double layer” was originally proposed by Helmholtz,⁵⁰⁰ later refined by Gouy and Chapman^501,502 and revised once more by Stern.⁵⁰³ The local diffuse charge leads to electroosmosis and diffusio-osmosis in response to an electric field and concentration gradient, respectively.⁹¹ The equations that govern transport in charged nanoporous media are therefore coupled, such that gradients in pressure, concentration, or electric potential (Figure 12b) lead to combined fluxes in fluid flow, salt transport, and electrical current.⁵⁰⁴ Gross and Osterle described the set of coupled transport phenomena by the Poisson–Nernst–Planck–Stokes (PNPS) equations, with assumptions of local equilibrium and electroneutrality in the cross sections of pores.⁵⁰⁵ This approach presents a unified theoretical framework to describe ion and fluid transport in charged porous media down to the nanoscale, and it can be used to model ED, EDI, and shock ED. Recent work extended these results to numerically solve the PNPS equations⁵⁰⁶ and describe nanofluidic transport in nanopores.⁵⁰⁴ Systems modeled in this way include homogeneous networks with pores connected in series or in parallel as well as heterogeneous networks with pores of varying cross section.^507−509 The homogenization of the PNPS equations to heterogeneous porous media revealed complex flow patterns and vortices due to parallel connectivity in these materials.⁵¹⁰ Other models incorporated chemical interactions, mismatch in the diffusion coefficients,⁵¹¹ multicomponent electrolytes,^512−514 electroviscous effects,⁵¹⁵ nonionic solutes,⁵¹¹ salts with asymmetric valences,^516,517 reactions of fixed charges,^518,519 temperature gradients,⁵²⁰ electron conducting⁵²¹ or polarizable nanopores,⁵²² and dielectric effects^{512,513,523−525} to describe the distribution of ions and solvent molecules in the pores. Analytical approximations to the nonlinear transport equations are valid when the surface potentials are small⁵²⁶ or when the EDLs are overlapping.⁵⁰⁴

Perhaps the most fundamental principle of nanofluidic transport is electroneutrality, which is an implicit assumption embedded in the models previously discussed that dictates that the charge within a pore must balance the charge on the pore walls. Indeed, in deionization processes such as EDI and shock ED, the strong tendency toward electroneutrality in charged nanopores allows for residual conductivity, even as the concentration of ions in the bulk goes to zero, because counterions screening the surface charge are always present at a concentration determined by this charge.^{94,353,527,528} Levy et al., however, recently showed that one-dimensional confinement of ions to an isolated nanochannel in a dielectric matrix can lead to breakdown of cross-sectional electroneutrality, as the screening length diverges exponentially with decreasing ion concentration.⁵²⁹ Under these conditions, which are typical of biological ion channels,¹²⁷⁰ the role of ion-specific chemical interactions becomes more significant than that of electroneutrality, although this order of importance is reversed in membranes where the distance between parallel channels becomes smaller than their lengths.¹²⁷¹ In addition to energy barriers to enter a pore, intrapore energy barriers can govern ion transport in one-dimensional pores, where ions are confined to the molecular scale.⁵³⁰

Other open questions exist about the structure and dynamics of ions and solvent molecules in the EDL, particularly at high ionic strength^531,532 or in the case of extreme confinement.⁴⁹⁰ Discrepancies between molecular dynamics (MD) simulations and continuum electrokinetics predictions occur for channels with widths of ∼1 nm.^490,533 As another example, overscreening of surface charge at high concentration can lead to reversal of electroosmotic flow,^534,535 and a high surface charge density can cause electric charge to crowd an interface.^532,536,537 A theoretical explanation of these phenomena requires consideration of electrostatic correlations in the EDL that go beyond the mean-field electrostatic potential.^534,538,539 Transport models of dense ionic solutions can account for these correlations as well as coupled transport modes due to friction between pairs of species. For example, transport equations based on the Stefan–Maxwell model have been used to describe the coupled fluxes of ions and solvent molecules passing through nanoporous membranes.^540−543 Regardless of the model, fixed surface charge influences the distribution of ions in a nanopore and in turn the observed transport properties. The regulation of surface charge via chemical reactions also affects transport dynamics.^{372,375,544−547} While the surface charge is usually assumed to be in equilibrium, nonequilibrium reactions can participate in electrokinetic transport.^548,549

3.4. Ion Concentration Polarization in Microfluidics

The physical phenomenon that enables shock ED, the deionization shock wave, was first discovered and studied in the seemingly distinct context of ICP in microfluidic devices.^{92,93,353,550−556} What we now understand to be stationary deionization shock waves were observed by Wang et al. as early as 2005 in a microchannel near a nanochannel junction and used to trap and concentrate biomolecules in cross-flow, though without a theoretical explanation of the mechanism.⁵⁵⁷ In 2010, Kim et al. reported a microfluidic system that used ICP to desalinate seawater by applying current in the direction of flow and across a nanochannel junction,⁵⁵⁴ albeit at very low efficiency and in nanoliter volumes.⁵⁵⁵ A sharp, stationary depletion region was observed behind the nanojunction, and the deionized fluid below the shock was separated from the enriched brine into different microchannels. Separation of all charged solutes into the brine channel was also claimed,⁵⁵⁸ although this would appear to contradict the theory of electrodiffusiophoresis,⁵⁵⁹ which predicts the dominance of electrophoresis (motion in the direction of the Coulomb force) over diffusiophoresis (climbing salt concentration gradients, regardless of the charge sign) and suggests that fast-moving, positively charged particles in the deionized channel may not have been detected in the experiments.

The breakthrough in understanding microfluidic ICP that paved the way for shock ED occurred in 2009, when Mani, Zangle, and Santiago theoretically modeled⁹² and experimentally observed⁹³ propagating ion concentration shock waves for the first time in a microchannel–nanochannel device with negatively charged surfaces. Two microchannels were filled with a stagnant electrolyte and were separated by a nanochannel with thick, overlapping EDLs, which behaved like a CEM to induce ICP and initiate a shock wave of salt depletion in one direction and salt enrichment in the other. Mathematical modeling revealed how the nonlinear drift arising from electromigration (competing with electrodiffusion) can create breaking waves of salt concentration and propagating shocks,⁹² analogous to the shock waves that arise in isotachophoresis^560,561 and capillary electrophoresis.^562,563

3.5. Shock Electrodialysis

3.5.1. Deionization Shock Waves in Microstructures

While developments in microfluidics have been critical to advancing the scientific understanding of ICP under strong confinement, microfabricated lab-on-a-chip devices are neither scalable nor efficient for macroscopic ion separations and water treatment. To produce any meaningful volume of water, a prohibitively large number of channels or devices would need to be fabricated and operated in parallel. Porous media, on the other hand, represent much more compact systems of interconnected microchannels in parallel that can be manufactured at scale, and these include ion exchange resin beads, ceramics, clays, and porous glass. Moreover, theoretical and experimental work have demonstrated that overlimiting current can in fact be sustained in charged porous media.^{94,95,565−568}

In 2011, Mani and Bazant demonstrated the propagation of deionization shock waves in porous microstructures by theory and simulation.⁹⁴ At the same time, Dydek et al. described the transport processes that enable overlimiting current in a charged microchannel, as shown in Figure 13.⁵²⁷ This work explained that electroosmosis dominates the transport of ions when an electrolyte is confined to small pores (∼100 μm). When the pores are even smaller (∼1 μm), surface conduction becomes the dominant mechanism of overlimiting current. Experimental visualizations by Nam et al. later confirmed the theoretical predictions of the regimes in which surface conduction and electroosmosis sustain overlimiting current in a microchannel.^556,569 These observations and discoveries served as the foundation for developing a novel technology that can separate ions by deionization shock waves in porous media. Deionization shocks have also proven useful in controlling the high-rate electrodeposition of metals in charged nanopores,⁵⁶⁷ where shock electrodeposition can suppress dendritic growth⁵⁷⁰ and enable resistive switching memory⁵⁷¹ and rechargeable metal batteries.⁵⁷²

In 2013, Deng et al. reported experimental evidence for overlimiting current, sustained by surface conduction and electroosmosis, and deionization shock waves in a charged porous material.⁹⁵ This study was the first to observe these phenomena in porous media, and it demonstrated the feasibility of shock ED as a new platform to desalinate water and perform electrochemical separations.^95,573 The basic design of the original shock ED system used a silica glass frit (weakly charged porous material) sandwiched between two electrodes (which can be thought of as strongly charged porous material; recall the microchannel–nanochannel geometry of Mani et al.⁹²), as shown in Figure 14. (Note that the porous material and the IEM must have surface charges of the same sign. This condition ensures the propagation of depleted and enriched regions.⁹³) Because overlimiting current is sustained in shock ED to form a deionization shock wave near the cathodic CEM, as shown in Figure 14c, it is possible to deionize the feed to concentrations well below what could be achieved by standard ED. The experimental work by Deng et al. indeed demonstrated that shock ED can produce deionized water with concentrations of ∼10 μM (see Figure 14d).⁹⁵ In the same year, the nonlinear dynamics of ICP in porous media were mathematically described by Dydek et al. using a homogenized model (which assumes that the EDLs are thin relative to the nominal pore size), with emphasis on water treatment by shock ED.⁵²⁸ These two publications also proposed a scalable shock ED system that would be operated continuously.

Figure 14.

Prototype “button cell” used for shock ED. (a) Rectangular cross section and (b) photograph of the button cell prototype. This device is a cylindrical macroporous material submerged in an electrolyte and placed between copper electrodes. Electrical current is driven by the electrodeposition of copper onto the bottom electrode. (c) Current–voltage relationship in shock ED with 1 mM aqueous CuSO₄ as the electrolyte; the effect of the sign of surface charge is also shown. Dashed lines represent experimental data, and solid lines represent the fit to a theoretical model. (d) Energy cost per unit volume of deionized water with (dashed lines) and without (solid lines) the series resistances due to the reservoir and electrodes. Reproduced with permission from ref (95). Copyright 2013 American Chemical Society.

The original shock ED system by Deng et al. was a radially symmetric button cell that had to be operated in batch mode.⁹⁵ This geometry made it difficult to scale up the system, and it was clear to the authors at the time that a continuous process was more desirable. The second generation of shock ED devices, built by Schlumpberger et al. then improved by Conforti and Alkhadra, had a rectangular geometry and was shown to remove more than 99.9% of the salt present in solution.^87,574 This new system used a silica frit sandwiched between two CEMs and included a splitter at the outlet between the two membranes to partition the flow into a diluate and a concentrate, as shown in Figures 15 and 16a. In the range of concentrations tested using binary electrolytes of monovalent ions, Schlumpberger et al. showed that desalination was a function of only dimensionless current:

4

where z₊ and c₊ correspond to the charge and concentration of the cation and Q is the volumetric flow rate of the feed. This definition of I_lim can be interpreted as the rate of advection of positive charge across the device; it is assumed that the flux of anions is zero across ideal CEMs. The authors also described the effect of electroosmosis on the fraction of liquid recovered as desalinated water from the contaminated feed, often referred to in the literature as the water recovery. Because the zeta potential of the glass frit is negative, the applied electric field induces electroosmosis in the same direction^348,575 (i.e., toward the depleted region near the cathode) and as a result increases water recovery (to above 80% at high current⁸⁷). This effect resembles the function of an electroosmotic pump originally used in microfluidic devices^576,577 and later studied in relation to ICP in microchannels.⁵⁷⁸

Figure 15.

Shock ED as a continuous process. (left) Photographs and (right) illustration of the device that shows assembly: a working device consists of electrodes, wire, and a porous material sandwiched between identical IEMs. Inlet (outlet) streams are labeled contaminated, anolyte (anolyte out), and catholyte (catholyte out); fluid leaving the top edge of the porous material is split into fresh and brine streams. Reproduced with permission from ref (574). Copyright 2019 American Chemical Society.

Figure 16.

Operating principles and representative simulation results of shock ED. (a) Ions are transported by an electric field perpendicular to the flow, and the electrical current is driven by water redox at the electrodes. (b) Schematic to illustrate ion transport by advection (blue arrows) and streaming potential (red arrows) at the interface between a charged channel and the outlet. u^P is the hydrodynamic flow velocity, and q_kE_str is an electrostatic force generated by the streaming potential, where q_k is electric charge and E_str is electric field. Reproduced with permission from ref (581). Copyright 2021 American Chemical Society. (c) Steady-state profiles of concentration, c, electric potential, ψ, and velocity, u, at a dimensionless current of five. The velocity profile shows distortion of the depletion zone by electroosmotic vortices. The feed to all channels is 10 mM NaCl aqueous solution at a pH of seven. Reproduced with permission from ref (580). Copyright 2021 Elsevier. (d) Contour plots of dimensionless, depth-averaged concentration and flux vectors for Na⁺ and Pb²⁺ at six sample locations in the porous material. Reproduced with permission from ref (581). Copyright 2021 American Chemical Society.

To improve the mathematical model developed by Dydek et al. following these modifications,⁵²⁸ Schlumpberger et al. considered the role of linear electroosmotic flow and captured some aspects of deionization and water recovery observed experimentally.⁵⁷⁹ The model, however, significantly overestimates the extent of deionization and underestimates the overlimiting conductance. Recently, Tian et al. developed a more comprehensive model of shock ED that is valid for multicomponent electrolytes (irrespective of EDL thickness), captures the phenomena of electroosmosis, diffusioosmosis, and water dissociation, and incorporates more realistic boundary conditions.⁵⁸⁰ This model also describes the role of electroosmotic vortices at the inlet and outlet of the system, and it considers the effect of hydronium transport. Figure 16c shows the profiles of concentration, electric potential, and velocity obtained from numerical simulations, which have enabled quantitative predictions of deionization and overlimiting conductance that agree with the experimental data in ref (87).

3.5.2. Selective Separations by Shock Electrodialysis

The design introduced by Schlumpberger et al. was the first example of a scalable device that could be used for shock ED.⁸⁷ Recent work by Čížek et al. upgraded this design by using a multistack device with AEMs and various porous materials to remove Na⁺.⁵⁸² In both studies, however, only binary electrolytes of monovalent ions were considered, which encouraged examination of selective removal of multivalent ions. This research began with a systematic study in which Mg²⁺ was selectively removed from an aqueous mixture of NaCl and MgCl₂.^109,583 For a feed with Na⁺ and Mg²⁺ in molar proportions of nine to one, more than 99% of the Mg²⁺ was removed, even though the total desalination was only 68%. This outcome implied that the divalent ion was selectively removed in the presence of a competing monovalent ion. The results of this work inspired further research into the kinds of contaminants that can be targeted by shock ED in the presence of competing ions. The first of these studies used shock ED in two passes to desalinate artificial seawater (3.5 wt %), from which 99.8% of the salt fed was rejected and more than 99.99% of the Mg²⁺ was removed, as shown in Figure 17a.⁵⁸⁴ These results also revealed selectivity toward Mg²⁺, which was preferentially removed relative to all other species by at least one (Mg²⁺:K⁺) and up to nearly two orders of magnitude (Mg²⁺:Ca²⁺). Scaled (retention) selectivity in the fresh stream was calculated as a function of dimensionless current between each pair of unique species i and j using the equation¹⁰⁹

5

For example, if S_j:i was greater than one, then species j was selectively removed relative to species i. Despite the high rejection of salt and extreme selectivity toward Mg²⁺, desalination of seawater by shock ED was energy intensive. It was therefore concluded that a more suitable application of this technology would be to target trace contaminants in dilute feeds, such as toxic heavy metals in drinking water or radioactive ions in the process water of a nuclear reactor (i.e., water used in the boiler or for cooling).

Figure 17.

Representative results for selective ion separations by shock ED. (a) Concentration per pass (top and middle) and scaled selectivity (bottom) as functions of dimensionless current for the most abundant cations in seawater. Reproduced with permission from ref (584). Copyright 2020 Elsevier. (b) Deionization per pass (bottom) and cumulative deionization (top) at a dimensionless current of five for three cations normally present in the process water of light-water nuclear reactors. Reproduced with permission from ref (574). Copyright 2019 American Chemical Society. (c) Normalized concentration of Pb²⁺ and water recovery as functions of dimensionless current for removal of Pb²⁺ from water in the presence of excess competing Na⁺. Reproduced with permission from ref (581). Copyright 2021 American Chemical Society.

Nuclear fission is a common method to produce energy, although this process leads to contamination of the process water with several dissolved species, some of which are radioactive.^585,586 It is often desirable to selectively accumulate the radionuclides to reduce the volume of nuclear waste and facilitate its containment or disposal.^587−589 Shock ED was thus used to selectively and continuously remove cobalt and cesium from a feed of dissolved lithium, cobalt, cesium, and boric acid.⁵⁷⁴ This formulation modeled the contaminated water normally found in light-water nuclear reactors and in other nuclear processes.^590−593Figure 17b shows deionization per pass and cumulative deionization for each species (with the exception of boron, which was present predominantly as the electrically neutral boric acid) in three passes. In each of the first two passes, all three species were removed in nearly equal proportions, but in the third pass, Co²⁺ was preferentially removed. Overall, the three-step process led to a high cumulative deionization for each species, ranging from 96.3% for Cs⁺ to 99.6% for Co²⁺. On the basis of these results, a clear and consistent trade-off was observed between the extent of purification (as well as water recovery) and the energy demand of the process. In general, however, the energy demand of this process (which has not yet been optimized) was low, ranging between 1.8 and 4.8 kW h m^–3, because only charged species were targeted and essentially no energy was expended, removing boric acid, the most abundant species in the mixture.

Similar to radioactive species, heavy metals can damage normal functions of the human body and lead to heavy metal poisoning. Affordable and effective removal of toxic heavy metals from water, especially in the presence of excess competing ions, has been a longstanding goal in environmental science and engineering.⁵⁹⁴ Tian et al. recently demonstrated low-cost, continuous, and selective removal of Pb²⁺ from simulated drinking water using shock ED.⁵⁸¹ As shown in Figure 17c, this process removed over 98% of Pb²⁺ (to safe levels⁵⁹⁵ below 1 ppb) with a water recovery of 70% and at an energy cost of 0.1 kW h m^–3. At the same time, only 40% of Na⁺ was removed, which maintained electrolyte conductivity and lowered the energy demand of the process relative to nonselective methods.

Although these selective separations have been successfully demonstrated in practice, the basic physical mechanisms of selective ion removal by shock ED are still under investigation. Using the mathematical model of shock ED in ref (580), Tian et al. proposed several possible mechanisms for selective removal of multivalent cations. These mechanisms include greater separation of multivalent cations by the deionization shock wave, their high affinity to the negatively charged surfaces of the porous medium where transport is slow (due to the condition of no slip at solid surfaces), and a strong resistance to their emergence in the fresh product due to streaming potentials.⁵⁹⁶ The last two of these mechanisms are shown schematically in Figure 16b. Tian et al. also used this model to explain the selective removal of Mg²⁺ relative to Na⁺ observed in refs (109 and 584). A recent publication adapted this model to simulate and interpret selective removal of Pb²⁺ in the presence of excess competing Na⁺ (see Figure 17c).⁵⁸¹Figure 16d shows representative simulation results of concentration profiles and ion fluxes for both Pb²⁺ and Na⁺. Indeed, the greater flux of Pb²⁺ relative to Na⁺ out of the feed and into the cathode stream (see Figure 16b) underpins the selective removal of Pb²⁺ by shock ED.

3.6. Fouling in Electrokinetic Systems

Fouling is the accumulation of unwanted material (known as foulant), such as mineral scale, organic compounds, colloids, and biomass, on solid surfaces. In electrokinetic systems, fouling typically occurs on the surfaces of membranes⁴¹⁷ and porous media (e.g., ion exchange resin),^597,598 which diminishes performance and increases electrical resistance. If left uncontrolled, fouling may introduce additional problems, such as increased pressure drop, flow blockages and instabilities, material fatigue, and premature system failure. In this section, we discuss how fouling occurs in electrokinetic systems as well as common methods used to prevent and control it.

Mineral scale is deposited on membrane surfaces (or within the membrane itself) when ions such as Ca²⁺, Mg²⁺, HCO₃^–, and SO₄ precipitate (or sometimes crystallize) from solution as solid salts.^417,599Figure 18 shows scanning electron microscopy (SEM) images of deposits of substrates on both CEM and AEM surfaces, particularly on the faces in contact with the concentrate.^600,601 In the concentrate compartments of ED and EDI cells, the concentration of an ion can approach or even exceed its saturation point, which leads to precipitation of the ion. Turek et al. developed a model to predict the formation of mineral scale by computing the saturation level of divalent cations in the concentrate compartment of an ED cell.⁶⁰² Precipitation of minerals also depends on the pH and temperature of the electrolyte.^{417,601,603,604} For example, a basic electrolyte is concentrated in OH^–, which reacts with Ca²⁺ and Mg²⁺ as follows:⁴¹⁷

6a

6b

Figure 18.

SEM images of deposits of CaCO₃, Ca(OH)₂, and Mg(OH)₂ on the surfaces of three IEMs (MK-40 and MK-40MOD are CEMs; MA-41 is an AEM). Reproduced with permission from ref (601). Copyright 2018 Elsevier.

As shown in Figure 19, these reactions may occur via several pathways, many of which begin with the dissociation of water into H⁺ and OH^– in the boundary layers of the diluate compartment.^368,605 (see ref (606) for a more detailed explanation of the reaction pathways in Figure 19). Cifuentes-Araya et al. demonstrated that scaling of the CEM can be reduced by using a modified cell configuration and pulsed electric fields to control water dissociation.^606−608

Figure 19.

Pathways of mineral scaling on CEMs and AEMs in ED, many of which are initiated by water dissociation in the boundary layers of the diluate compartment. Reproduced with permission from ref (606). Copyright 2014 Elsevier.

Colloids, which are nondissolved suspended solids (e.g., silt granules, clay minerals, colloidal silica, metal nanoparticles, organic colloids), represent another common source of membrane fouling.⁴¹⁷ The key property of colloids that makes them potential foulants is their charged surfaces, which can lead to electrostatic interactions with oppositely charged IEMs.^609−611 For example, Lee et al.⁶⁰⁹ and Mondor et al.⁶¹¹ reported that colloidal silica deposits irreversibly on the surfaces of AEMs and forms so-called “cake” layers that cannot be removed by chemical cleaning.⁶¹² Similar to colloids, organic compounds (e.g., humic substances, proteins, carbohydrates, organic acids) exhibit electrostatic interactions with the fixed charge of an IEM that can cause organic fouling.^613−617 Unlike colloidal fouling, however, organic fouling may also occur via hydrophobic interactions between the foulant and membrane.^{417,618−620}

Although chemical and mechanical procedures can be used to remove foulants from membrane surfaces,⁶¹⁶ cleaning protocols increase operating costs and user intervention. Frequent cleaning may also damage the membrane and shorten its lifetime. Several strategies have thus been developed to reduce the amount of cleaning and maintenance required, the earliest of which was to periodically reverse the electrical polarity in ED (electrodialysis reversal) and EDI (electrodeionization reversal) systems.^{418,458,621−623} Membranes with intrinsic antifouling properties have also been explored as a means to reduce cleaning requirements.^{419−422,624} Engineering membranes that are resistant to fouling requires careful design of chemical composition, surface hydrophobicity and roughness, and the surface charge of functional groups.⁴¹⁷ For instance, increasing the hydrophilicity of IEMs by modifying their surfaces with sulfonyl groups reduces the formation of mineral scale.⁶²⁵ Similarly, the introduction of polyelectrolyte layers inhibits organic fouling by electrostatic repulsion.⁶²⁶ Other methods to make membranes hydrophilic involve the use of additives such as inorganic particles,^{419,420,625,627} some of which could even be placed inside the membrane to modify its entire matrix. Membrane surfaces can also be made homogeneous to improve their resistance to fouling. For example, heterogeneous IEMs exhibit a dramatic reduction in scale deposition after being coated with a homogeneous layer of (hydrophobic) Nafion.⁶⁰¹ The improved resistance to scaling was attributed to enhanced electroconvective mixing, suppressed water dissociation, and decreased crystal growth at the membrane surfaces.^601,628

4. Electrosorptive Separations

Electrosorption refers to the adsorption of dissolved species from a solvent onto the surfaces or into the bulk of a charged electrode and is driven by an applied voltage. The ability to cyclically trap and release ions without the need for chemical additives or swings in temperature and pressure is a major advantage of this process. The performance of devices that employ electrosorption depends on both effective transport of species to the electrode and adequate storage capacity by the electrode. When subjected to an electric potential, the surface of an electrode becomes polarized, and the mechanism by which electrosorption occurs depends on both the identity of the dissolved species and the electrode properties. For instance, ions in an electrolyte can be electrosorbed within the EDL that forms on a polarized surface, and they can be released by either attenuating or reversing the applied electric field. This mechanism is the basis of CDI with porous carbon electrodes, as shown in Figure 20a.⁶²⁹ A number of emerging applications use CDI for water purification, including desalination of brackish water and selective electrosorption of target ions.^{115,150,630−632} While electrosorption is generally limited to removal of charged species, porous carbons have the well-known ability to adsorb organic compounds, which can enable simultaneous removal of salt and uncharged organic contaminants by CDI.⁶³³ Many novel designs and cell architectures have been invented and characterized, such as membrane CDI (MCDI, see Figure 33)^70,634−637 and flow electrode CDI (FCDI, Figure 20b).^638−642 MCDI can improve cell cycle life and energy efficiency, and FCDI enables continuous desalination using a single CDI cell. Moreover, the structure of CDI cells can be categorized as either flow-between (FB) or flow-through electrode (FTE), depending on the direction of the feed. In FB cells, the feed flows between the two porous electrodes, while in FTE cells, the feed flows directly through the electrodes..^643−645

Figure 20.

Schematics of various electrosorption processes. (a) Static electrode CDI has electrodes that are rigid solids, such as porous carbon and intercalation materials, whereas (b) flow electrode CDI has electrodes that are made of a suspension (or slurry) of carbon beads in an electrolyte. When oriented vertically, FCDI is often referred to as fluidized bed CDI, in which the flow of suspended carbon is impeded by gravity to establish a densely packed fluidized bed. (c) Faradaic electrosorption comprises redox-active electrodes that can selectively remove target ions. Reproduced with permission from ref (646). Copyright 2020 Royal Society of Chemistry. (d) Typical cyclic voltammograms for electrostatic electrosorption, where double-layer charging is associated with relatively constant capacitance over a wide range of potentials, versus Faradaic electrosorption (or electrochemical adsorption), where redox reactions that transfer electrons between certain ions and the electrode yield peaks in the voltammogram at specific potentials.

Figure 33.

Comparison between the designs of conventional CDI and MCDI. Membranes in the latter repel co-ions and prevent them from escaping to the bulk, which in turn attracts more counterions to preserve electroneutrality. Reproduced with permission from ref (1032). Copyright 2014 American Chemical Society.

Selective ion electrosorption is an emerging and promising application of electrosorption technologies.^154,632,642 Compared to other methods such as RO and ED, standard CDI is membraneless and its active elements are the electrodes. These properties enable unique functionalities and mechanisms of selectivity not accessible to membrane-based systems. For example, the selectivity of inexpensive and widely available activated porous carbon electrodes in CDI can be tuned by various parameters such as electrode voltage, chemical charge, pore size, and cell charging time.^647−651 Through design and selection of functional materials, technologies for selective electrosorption are now well developed and can be used to separate both charged and neutral species from water.

Faradaic electrosorption, in particular, relies on the transfer of electrons to a distinct redox-active species bound to the electrode, which changes the oxidation state of that electrode and enables tunable binding of ions based on chemical affinity.^96,652,653 Systems for Faradaic electrosorption include crystalline intercalation materials, which operate on the same principles as lithium-ion batteries for charge storage, and redox-active polymers, which selectively interact with species through single-site binding. By changing the molecular structure of the functional material, the mechanism of binding can be adapted to a specific class of ions. In this section, we describe the general principles of electrosorption in CDI and provide a review of porous carbon, intercalation, and emerging redox-active electrodes for desalination and selective separations.

4.1. Capacitive Deionization With Porous Carbon Electrodes

4.1.1. Electrostatic Electrosorption in Electric Double Layers

The origins of CDI and electrosorption can be traced back to the work of Helmholtz, who in 1853 developed the first EDL theory to describe a monolayer of counterions at the surface of an electrode in terms of a mechanism of charge storage similar to a capacitor.⁵⁰⁰ In the early 1900s, Gouy revised this theory by claiming the electrolyte in the EDL to be mostly diffuse and by allowing both co-ions and counterions to spread over a large distance.⁵⁰¹ Chapman further improved the model by deriving the EDL differential capacitance for a symmetric binary electrolyte,⁵⁰² and Stern introduced a series capacitance to account for surface solvation.⁵⁰³ The Gouy–Chapman–Stern (GCS) theory emerging from these developments has become the most widely used model to understand EDLs, and it is relevant for many electrosorption systems, especially in the dilute limit. The theory of electrocapillarity and EDLs was extensively discussed by Grahame in 1947,⁶⁵⁴ and there have since been several publications addressing the application of this theory to electrosorption and molecular separations.^150,655,656

CDI models were first proposed by Johnson and Newman in 1971 based on an equivalent circuit for charging of EDLs in porous electrodes, and the models included an empirical factor for salt removal efficiency.⁶⁵⁷ The first self-consistent microscopic theory for EDL charging and salt removal, however, was not developed until the work of Bazant, Thornton, and Ajdari in 2004.⁶⁵⁸ This theory was motivated by the modeling of induced-charge electrokinetic phenomena,^120,532 where large applied voltages at blocking porous electrodes, including voltage steps⁶⁵⁸ and alternating voltages,⁶⁵⁹ lead to nonlinear double-layer responses such as transient salt depletion and space charge formation. From this work, the first mathematical model for CDI with porous carbon electrodes, which bridged EDL theory with applications in water purification and desalination, was developed in 2010 by Biesheuvel and Bazant⁶⁶⁰ and validated by direct numerical simulations with corrections for surface conduction.⁶⁶¹ Further extensions of the theory by Biesheuvel et al. include the addition of multicomponent asymmetric electrolytes, a generalized Frumkin–Butler–Volmer (FBV) model for parasitic Faradaic side reactions, a modified Donnan model for strongly overlapped EDLs, and a model for attractive ion image forces in metallic micropores.^662−665

Storage of ions in EDLs for the purpose of water purification was first proposed and demonstrated in 1960. In the decades that followed, this area of research advanced slowly, although interest has grown quickly since 2010 with breakthroughs in theoretical modeling, experimental methods, material performance, and surface modifications.^{115,154,632,666} The most widely used materials for ion electrosorption are microporous carbons due to their broad commercial availability, electrochemical stability, and high electrical conductivity.¹¹⁵ Prominent examples of microporous carbons include activated carbon, carbon aerogels, graphene, and CNTs.⁶²⁹ Early theoretical frameworks of electrosorption in these materials are rooted in EDL models for planar surfaces (e.g., Helmholtz, GCS).^658,660,667 One of the main theoretical advances in the past decade was to treat the EDLs in carbon micropores as EDLs in an IEM because geometric confinement in both cases leads to EDL overlapping. Current theories also assume a uniform potential in the liquid phase of a micropore as well as a Donnan potential drop between a micropore and its adjacent macropore. With these innovations, numerous modified Donnan-type models have been proposed and validated.^{664,668−670} One benefit of such EDL models is their simplicity, which facilitates their integration into an electrode or cell-level transport theory.⁶⁷¹

4.1.2. Ion Selectivity

Ions constitute a major subset of contaminants found in water. When present even at low concentration, ions like F^–, CrO₄^2–, AsO₄, Hg²⁺, and Pb²⁺ can pose a threat to the health of humans and animals.^672−674 For this reason, researchers have studied and developed platforms for targeted removal of ionic contaminants using CDI.^675−687 Selective adsorption by CDI can also be employed to recover valuable elements, such as lithium,^688−696 phosphorus,^697−702 and nitrogen.^{701,703−710} In this section, we briefly review several experimental works that focus on selective separation of ions from multicomponent solutions using porous carbon electrodes. In particular, we focus on studies that involve either two monovalent ions or one monovalent and one divalent ion, and we exclude studies that involve mixtures of more than two competing ions because of the complexity of these systems.^711−720 We then discuss the quantification of ion selectivity via a separation factor. Next, we consider the use of composite and functionalized electrodes for enhancement of ion selectivity. Lastly, we present the foundations of selectivity modeling at equilibrium. Several studies focus on theories that capture selectivity based on charge, size, or affinity, which are not reviewed in detail here.^649,651,664

Early work by Avraham et al. reported the possibility of selective removal of ions by tuning the pore size of microporous carbons.⁷²¹ In this work, the electrodes were modified by chemical vapor deposition (CVD) using toluene as a precursor. This treatment influenced the outer surface of the micropores and reduced their size, so only small enough ions could access the pores. Results of this study demonstrated facile storage of the smaller, monovalent cation (Na⁺) with reduced storage of larger, divalent cations (Ca²⁺ and Mg²⁺). A follow-up study by Noked et al. also used this CVD approach for selective removal of NO₃^– from a mixture of NO₃ and Cl^–.⁷²² NO₃^– and Cl^– are hydrated anions of similar size, but the openings of the pores were controlled for NO₃ to preferentially enter by virtue of stereoselectivity. That is, the size and structure of unhydrated NO₃^– was what determined the capacity for electrosorption. Cerón et al. tuned the pore size of hierarchical carbon aerogel monolith electrodes by adjusting the activation time to toggle selectivity toward either Na⁺ or Ca²⁺.⁷²³ In this study, Ca²⁺ was nearly completely excluded from the small pores of the electrode. In the large pores, however, MD simulations suggested that Ca²⁺ selectivity is limited at high applied potential due to the competition between volume exclusion and electrostatic forces as well as the more favorable desolvation of Na⁺. Zhang et al. studied cation selectivity via pore sieving, charge, and desolvation effects, and they also examined time-dependent ion swapping.⁶⁵⁰ Han et al. further explored the impact of ion size and pore characteristics on selectivity.⁷²⁴ Eliad et al. showed that the capacity of an EDL can be independent of ion size when the pores are much larger than the ions, which effectively disables ion sieving.⁷²⁵

In addition to ion sieving, other widely employed strategies to selectively separate contaminants leverage the influence of ion charge and size on the EDL structure in micropores. Gabelich et al., for example, studied two different electrodes of aerogel carbon with relatively large nominal pore sizes (approximately 4 and 9 nm) and showed that charge had the greatest effect on electrosorption capacity.⁷²⁶ They also found that smaller, monovalent ions showed improved storage relative to larger, divalent ions. These results prompted additional research, for example by Hou et al., who combined grand canonical Monte Carlo (GCMC) simulations with experiments using electrodes made of carbon aerogel.⁷²⁷ When two ions of the same charge but different sizes were present in solution (e.g., K⁺ and Li⁺), the smaller ions favorably screened the surface charge while occupying a smaller volume in the liquid phase. This result was also predicted by using the closed-form Boublik–Mansoori–Carnahan–Starling–Leland (BMCSL) equation of state to model hydrated ions in the micropore EDLs as hard spheres,⁶⁴⁹ building on earlier work using either local-density approximations for ion crowding in thin EDLs⁵³² or the Carnahan–Starling equation for overlapping EDLs⁶⁶⁴ and consistent with later experimental observations.^683,690,728

Ion separations in CDI are complicated, however, and cannot be fully explained by molecular or continuum models of equilibrium EDLs, especially in the presence of multivalent ions. For example, when a monovalent ion competes with a divalent ion (e.g., K⁺ and Ca²⁺), the GCMC simulations of Hou et al. predicted preferential electrosorption of the divalent ion (Ca²⁺), but this result was not observed experimentally. This observation was later attributed to ion-specific electrosorption dynamics by Zhao et al.,⁶⁴⁸ who developed a general theory of time-dependent ion-selectivity in CDI, based on the interplay of voltage-dependent adsorption capacities in the EDL and ion-specific transport resistances in the electrolyte. The theory was validated by experiments involving mixtures of Na⁺ and Ca²⁺, which also confirmed the prediction that selectivity toward Ca²⁺ would become high only after several hours of charging in their system.⁶⁴⁸ To design more selective CDI electrodes for concentrated mixtures of multivalent ions, it may be important to extend these models to account for electrostatic correlations and hydration forces.⁷²⁹ These effects can be incorporated using either local approximations based on the Bazant–Storey–Kornyshev (BSK) equation^{534,535,538,539} or nonlocal weighted density approximations⁷³⁰ based on fundamental measure theory⁷³¹ and charged-shell electrolyte models.⁷³²

Separations of anions using CDI have also been studied. Tang et al. observed moderate selectivity of Cl^– relative to F^– under constant voltage charging,⁶⁸³ a result later observed and validated theoretically under constant current charging.⁷³³ Selective removal of NO₃^– is a particularly common topic. Chen et al. studied a mixture of Cl^– and NO₃ and showed that Cl^– is preferentially electrosorbed at early times, whereas Cl^– is displaced by NO₃^– in the EDLs at later times.⁷³⁴ The time dependence of selectivity observed by Chen et al. was not observed for a mixture of SO₄ and Cl^–, from which neither anion was preferentially removed despite the difference in charge.⁷³⁴ In contrast, Li et al. later observed dynamic replacement of Cl^– by both NO₃^– and SO₄ in separate experiments, and they suggested that selectivity and capacity are in fact determined by the hydration ratio, namely the ratio of the hydrated radius to the ionic radius.⁷³⁵ This work reported that monovalent ions with lower hydration ratios exhibit greater electrosorption relative to other monovalent ions and that divalent ions are preferentially stored over monovalent ions at equilibrium, which supports the results of Zhao et al.⁶⁴⁸ Nonetheless, there appear to be multiple mechanisms at play for the selective separation of NO₃^– from Cl^–. In addition to the work by Chen et al. and Li et al.,^734,735 Oyarzun et al. modeled the intrinsic selectivity of electrodes treated with cetyltrimethylammonium bromide (CTAB) toward NO₃ using surface group–ion equilibrium constants, and they obtained an observable NO₃^– selectivity factor of 6.5 relative to Cl^–.⁶⁵¹ Using MD simulations, Hawks et al. examined NO₃ selectivity in slit-shaped, subnanometer pores with approximately the same size as the hydrated diameters of NO₃^– and Cl^–. This study attributed the high selectivity to the slit-like shape and low hydration energy of NO₃, which allows its hydration shell to be removed more readily than those of Cl^– and SO₄^2–.⁷¹⁶ Mubita et al. found that the high selectivity of NO₃ relative to Cl^– was maintained even when the pore size exceeded that of the hydrated ion.⁷³⁶

Selectivity between competing ions can be quantified by a separation factor, defined as the ratio of the molar electrosorption of ion i (Γ_i) to that of ion j (Γ_j) scaled by the respective feed concentrations, and :⁶⁴⁹

7

If β_i:j > 1, ion i is preferentially removed relative to ion j. We note that eq 7, as it is written, does not hold for batch mode operation because the feed concentrations vary during charging and discharging. Considering the hydrated radii of the ions studied by Nightingale,⁷³⁷Figure 21 shows that β_i:j for competing monovalent ions, with ion i being the smaller of the two, is typically between one and five. One exception is for NO₃^–, which is larger than both Cl^– and Br^– but is selectively electrosorbed relative to these monatomic anions. For competing divalent (i) and monovalent (j) ions, β_i:j is time dependent and can range from below 1 to more than 10.⁶⁴⁸ Occasionally, selectivity is reported using a relation similar to the scaled retention selectivity introduced in section 3.5.2 for shock ED (see eq 5). The two definitions are related by rewriting the separation factor as provided the system is operated in a single pass; here, c_i and c_j are the concentrations in the desalinated product (see ref (738) for details on sample collection). Unlike in shock ED, selectivity in (membraneless) CDI arises due to the storage of ions in electrode pores. It is therefore preferable in CDI to use eq 7 because it explicitly accounts for ion electrosorption and also because effluent concentrations can vary in time even if pore concentrations reach steady state.

Figure 21.

Separation factor β_i:j calculated using eq 7 in MCDI, CDI with porous carbons, and CDI with intercalation or conversion electrodes for pairs of competing anions and cations. β_i:j > 1 implies selectivity toward ion i, β_i:j < 1 implies selectivity toward ion j, and β_i:j = 1 (dashed line) corresponds to no selectivity. Competing monovalent ions (ion valence 1:1) typically display β_i:j between 1 and 10. Competing divalent and monovalent ions (ion valence 2:1) display a wider range of β_i:j, typically between 0.01 and 24. References for valence 1:1 ion pairs in alphabetical order: Br^–:Cl^–,⁷³⁵ Br^–:F^–,⁷³⁵ Cl^–:F^–,^683,735 K⁺:Li⁺,^688,690 K⁺:Na⁺,^{712,728,735,739} Na⁺:Li⁺,⁶⁸⁸ NH₄⁺:Na⁺,⁷⁴⁰ NO₃:Br^–,⁷³⁵ NO₃^–:Cl^– (MCDI,^678,680 CDI^{651,709,734,735}), and NO₃:F^–.^709,735 References for valence 2:1 ion pairs: Ca²⁺:Li⁺,⁶⁸⁸ Ca²⁺:Na⁺ (MCDI,⁷⁴¹ CDI,^{647,648,712,723} intercalation or conversion^742,743) Mg²⁺:Li⁺,^688,689,744 Mg²⁺:Na⁺,⁷⁴² SO₄^2–:Cl^–,^734,745,746 and SO₄:NO₃^–.⁷⁰⁹

In addition to examining the mechanisms of selective electrosorption, researchers have sought to increase electrode capacity toward a specific ion by using composite carbon electrodes embedded with materials like polymers and metal oxides.^678,680,747 Kim et al. demonstrated this concept by coating the surface of a carbon anode with anion exchange resin that is selective toward NO₃^–, which more than doubled the separation of NO₃ to Cl^– compared to the commercial AEM Neosepta.⁶⁸⁰ By applying constant current to the device, the same group demonstrated even greater separation using the IEMs and composite electrodes reported by Kim and Choi.⁶⁷⁸ Similarly, Tang et al. showed slight preferential removal of SO₄^2– by using an MCDI system operated at constant current.⁷⁴⁸ It is important to distinguish composite electrodes, where materials are affixed directly to carbon, from MCDI, where free-standing membranes are layered onto the electrodes. MCDI can be used for selective removal of ions, as shown in Figure 21, but because ion selectivity in MCDI is primarily determined by the membranes rather than by the electrodes, we do not discuss this method further in this section. Besides using composites to tune selectivity, researchers have explored chemical treatments to functionalize electrode micropores for enhanced selectivity. Guyes et al. oxidized the negative electrode to add carboxylic groups to enhance ion selectivity based on size, which increased the selectivity factor of K⁺ relative to Li⁺ from approximately 1 to 1.84.⁶⁹⁰ In a later study, Guyes et al. used a cell with a sulfonated cathode to increase the selective removal of Na⁺ relative to Ca²⁺ at early charging times and high charging voltages.⁶⁴⁷ Uwayid et al. then published the first demonstration of perfect Ca²⁺ selectivity relative to Na⁺ in a CDI cell using a sulfonated cathode with long charging times and nonzero discharging voltages (i.e., only Ca²⁺ was removed during charging).⁷⁴⁹

The effects of ion size, chemical charge, and affinity can be mathematically modeled using continuum theories. Here, we briefly review the principles of modified Donnan (mD) EDL theory to describe some aspects of selectivity in porous carbon electrodes with micropores larger than the hydrated ions present in solution. In this framework, we spatially average over micropore volume to neglect the local pore structure and simplify the governing equations. We do not cover CDI models based on transport equations, which are needed to describe time-dependent electrosorption and selectivity.^647,750 An alternative yet related framework to the mD model, termed the amphoteric Donnan (amph-D) model, considers the electrodes to comprise separate regions of positive and negative surface charge.⁷⁵¹

The mD model employs the Donnan approximation of spatially constant micropore potential and assumes equilibrium between the macropores and micropores, which results in a Boltzmann distribution for concentration:

8

The terms c_i,k^mi and c_i,k are the micropore and macropore concentrations, respectively, of ion i in electrode k = A (anode) or C (cathode), Δϕ_D,k is the dimensionless difference in potential between the micropores and macropores, Δμ_i,k^ex is the dimensionless difference in excess chemical potential between the micropores and macropores, and z_i is valence; all potentials are scaled by the thermal voltage (see section 3.1.1). For single-pass operation and at cell equilibrium, it is usually assumed that the macropore concentration equals the feed concentration: . The term accounts for nonideal effects, such as an affinity between the electrode surface and specific ions,⁷⁵² image forces acting on ions,⁶⁶⁴ and hard-sphere interactions between ions.^649,690

According to eq 8, higher background (macropore) concentration of a particular ion increases its micropore concentration. In the case of two competing ions with equal valence and excess potential, the ion with higher background concentration will be stored in larger quantities. In a charged cell, Δϕ_D,C is negative and so cations are stored in the cathode, while Δϕ_D,A is positive and so anions are stored in the anode. Because the Donnan potential is multiplied by valence, multivalent (|z_i| > 1) counterions are preferentially stored relative to monovalent (|z_i| = 1) counterions. As captured by the term Δμ_i,k^ex, a larger excess potential reduces ion concentration in micropores. Considering effects of volume exclusion due to finite ion size, a small ion has a lower excess potential and is thus preferentially stored relative to a larger ion with equal valence.⁶⁴⁹ The value of Δμ_i,k is also affected by electrode properties such as the chemical charge of surface groups, which can be tuned to improve selectivity toward target ions.⁶⁹⁰ Moreover, it was recently demonstrated that volume exclusion may enhance selective removal of larger ions with higher valence due to the complex interplay between electrostatic forces and volume-exclusion effects.⁷⁴⁹

4.1.3. Energy Consumption

Energy consumption by CDI has been extensively studied over the past five years.^{90,636,753−761} Recently, Hawks et al. proposed metrics and methodologies to compare energy consumption between different CDI systems and other desalination technologies.⁷³⁸ They suggested that a target separation should be implemented across all systems being compared by specifying feed concentration, c₀, concentration reduction, Δc, and water recovery, γ. The systems can then be compared in terms of volumetric energy consumption (in units of W h m^–3 of treated water) versus productivity, , defined as the throughput of fresh water scaled by the projected cross-sectional area of the electrode and multiplied by the number of cells (see eq 24).⁷³⁸ A related metric to quantify energy demand was used by Lin,⁷⁶² namely the specific energy consumption, SEC, scaled by either the amount of adsorbed salt^759,761

9

or the volume of produced water^738,762

10

where E_ch is energy consumption during charging, M_salt is the molecular weight of the salt, Q is volumetric flow rate, and t_ch is charging time. Another common metric in the literature is energy normalized adsorbed salt (ENAS),^90,706,738 which equals SEC_ion^–1. To ensure repeatable results, energy consumption should be determined from the CDI cycle after the cell has reached a dynamic steady state.⁹⁰ Dynamic steady state is usually reached after 3–5 cycles, after which the system exhibits steady behavior and the amount of salt adsorbed during charging equals the amount released during discharging.

In CDI, energy can be recovered from the discharging step to reduce the total energy consumption.⁷⁵⁶ Energy recovery is often quantified as the ratio of energy recovered during discharging to energy consumed during charging.^763−767 Early studies of energy recovery used either only the consumer unit, a CDI cell,⁷⁶⁶ or only the storage unit, a buck-boost converter,^763−765 and reported recovery ratios of up to 83%. Kang et al. later examined a system comprising both a CDI cell and a buck-boost converter, which achieved recovery ratios of up to 50% and that decreased with lower feed concentrations and faster desalination rates.⁷⁶⁷ Oyarzun et al. emphasized the importance of utilization efficiency, defined as the recovery efficiency over a full cycle, including charging and discharging of the DC–DC converter used for energy storage.⁷⁶⁸ The nature of the power source used to drive the CDI cell is also an important consideration. As demonstrated by Tan et al.,^769−771 photovoltaics are an adequate power source due to the low applied voltage required for CDI cell charging (<1.2 V), which makes CDI a candidate for off-grid desalination.

The energy efficiency of CDI systems is normally reported as a thermodynamic efficiency, η_thermo (see eq 21),^{90,150,757,761,762} defined as the ratio of the theoretical minimum energy needed to achieve a target separation (see eq 19) to the energy consumed in practice. For comparisons of η_thermo to be meaningful, the systems being compared should have the same productivity and realized ion separation.^761,772 Porada et al. presented an MCDI cell with an efficiency of 16.04% at a productivity of 11.8 L h^–1 m^–2 and current density of 18.5 mA m^–2.⁶³⁶ Hemmatifar et al. analyzed a membraneless cell designed to achieve high efficiency with total recovery of the energy released during discharging.⁹⁰ This system achieved an efficiency of 8.89% at a productivity of 3 L h^–1 m^–2 and current density of 4 mA m^–2; increasing productivity to 6 L h^–1 m^–2 and current density to 8 mA m^–2 lowered the efficiency to 6.81%. Ramachandran et al. reported a cell with an efficiency of 8% when operated at constant current and 5% when operated at constant voltage, assuming total energy recovery.⁷⁶⁰ A comprehensive review of CDI efficiencies was presented recently,^761,762 and it reported values of up to approximately 6%.^750,773

Three efficiency indicators have been proposed to study mechanisms of charge and energy loss: Coulombic efficiency, charge efficiency,^{90,738,774−776} and flow efficiency.^{636,657,758,760} Coulombic efficiency quantifies the ratio of charge released during discharging to charge delivered to the cell during charging. Values below one are due to parasitic Faradaic reactions that can occur during charging,⁷⁷⁷ such as oxygen reduction at the cathode and carbon oxidation at the anode (see eq 11). Parasitic reactions, which can degrade salt storage capacity in the electrodes, are more likely to occur when operating at higher charging voltages and for longer times.^{755,756,778−781} Charge efficiency is defined as the ratio of moles of salt removed from the feed to moles of electrons transferred between the electrodes during discharging,^775,782,783 and typical charge efficiencies in CDI systems are in the range of 60–90%.^115,784 Many strategies have been explored to improve charge efficiency, such as using high discharge voltages (with 0.3 V as a minimum value),^756,785 chemically functionalizing electrodes,⁷⁸⁴ or layering IEMs over the electrodes, as in MCDI.^70,786 Flow efficiency is defined as moles of salt removed from the feed divided by moles of salt adsorbed by the electrodes,⁷⁶⁰ and it captures unwanted effects like direct mixing of diluate and concentrate when switching between charging and discharging. Hawks et al. demonstrated that flow efficiency can be improved by adding a high throughput flush step between charging and discharging.⁷⁵⁸

In addition to energy losses caused by electrons participating in side reactions, there are also resistive losses caused by electronic and ionic resistances.^{753,754,761,787} Electronic losses arise mainly from resistances of the external circuit components, contact resistance between current collectors and electrodes, and resistances in the solid phase of the electrodes.^757,761 Electronic losses can be reduced by improving contact between the current collectors and electrodes,⁷⁶¹ which can be achieved by surface treatments of the current collectors and electrodes⁷⁵⁷ or by using conductive epoxies.⁷⁵³ Electronic losses are expected to be greater when the cell is operated at constant voltage, as current can be high at early times.^90,755,756 Ionic losses occur mainly due to the resistance of the electrolyte in the separator and electrode. Common methods to lower these losses are to reduce spacer thickness or increase spacer porosity.⁷⁵⁷ Although operating at constant current is often considered more energy efficient than operating at constant voltage,^755,788 Dykstra et al. showed that the reverse can be true when operating at certain conditions.⁷⁵⁶

Qin et al. recently analyzed brackish water desalination by RO and by CDI, and they concluded that RO is much more energy efficient, with no expected change in the future.^789,790 However, several serious issues with this analysis were later identified, including unrealistic values of electrical resistance, unphysical trends in energy consumption, and inaccurate predictions of performance.⁷⁹¹ Porada et al. also highlighted that the comparison by Qin et al. was done using inconsistent definitions of salt rejection.⁶³⁶ While the common definition given by R_j = 1 – c_D/c₀ (where c_D is diluate concentration) was used for CDI, salt rejection in RO was defined as R_j = 1 – c_D/c_B (where c_B is brine concentration).⁷⁸⁹ Porada et al. thus repeated the analysis with consistent definitions of salt rejection and demonstrated that CDI can indeed be competitive with RO for brackish water desalination, especially when water recovery is high.⁶³⁶

Many applications in desalination and water purification demand high water recovery to reduce the volume of waste generated.^{636,636,757,760,761,789} It was recently shown that water recovery in CDI could be significantly improved, with only a small increase in energy consumption, by decreasing the discharging flow rate.^760,792 For example, Ramachandran et al. found that setting the discharging flow rate to 10% of the charging flow rate was most suitable in their study,⁷⁶⁰ and Tan et al. used a stopped flow discharge process.⁷⁹² Water recovery can also be improved by adjusting the operating parameters, which include charging and discharging times as well as charging and discharging voltages (or currents).⁷⁵⁶ Another method to increase water recovery is by flowing the brine product through the cell as a washing solution during discharging.⁷⁹³

4.1.4. Flow Electrode Capacitive Deionization

Research and commercialization efforts in CDI have grown significantly over the past 10 years, and much of this growth has been catalyzed by the invention of new architectures and designs for CDI systems.^115,794,795 In 2011, Duduta et al.¹²⁷² introduced semisolid lithium-ion flow batteries with flowable intercalation electrodes, a concept that has since been applied to other rechargeable batteries^1273−,1275 as well as CDI. In 2013, Jeon et al. introduced flow electrodes for CDI, which were a suspension of carbon particles that are transported along with the flowing electrolyte, as shown in Figure 20a.⁶³⁸ FCDI offers two major improvements to CDI with static electrodes, namely the ability to operate the system continuously rather than cyclically and an increased capacity for salt storage.¹¹⁵ Transforming CDI into a continuous process requires discharge of the carbon particles (and formation of the brine) downstream, either in a mixing tank or in a second device.^{641,796−803} FCDI has also demonstrated the possibility of desalinating feeds that are more concentrated than what is processed by static electrode CDI.^638,804

FCDI is sometimes used to concentrate inorganic nutrients and organic molecules for later recovery.^{699,700,805−808} Bian et al. introduced FCDI as an alternative to a complex and costly nutrient removal system for the extraction and up-concentration of inorganic nutrients, such as NH₄⁺, NO₃, and PO₄^3–, from wastewater.⁸⁰⁵ Although the system achieved a high recovery of NH₄ (89–99%), NO₃^– (83–99%), and PO₄ (49–91%), it required post-treatment to further separate inorganic nutrients from desalinated salts (e.g., NaCl). Several modifications of FCDI then enabled the recovery of inorganic nutrients, especially phosphorus. For instance, PO₄^3– was selectively recovered as H₃PO₄ by leveraging inevitable side reactions like the generation of H⁺ by water splitting.⁸⁰⁶ In this study, operating the system between 1.5 and 2.1 V resulted in the dissociation of water, while both PO₄ and Cl^– were adsorbed by the positively charged carbon particles. As a result, both H₃PO₄ and H⁺ were produced at the same electrode. During regeneration, the electrode released most of the adsorbed Cl^– back into the original chamber, whereas neutral H₃PO₄ remained in the electrolyte (with a selectivity factor of > 2 relative to Cl^–) to produce a solution rich in phosphorus.⁸⁰⁶ Moreover, Zhang et al. recovered PO₄^3– with magnetic carbon particles (impregnated with Fe₃O₄) that showed strong affinity toward phosphorus.⁷⁰⁰ This study proposed a two-step process for the recovery of highly purified PO₄ by concentrating the ion using FCDI and crystallizing it into vivianite (a hydrated iron phosphate mineral) using a fluidized bed crystallization column. The applications of FCDI have expanded to include recovery of carboxylates, such as acetate and oxalate, downstream of advanced oxidation processes.⁸⁰⁷

In contrast to solid carbon electrodes which have electrical conductivities in excess of 1000 mS cm^–1, flow electrodes have electrical conductivities mostly below 10 mS cm^–1 because a suspension of carbon particles can form a largely discontinuous network for electron transport.^666,809 The electrical conductivity of flow electrodes is a function of the amount of suspended carbon, typically quantified by the weight fraction of carbon in the electrode (i.e., the dry weight of carbon particles divided by the total weight of the electrode). Increasing the fraction of carbon boosts electrical conductivity, although slurries used as flow electrodes become viscous and difficult to pump when the fraction of carbon exceeds about 20 wt %.⁶⁶⁶ On the other hand, low electrical conductivity adversely impacts performance, particularly in the extreme case that the slurry does not percolate electric charge. In this limit, carbon particles through much of the network cannot be charged and do not store salt, and the mechanism of desalination reverts to that of an ED system driven by a steady Faradaic current.^639,810 This functionality is sometimes mistaken for capacitive storage of salt in carbon particles, which has led to spurious reports of unusually high salt adsorption capacities of flow electrodes.^810,811

To overcome the limitations of weakly conductive flow electrodes, researchers have proposed new platforms that increase the loading of suspended carbon. These platforms include fluidized bed electrodes, a combination of these electrodes and slurry particles, conductive chemical additives, and redox electron mediators.^{641,809,812−816} Fluidized bed electrodes are distinct from slurries used as flow electrodes because the former uses large (∼100 μm), porous beads made of carbon that are pumped with the electrolyte vertically upward. The weight of the beads causes them to travel more slowly than the surrounding fluid, which leads to a densely packed electrode where the fraction of carbon can reach over 30 wt %.⁶⁴¹ Although these fluidized beds can be dense, the electrical conductivity often remains less than 1 mS cm^–1.⁸⁰⁹ One way to further increase the loading of carbon beads is to enrich the bed with a slurry of smaller particles, which feel a negligible force due to gravity. Cohen et al. showed that although this slurry itself could have a conductivity greater than 1 mS cm^–1, the conductivity of the combined system would only just reach 1 mS cm^–1.⁸⁰⁹ This result suggests that a more detailed understanding is needed of the relationship between the properties of carbon particles and the conductivity of the electrolyte. Ma et al. found that by adding redox-active quinone to a flow electrode with carbon particles of 100 μm loaded at 1 wt %, the rate of salt removal can be increased in potentiostatic mode.⁸¹² Similarly, other researchers have attempted to increase the conductivity of flow electrodes by introducing conductive additives such as multiwalled CNTs and carbon black.^813,814 Recently, Halfon and Suss demonstrated a system that comprised metal particles under two operating modes: the first was a static mode in which conductivity exceeded 10 000 mS cm^–1, and the second was a flow mode in which the particles were discharged.⁸¹⁷ In the future, this concept may be adapted to FCDI by replacing the metal particles with carbon.

The latest advances in the field of FCDI have focused on scaling up, discovering innovative alternate applications, and better understanding the overall process. Today, FCDI is scaled up by either fabricating a stack with multiple unit cells or using three-dimensional honeycomb cells with multiple channels that are interconnected by porous carbon supports.^818−821 At the same time, researchers have adapted FCDI for capacitive neutralization desalination, double displacement reactions,⁸²² ion separations,^{704,707,823−828} and resource recovery.^699,700 To better understand the process, theoretical models have been proposed,^800,829,830 and the governing transport mechanisms have been studied.⁸³¹ Overall, FCDI has the potential to be used for water purification, but for this field to reach its potential, the community must develop ways to either improve or compensate for the low electrical conductivity of flow electrodes.

4.1.5. Parasitic Faradaic Reactions

Faradaic side reactions in an electrochemical system are often (but not always) undesirable and can limit the performance and stability of the process.^652,832 The two reactions that are believed to have the greatest impact on the performance of CDI are carbon oxidation at the positive electrode and oxygen reduction at the negative electrode:^779,780

11

Managing the oxidative step in the corrosion of carbon plays a key role in the long term stability of a CDI device, as the electrodes are repeatedly charged and discharged. While this reaction can lower Coulombic efficiency, reduce the conductivity of the electrode, and distort its microporous structure, the most severe effect is likely the production of carboxyl surface groups (COO^–).^652,778,782 As the concentration of these surface groups increases in the micropores of the positive electrode, so does the concentration of co-ions to compensate for the increased negative charge.⁷⁸⁴ The result is that a larger fraction of the applied current is consumed during charging to expel co-ions from the positive electrode, which lowers charge efficiency. This phenomenon was observed in the form of “inversion peaks” in the effluent concentration profile during charging after several cycles.^778,784 In addition, X-ray photoelectron spectroscopy revealed that, after cycling, the positive electrode was oxidized to the point that its surface was highly oxygenated,⁷⁷⁸ and measurements of pH indicated asymmetric electrosorption as well as irreversible consumption of charge.⁷⁷⁹ Although salt removal significantly decreases due to corrosion of the positive electrode,⁸³³ charge storage capacitance often remains largely unaffected.⁸³⁴ For a cell with an oxidized cathode, it was found that the anode was relatively stable during cycling, but the cathode degraded due to loss of its negative chemical charge.^690,781 Uwayid et al. recently compared the lifetimes of oxidized cathodes and sulfonated cathodes, and they found that the sulfonic acid groups led to a more stable cell.⁸³⁵

When no measures are taken to mitigate this corrosion, a CDI cell is often limited to 50 cycles before the capacity for salt adsorption exhibits a significant loss,^779,781,836 where FTE CDI cells are reported to degrade faster than FB cells.^640,837 Known strategies to mitigate corrosion include periodically reversing cell polarity to slow the oxidation of carbon,^778,838 optimizing the voltage window to reduce electrode degradation,^839,840 reducing the concentration of dissolved oxygen (DO) by sparging with gaseous nitrogen,⁷⁷⁸ decorating the electrodes with titania to enhance electrochemical reduction and prevent DO from participating in corrosion reactions,⁸³⁶ treating the surfaces of the positive electrode to make it negatively charged (inverted CDI),^838,841,842 and reducing the charging time to inhibit Faradaic reactions.^647,781 Other proposed methods include recovering the electrodes by thermal treatment⁸³⁴ and adding an AEM between the main channel and the positive electrode to both reduce corrosion and extend cycle life.^115,635,843 While the positive electrode may still corrode, its capacitance can be preserved, as the AEM is the active element in this desalination.

By measuring the concentration of DO in the product, researchers discovered that oxygen is reduced at the negative electrode during charging, while the concentration of DO recovers during discharging.^779,780 It was proposed that DO reduction at the cathode contributes to corrosion of the anode carbon because it causes asymmetry in the electric potential⁷⁷⁸ and produces hydrogen peroxide (see eq 11),^640,780,844 which accelerates oxidation of the positive electrode.⁸³⁶ The reduction of oxygen, which dominates when the applied voltage is less than approximately 1 V, can severely diminish Coulombic efficiency under certain conditions.⁷⁷⁹ The main approach to suppress this parasitic reaction is to dilute the DO by displacing it with gaseous nitrogen.⁷⁷⁸ As with corrosion of carbon, the rate at which oxygen is reduced is lowered in MCDI systems⁶³⁵ because the membranes impede transport of DO inside the device.

Besides the two reactions in eq 11, several other parasitic reactions may occur, although they usually have a minor impact on the performance of the cell.^652,686 One such reaction is electrolysis of water, which produces gaseous oxygen at the positive electrode and gaseous hydrogen at the negative electrode. The fact that this reaction is sluggish on carbon electrodes, however, reduces its impact on desalination by CDI. Moreover, electrolysis of water can be prevented by operating the system at or below the standard potential of water electrolysis (1.23 V). Recent studies of Faradaic side reactions in CDI with carbon electrodes have enabled researchers to limit the destructive effects of these reactions.^781,837,839 While the use of IEMs can reduce the impact of side reactions and increase cycle life, future work in this area should focus on delivering stable electrodes without the need for potentially costly membranes.

4.1.6. Fouling in Capacitive Deionization with Porous Carbon Electrodes

CDI devices with porous carbon electrodes are susceptible to fouling, which may either occur on electrode surfaces or block electrode pores. In MCDI devices, fouling can also occur on membrane surfaces or inside the membrane structure. Here, we discuss common types of fouling and their mechanisms in CDI; we limit the discussion to fouling of electrodes, as membrane fouling was explained in section 3.6. Fouling of electrodes by organic matter has received the most attention to date, and researchers generally report that salt capacity is reduced in the presence of organic matter.^{558,726,845−851} Zhang et al. showed that the level of dissolved organic matter can determine the extent of fouling. In particular, treatment of lake water with low levels of organic matter led to little fouling and only a 13% loss in capacity over 2 weeks, whereas treatment of lake water with high levels of organic matter led to a 75% loss in capacity in the same amount of time.⁸⁴⁶ Organic fouling also increases energy consumption due to adsorption of organic species that block active sites on the electrodes and inhibit fluid flow.^846,851 For example, Chen et al. found deposits of alginate gel and humic acid in the bulk electrode structure, and these deposits were more noticeable in the presence of Na⁺ and Ca²⁺.⁸⁵¹ The same authors observed similar deposits on the surfaces of IEMs in MCDI, although the alginate gels formed in the presence of Ca²⁺ were denser than those formed in the presence of Na⁺. Organic foulants can also be adsorbed within carbon pores, although this adsorption may not significantly impact performance. Liu et al. found that while humic acid was adsorbed in both microporous and mesoporous activated carbons, salt capacity decreased by only 5% over 30 cycles.⁸⁴⁸

Mineral scale can form on the electrodes in CDI when ions such as Ca²⁺, Mg²⁺, HCO₃^–, and SO₄ are present in solution. Several studies found that scaling by Ca²⁺ and Mg²⁺ is minor,^845,846 although Mossad et al. observed that Fe³⁺, even at concentrations as low as 2 mg L^–1, formed deposits which reduced salt removal by 7.4% and flow rate by 13% over 30 hours.⁸⁴⁵ Moreover, Pb²⁺ and Cu²⁺ poorly desorb during cell discharging, which indicates that these species can remain bound to the electrode.⁸⁵²

A number of techniques have been studied to control fouling and prolong device lifetime, several of which are similar to the methods discussed in section 3.6. Pretreatment involves removing foulants and scalants upstream of CDI using chemical additives and antiscaling compounds. Often, these compounds are added directly to the feed, although in some cases they may interfere with the electrosorption process. Special electrodes (e.g., TiO₂–RGO nanocomposites⁸⁵³) can also be designed to have intrinsic antifouling properties. Finally, cleaning solutions such as citric acid, NaOH, and HCl are used to regularly remove deposits and restore performance. A more detailed review of methods to control fouling in CDI can be found in ref (854).

4.2. Capacitive Deionization With Intercalation and Conversion Electrodes

4.2.1. Faradaic Electrosorption Involving Electron Transfer Reactions

CDI cells have historically relied on porous carbon electrodes¹¹⁰ because carbons are inexpensive, widely available, and capable of reversible positive and negative polarization across a wide range of potentials.^855−858 Despite these advantages, however, carbon electrodes have several limitations. First, the extent to which a porous carbon can store salt (expressed as the mass of stored material per unit mass of material available for storage, often in mg g^–1) is limited and rarely exceeds 20 mg g^–1 for a typical salt like NaCl. Second, carbon electrodes must be charged to a relatively high voltage of 0.8–1.2 V for the amount of salt stored per unit charge to be meaningful. Third, the anode is unstable when exposed to water with a high concentration of DO, as discussed in section 4.1.5.^837,859 Researchers have thus sought to develop novel materials that can address the shortcomings of porous carbons.^860,861

One class of intercalation materials traps ions between the closely spaced layers of the electrode mainly by electrostatic forces, although these materials often exhibit redox reactions as well. Examples of such materials include the inorganic compounds MXene^{128,862−864} and MoS₂ (shown in Figure 22a,b),^865−868 both of which can be used as a cathode and, to a lesser extent, as an anode in CDI cells. Intercalation in these materials is largely driven by the electrostatic attraction between the ions and electrode, with only limited influence of redox reactions, as signified by wide “box-like” cyclic voltammograms such as the one shown in Figure 22b. MXenes have so far demonstrated electrosorption capacities of NaCl of up to 45 mg g^–1 and charge efficiencies greater than 90%,^862,863 but degradation caused by oxidation and hydrolysis is a limiting factor.⁸⁶⁹ Furthermore, these materials are susceptible to deformation (e.g., swelling) during operation because intercalation and ion exchange can alter the interlayer spacing,⁸⁷⁰ as also observed in porous carbons adsorbing larger ions from room-temperature ionic liquids.⁸⁷¹ In MoS₂, salt capacity and charge efficiency have been shown to increase with increasing salt concentration in the feed, a trend opposite to what is observed in porous carbons.⁸⁶⁶

Figure 22.

Representative intercalation and conversion electrodes used for electrochemical separations. Structural image and cyclic voltammetry of (a,b) MoS₂, (c,d) NMO, and (e,f) Ag/AgCl. (a,b) Reproduced with permission from ref (865). Copyright 2015 Springer Nature. (c) Reproduced with permission from ref (872). Copyright 2015 Royal Society of Chemistry. (d) Reproduced with permission from ref (843). Copyright 2014 Royal Society of Chemistry. (e,f) Reproduced with permission from ref (873). Copyright 2020 Elsevier.

A second class of cells that do not use porous carbons involve asymmetric combinations of an electrostatic insertion cathode with a Faradaic conversion anode, such as Ag/AgCl,^874,875 Bi/BiOCl,⁸⁷⁶ or MnO₂.^877,878 In this design, an important feature is that the metallic phase donates or accepts electrons in a redox reaction where ions are adsorbed from the liquid phase. The most familiar example of this process is the oxidation of Ag with Cl^–, which produces the solid salt AgCl on the surface of the electrode until the Ag is fully converted. As shown in Figure 22e,f, the reversible electron transfer reaction is associated with rate-dependent, symmetrically shifted peaks in the cyclic voltammograms, which are the most familiar waveforms observed in voltammetry.⁸⁷⁹ Another such reaction is that of Bi with Cl^– to form BiOCl according to the chemical equation,⁸⁷⁶

12

and Bi-based electrodes enable high storage of Cl^– (∼80 mg g^–1).⁸⁷⁶ MnO₂ is also used as an anode because of its stability, high capacity for storage, and inhibition of parasitic side reactions,⁸⁷⁷ and the redox reaction in which MnO₂ participates is analogous to that observed for Prussian Blue analogues (PBAs, discussed in more detail below).⁸⁸⁰ Similarly, RuO₂ and TiO₂ increase electrosorption capacity when deposited onto a porous carbon electrode.^881,882 Overall, these inorganic materials are functional and selective, although they are limited in that only the primary atom in the host structure is reactive and that reactions of this kind can influence chemical structure.

Another class of intercalation materials is based on redox-active solid insertion electrodes, which enable ion storage throughout the (typically crystalline) solid bulk of the material.^860,883,884 In this section, redox-active crystalline solids are distinguished from redox-active polymers, which are discussed as a separate class of materials for Faradaic electrosorption in section 4.3. The reactions that enable ion storage in solid insertion electrodes typically require less than 1 V.^{715,843,880,885} Moreover, a high diffusivity of ions within the crystal structure is desirable to facilitate ion insertion into the electrode, which increases the capacity of adsorption and reduces energy consumption.⁸⁸ High rates can also be achieved using nanoparticles with small diffusion lengths, as in most intercalation batteries.^357,886,887 The enhanced salt capacity in redox electrodes was described theoretically by He et al., who generalized the EDL theory used for capacitive CDI electrodes to include redox-active groups.⁸⁸⁸

So far, three major classes of redox-active materials have been explored for use in CDI, namely Na₂Mn₅O₁₀ (NMO, shown in Figure 22c,d),^843,872,889 metal phosphates,^885,890 and PBAs.^99,880,891 PBAs are analogues of Prussian Blue in which a fraction of the iron is replaced by another metal like nickel or copper, and they are widely investigated because of their potential longevity, high specific capacity, and facile intercalation of cations, particularly K⁺.^{715,739,892−896} Other materials have also been introduced in recent years, including V₂O₅,⁸⁹⁷ TiS₂,⁸⁹⁸ and sodium superionic conductors (NASICON), the last of which includes NMO. When subjected to an applied voltage, cations are selectively and reversibly inserted into the negatively charged crystal structure of these materials, whereas anions are electrostatically rejected. As the electrodes are charged more negatively (at increasingly positive cathodic potentials), the redox-active atoms are reduced further and more cations are drawn into the pores, which leads to desalination of the water surrounding the electrodes.⁸⁸ In practice, these materials are often combined with anodes such as Ag/AgCl and carbon to selectively adsorb the anions.^{874,875,899,900}

Hybrid CDI (HCDI), first introduced by Lee et al.,⁸⁴³ demonstrates this idea by pairing a redox-active electrode with a porous carbon anode.^{694,877,881,896,901} Smith and Dmello discovered that redox-active materials can even be used as anodes as long as the two electrodes are separated by an IEM.¹²⁷ In the case of electrodes that selectively capture cations (e.g., NMO, metal phosphates, PBAs), the membrane must be an AEM, which functions as the anion selective surface in the device.^116,127,739 The AEM in this symmetric design partitions the feed into two channels in a way that resembles ED, except the electric field is still cyclic. This mode of operation was mathematically modeled by Singh et al.⁸⁸

Recent advances in electrosorption with intercalation materials have provided opportunities for combined theoretical and experimental studies,^{86,96,127,631,888} which build on earlier models of ion intercalation in porous electrodes of lithium-ion batteries.^{357,777,886,887} Redox-active systems have an electric charge that depends on the applied potential, and this variable charge yields storage capacity and ionic selectivity.^105,902 In 2017, Smith used nickel hexacyanoferrate (NiHCF) and NMO electrodes in an ED cell and developed the first model to account for two-dimensional ion transport in ED with intercalation electrodes.¹¹⁶ In 2018, He et al. proposed the first model to account for the thermodynamics of electrosorption based on variable chemical charge at electrodes whose surfaces were decorated with redox-active species.⁸⁸⁸ This same framework was adapted to generate an equivalent circuit model to predict and validate the electroanalytical performance of intercalation materials.⁹⁰³ He et al. further expanded the theory of electrosorption by redox-active materials by developing equivalent circuit models and a model based on coupled diffusion, convection, and electromigration with surface reaction kinetics.^903,904 Singh et al. also presented a theory for CDI with porous electrodes comprising nanoparticles made of a redox-active intercalation material, and the authors described the dynamics of this system in terms of concentration of the product, distribution of intercalated ions, cell potential, and energy consumption.⁸⁸ These models, however, neglect variations in ion concentration inside the particles and assume fast reactions with quasi-equilibrium adsorption isotherms. Inspired by recent progress in modeling intercalation batteries with multiphase layered materials,^{777,886,887,905,906} it would be interesting to further consider the effects of coupled ion–electron transfer reactions, mechanical deformations, and phase transformations resulting from Faradaic electrosorption in porous electrodes.

Considering the many recent advances in intercalation materials, models, and cell designs summarized here, it is naturally useful to compare their desalination performances to those of traditional porous carbons. In a comparison of nine electrode materials, Pothanamkandathil et al. found that energy demands for NiHCF were lower than those of carbon materials when desalinating a 20 mM feed of NaCl without energy recovery; they were comparable, however, when assuming complete energy recovery.⁹⁰⁷ Metzger et al. analyzed CDI with only intercalation electrodes and found that this CDI module costs approximately 27% of a typical MCDI module and is nearly four times smaller in volume due to the larger capacity of intercalation electrodes.⁹⁰⁸ Liu et al. showed, however, that intercalation electrodes are more susceptible to declines in performance from organic fouling.⁸⁴⁸

4.2.2. The Role of Electron Transfer in Ion Intercalation

It is a common misconception that ion intercalation in solid host materials is a physical electroadsorption process, whenever the ion does not itself participate in redox reactions. Indeed, lithium-ion batteries derive their name from the fact that Li⁺ is inserted reversibly in each electrode without being reduced to metallic lithium, effectively shuttling back and forth during battery cycling like a “rocking chair.” This picture, reminiscent of physical adsorption, might appear to be inconsistent with peaked voltammograms for lithium-ion battery materials, which clearly signify redox reactions, similar to those shown in Figure 22d for intercalation electrodes used in electrochemical separations.

In the context of lithium-ion batteries, the resolution of this paradox began with the discovery of Bai and Bazant in 2014 that ion intercalation can be limited by solid–solid electron transfer reactions that reduce and oxidize nearby transition metal ions in the host crystal.⁹⁰⁹ The authors constructed Tafel plots (logarithm of current versus overpotential) for the high-rate cathode material Li_xFePO₄ in quantitative agreement with those predicted by the quantum mechanical Marcus theory of electron transfer,^777,910,911 specifically the Marcus–Hush–Chidsey formula for electron transfer from a metallic electrode (Figure 23a),^912,913 and they postulated a rate-limiting step of electron transfer from the carbon coating to the iron redox site, Fe²⁺⇌ Fe³⁺ + e^– (Figure 23b). They also showed that the temperature-independent curvature of the Tafel plot (Figure 23a) is controlled by the reorganization energy, λ, of the local “solvent” (dielectric solid environment), which is well approximated by the Marcus estimate for outer-sphere electron transfer:

13

based on the optical and static dielectric constants of the solid, ε_op and ε_s, respectively, as well as the effective radius of the reactant a₀ and the distance for electron transfer d, each set to the Fe–O bond distance assuming direct contact of FePO₄ octahedra with a metallic (and sufficiently thick) carbon coating.

Figure 23.

CIET mechanism for ion intercalation in redox-active solid electrodes. (a) Curved Tafel plots consistent with Marcus–Hush–Chidsey theory, indicating electron transfer limitation in LiFePO₄. (b) Sketch of lithium ion transfer (IT) coupled to electron transfer (ET) from the carbon coating to the nearest iron redox site. (c) Excess chemical potential landscape for CIET, combining the classical ion transfer coordinate with the solvent reorganization coordinate for quantum mechanical electron transfer. (a,b) Adapted with permission from ref (909). Copyright 2014 Nature Research. (c) Adapted with permission from ref (914). Copyright 2021 Elsevier.

The emerging understanding is that both mechanisms occur simultaneously, as ion intercalation in redox-active solids occurs by coupled ion–electron transfer (CIET).⁹¹⁴ In this new picture, classical transfer of an ion over the intercalation energy barrier (dominated by the ion–electron Coulomb energy and the local dielectric response of the interface) facilitates the instantaneous quantum mechanical transfer of an electron tunneling between a localized state in the host crystal and a delocalized state in the electrode (Figure 23c). CIET theory is becoming widely used in modeling lithium-ion batteries,⁸⁸⁷ and it could also be used to guide the design of Faradaic electrosorption systems from microscopic first principles. In contrast to the empirical Butler–Volmer equation, CIET theory predicts a reaction limited current, which can be modified by tuning the electrostatic and dielectric properties of the electrode–electrolyte interface and the electronic structure of the electrode.^914,915 For example, the theory attributes the enhanced intercalation rate in Li_xFePO₄ achieved experimentally by anionic surface charge modification⁹¹⁶ to lowering of the ion-transfer barrier of lithium. CIET theory also reveals the roles of dielectric properties and crystal structure of the intercalation host, including generalizations of eq 13 for the CIET barrier.⁹¹⁴

4.2.3. Ion Selectivity in Intercalation Materials

Unlike the selectivity in capacitive materials described in section 4.1.2, selectivity in redox-active intercalation materials is achieved by insertion of specific ions into the crystalline structure, as shown in Figure 24a. Cyclic voltammetry of NaTi₂(PO₄)₃ and Na₄Mn₉O₁₈ indeed showed nearly one order of magnitude greater capacity for Na⁺ compared to K⁺, Ca²⁺, and Al³⁺. It was also concluded that the rate at which ions were removed from a dilute feed was limited by their transport from the bulk to the surfaces of the electrodes as well as by the concentration of noninserting (or electrochemically inert) ions.

Figure 24.

Overview of methods for water purification involving intercalation and conversion processes by (a) crystalline redox materials and (b) polymeric, single-site redox materials. Adapted with permission from ref (631). Copyright 2016 Wiley-VCH.

In addition to the innovations already described in this section, several research groups developed novel device architectures and designs with the intent of selectively removing specific ions. For example, Pasta et al. created an asymmetric system in which one electrode was made of (redox-active) Na₂Mn₅O₁₀ and the other of (metallic) AgCl to desalinate seawater.⁷¹¹ In contrast to carbon-based electrodes which are slightly selective toward K⁺, this system could selectively remove Na⁺ with a separation factor of approximately 2.8 relative to K⁺. Separation factors of more than ten were also reported for electrodes made of (redox-active) LiMn₂O₄ in electrolytes comprising Li⁺ as well as the competing cations Na⁺, K⁺, Ca²⁺, and Mg²⁺.⁶⁸⁸ Kim et al. reported an even higher separation factor of 16 for Na⁺ relative to K⁺ in the presence of competing ions such as Ca²⁺ and Mg²⁺ by using (redox-active) Na_0.44MnO₂ and AgCl electrodes.⁹²⁷ Among the most extreme separations of Li⁺, however, were those achieved by Trócoli et al.,⁹¹⁷ who reported separation factors of 240 relative to Na⁺ and 320 relative to K⁺ by using (redox-active) LiFePO₄ and AgCl electrodes.^918,919 Recent studies that used PBAs for water desalination showed selectivity factors between 2.2 and 3.4 for K⁺ relative to Na⁺.^715,739,895 Although PBAs are usually selective toward monovalent cations, Singh et al. used the PBA vanadium hexacyanoferrate (VHCF) to preferentially remove Ca²⁺ in the presence of Na⁺ with a selectivity factor of 3.5.⁸⁹⁶

Intercalation materials that are selective toward anions have also been examined in conventional CDI systems for the selective removal of PO₄^3– and F^–. For instance, Hong et al.⁹²⁰ recently introduced ZnAl-layered double hydroxide (LDH, commonly used as an adsorbent for PO₄) to an electrode material that is itself selective toward PO₄^3–.^921,922 LDH was composited with RGO to increase conductivity, and the LDH/RGO composite electrode (with activated carbon as the counter electrode) was used for selective separation of PO₄. This composite electrode is selective toward PO₄^3– because as RGO boosts the conductivity of the composite electrode, the electron density of the transition metal sites in LDH is effectively reduced. The reduced metal sites then allow PO₄ to form inner-sphere complexations via an exchange reaction of ligands between the OH^– groups of the transition metal sites and PO₄^3–. As a result, the LDH/RGO composite electrode is selective toward PO₄ (with a selectivity factor of 6.1), even in the presence of Cl^– at 10 times the concentration of PO₄^3–.⁹²⁰ Similarly, hydroxyapatite (HA) was composited with RGO and used to selectively remove F^–:⁹²³ as RGO boosts the conductivity, the RGO/HA composite electrode forms a region concentrated with anions near the electrode, where ion exchange occurs efficiently between OH^– in the HA metal sites and F^–.^923,924 With this composite electrode, the CDI system exhibited five times greater removal of F^– compared to an equivalent system with activated carbon electrodes when both were applied to an equimolar solution of F^–, Cl^–, and NO₃.⁹²³

4.2.4. Electrochemical Systems Design With Redox Reactions

Various cell architectures have been developed to use intercalation and conversion materials more efficiently. For instance, HCDI (shown in Figure 25a) is designed to increase the extent of desalination by using the Faradaic electrode as a working electrode. Because this design slows desalination, a capacitive carbon electrode is included to mitigate the reduction in the rate of ion removal.^843,885,925 HCDI systems usually comprise intercalation or conversion materials, such as Na₄Mn₉O₁₈⁸⁴³ and Na₂FeP₂O₇,⁸⁸⁵ as the negative electrode to selectively capture Na⁺, along with porous carbon and an AEM as the positive electrode to capture the anions. Na₄Mn₉O₁₈ and Na₂FeP₂O₇ could be categorized as one-dimensional insertion materials that store and diffuse cations, especially Na⁺, in their tunnel structures.^926,927 Compared to a system with ordinary capacitive electrodes (which had a capacity of 13.5 mg g^–1 and a maximum removal rate of 0.048 mg g^–1 s^–1), Lee et al. demonstrated major improvements in the capacity (31.2 mg g^–1) and maximum removal rate (0.066 mg g^–1 s^–1) by using HCDI with an NMO electrode, as shown in Figure 26a.⁸⁴³ The HCDI system comprising Na₂FeP₂O₇ was also compared to an MCDI system based on capacity and rate capability using a CDI Ragone plot (see Figure 26b).⁸⁸⁵ The Ragone plot of the HCDI system showed improvement in the capacity of ion removal by up to 150% relative to the MCDI system, which was attributed to the high capacity of Na₂FeP₂O₇. Moreover, the HCDI system achieved a rate capability comparable to that of its MCDI counterpart.⁸⁸⁵ The hybrid use of electrodes has also been shown to enhance desalination performance in FCDI systems. As illustrated in Figure 25b, Chang et al. used copper hexacyanoferrate (CuHCF, a PBA) as the flow-negative electrode and activated carbon as the flow-positive electrode.⁹²⁸ This system exhibited a higher rate of salt removal (up to 0.12 mg cm^–2 min^–1) and a greater efficiency (up to 96%) compared to a system with only activated carbon electrodes.

Figure 25.

Schematic illustrations of CDI systems that use intercalation and conversion materials as electrodes. Adapted with permission from ref (929). Copyright 2020 Multidisciplinary Digital Publishing Institute.

Figure 26.

Representative results and mechanisms of CDI with intercalation and conversion electrodes. (a) Ion removal capacity and maximum removal rate in HCDI applied to a 10 mM solution of NaCl; the system is operated at 1.2 V for 15 min during the capture step and at −1.2 V for 15 min during the release step. Reproduced with permission from ref (843). Copyright 2014 Royal Society of Chemistry. (b) Ragone plot of HCDI (with Na₂FeP₂O₇) and MCDI for feed concentrations of 10 mM and 100 mM. Reproduced with permission from ref (885). Copyright 2016 Elsevier. (c) Mechanism and (d) performance metrics of RCDI applied to a 600 mM solution of NaCl; the system is operated at a cell voltage of 200 mV. (c) Reproduced with permission from ref (715). Copyright 2017 American Chemical Society. (d) Reproduced with permission from ref (930). Copyright 2019 Royal Society of Chemistry.

Recently, several studies introduced dispersed redox species in various multichannel architectures to further improve performance.^{812,931−934} By using two IEMs to provide different environmental conditions between the feed and electrolyte, the redox reaction can occur without contaminating the feed.^{812,929,931−934} For instance, the use of redox-active species significantly improves charge transfer in FCDI by overcoming some of the limitations of charge transfer between current collectors and particles in the flow electrodes (see Figure 25c; details are provided in section 4.1.4).^812,931 Although many studies have suggested increasing either carbon content or salt concentration to reduce the internal resistance of the electrode channels, the conductivity of these channels could be compromised if the carbon particles were to aggregate.^812,935 Ma et al. thus proposed the use of hydroquinone (H₂Q) as an electron mediator that shuttles electrons from the current collector to the particles in the flow electrodes.⁸¹² As a result, the deionization system exhibited an increase of 131% in the average rate of salt removal compared to the same system in the absence of H₂Q.⁸¹² Kim et al. used redox species as an additional means for salt removal in the multichannel redox system (see Figure 25d).⁹³² In this system, the redox reaction between Fe(CN₆)^3– and Fe(CN₆)^4– further remove ions from the feed to maintain bulk electroneutrality in the electrolyte,⁹²⁹ and the redox species are continuously regenerated as the redox couple circulates the electrodes. Overall, the desalination performance of this system (with a capacity of 67.8 mg g^–1) was improved by more than a factor of three compared to the same system in the absence of the redox species (with a capacity of 20.0 mg g^–1).⁹³²

Another active area of research is on the development of various configurations of battery desalination,^{711,715,873,876,930,936−939} where both capacitive electrodes are replaced by intercalation or conversion materials. This concept, shown in Figure 25e, was first introduced by Pasta et al., who used NMO to capture Na⁺ and Ag to capture Cl^–.⁷¹¹ Following this study, several intercalation and conversion materials have been used in desalination battery systems to overcome the limited capacity of capacitive electrodes. Electrodes used to capture Na⁺ include NMO,⁷¹¹ NaTi₂(PO₄)₃,^876,938 and PBAs,⁹⁴⁰ and electrodes used to capture Cl^– include Ag and Bi. Because conversion electrodes have high theoretical specific capacities, they are typically used as counter electrodes.^{711,876,938,940−942} Conversion reactions, however, can lead to poor electrode stability due to repeated volume expansion and contraction over many cycles.⁹⁴³ Despite this drawback, battery desalination systems have been used to recover Li⁺ using selective electrodes such as LMO⁹⁴¹ and LiFePO₄⁹⁴² because of the structural advantages these materials exhibit toward Li⁺. For instance, the tetrahedral sites of LMO and LiNi_0.5Mn_1.5O₄ are narrowly spaced and enable selective removal of Li⁺ relative to larger cations.^944−946 Although the use of Faradaic materials can significantly increase the capacity of salt removal, they tend to limit the rate at which salt is removed.^117,947 Initial studies showed that CDI with activated carbon permits an applied current greater than 1 mA cm^–2,¹¹⁵ whereas intercalation electrodes limit the applied current to ∼100 μA cm^–2.^115,715,739

Lee et al. recently proposed a multichannel desalination battery, as shown in Figure 25f, to overcome the mass transfer limitations of traditional desalination batteries.⁹³⁹ The multichannel system was designed to have two independent channels (one for both electrodes and one for the feed) by placing an IEM at each interface.^939,948 The two electrodes (NaNiHCF and Ag) were exposed to a concentrated solution (1 M of NaCl) to reduce the resistance of the system, where the high salinity of the electrolyte significantly improved the capacity (52.9 mg g^–1) and removal rate (0.0576 mg g^–1 s^–1).⁹³⁹ Lee et al. also demonstrated the continuous battery desalination system known as “rocking-chair” CDI (RCDI; Figure 25g).⁷¹⁵ This configuration is usually made of two different PBAs, such as NaNiHCF and NaFeHCF, separated by an AEM. During the charging step, the working electrode (NaFeHCF) captures cations, while anions move to the counter electrode (NaNiHCF) to maintain bulk electroneutrality. During the discharging step, the working electrode is regenerated while the counter electrode captures cations, which enables continuous desalination with a removal capacity of 59.9 mg g^–1 (Figure 26c).^715,936,937 In the same manner, the cation selective electrodes can be replaced with anion selective conversion electrodes such as Ag/AgCl for even greater continuous desalination (85⁸⁷³ to 115⁹³⁰ mg g^–1) at a low operating voltage (≈ 200 mV; Figure 26d).^873,930 During oxidation of the Ag electrode, Ag reacts with Cl^– in solution to form AgCl by breaking Ag–Ag bonds. Because both the oxidation and reduction peaks of Ag/AgCl occur at a low cell potential (220 mV versus SHE), this Faradaic process requires only a small input of energy (ranging between 2.5 k_BT ion^–1930 and 10 kJ mol^–1 of salt⁸⁷³). By adapting these cell architectures to existing designs, the desalination performance of conventional systems can be greatly improved using intercalation and conversion electrodes.

4.3. Electrosorption by Redox-Active Polymers

As illustrated in Figure 24, Faradaic redox reactions occur when there is transfer of n electrons at the surface of an electrode that changes the oxidation state of a reactant.^949,950 The equilibrium thermodynamics of these reversible Faradaic reactions are governed by the Nernst equation.^{777,951−953} In this review, the notion of a Faradaic process encompasses reactions at electrodes regardless of the phase in or across which the reactants are moving or are present. For instance, a Faradaic process could be either a redox event at an electrode surface⁸⁶ or a variety of side reactions in which the electrons are transported across the electrolyte (e.g., water splitting, oxygen reduction).^89,954 Faradaic processes, combined with rich redox chemistry, have been extensively studied for a variety of applications, including energy storage,^955,956 bioelectrochemistry,⁹⁵⁷ sensing,^958−960 and electrocatalysis.^961,962 As explained in section 2, nonelectrosorptive electrochemical methods rely on Faradaic reactions, be it through dissolution of a metal electrode or changes in the oxidation state of a dissolved species for plating. In the context of water purification by capacitive electrosorption, electron transfer and redox reactions are common, even without the use of external electroactive species. If left unregulated, these reactions could be parasitic and in turn degrade the electrodes, produce undesired byproducts, and reduce the efficiency of the primary electrochemical process.^150,954

In this section, we focus on electrosorption promoted by redox reactions at active polymer electrodes, which boost energy efficiency and, more importantly, enable molecular selectivity. Materials with intrinsic Faradaic properties have long been studied and used for applications in energy storage.^963,964 Batteries, for example, store electric charge by Faradaic intercalation reactions and by changes in the oxidation state of a crystalline solid subjected to an electric field. These synthetic materials confine the redox reactions to within the solid electrode itself, without loss of electrons or transfer of ions across the electrolyte. Moreover, the surface reactions that occur usually promote specific binding of ions by creating sites of fixed charge that enable noncovalent interactions, such as hydrogen bonding, charge transfer processes, and hydrophobic transitions. From the perspective of electrochemical engineering, Faradaic materials also enable greater control of the window of operating voltages and suppress leakage currents and parasitic side reactions that would otherwise be detrimental to the longevity of the system.⁹⁰² The properties and applications of redox-active polymers are described in the sections below.

4.3.1. Overview of Redox-Active Polymers

Electrochemically tunable redox polymers, defined by the IUPAC as polymers with groups that can be either oxidized or reduced, play a key role as stimuli-responsive materials in many chemical and biochemical applications.⁹⁶⁵ These materials have several properties that can be modulated electrochemically, such as reactivity,⁹⁶⁶ mechanical actuation,⁹⁶⁷ sensing,⁹⁶⁸ and energy storage.⁹⁶⁹ As shown in Figure 27, the redox group can exist either directly on the backbone of the polymer chain, as is the case of polyanilines, polypyrroles, and polyquinones, or in pendant groups like ferrocene.⁹⁷⁰ Redox-active polymers with electroactive units in the main chain are often conjugated⁹⁷¹ and comprise π-bonds that enable semiconductivity. While some of these semiconducting polymers have linear backbones, such as polyacetylene, many have conjugated backbones made of aromatic groups, such as polyaniline (PANI) and polypyrrole (PPy). Both PANI and PPy have been used extensively in electrosorption, particularly for enhanced CDI.^96,972−976 Redox polymers with active units in pendant groups, like pendant-group metallopolymers, are charged by a combination of bounded diffusion and electron hopping (or free diffusion).⁹⁷⁷ Although these materials are less electrically conductive, they exhibit unique electronic properties and charge transfer interactions that make them highly selective toward target ions.^96,102,978

Figure 27.

Chemical structures of representative redox-active and conducting polymers used to study selective electrochemical separations.

Faradaic electrosorption relies on a change in the oxidation state of the electrode by electron transfer, which in turn influences the environment of the material. In a complementary manner, the solvent and immediate environment affect the behavior of redox polymers.^979,980 The oxidation or reduction of a redox polymer, for example, creates a fixed charge that selectively binds counterions by either ion exchange or insertion and release of the ions.^{86,96,979−982} For redox systems that favor oxidation, the following chemical reactions are observed:

14a

14b

The first reaction describes a redox polymer P⁺ (e.g., PPy) in which the weakly bound ion X^– is exchanged with the substituent ion A^n–. The second reaction describes a redox polymer P_n that is initially uncharged but, after undergoing oxidation (e.g., the transition from ferrocene to ferrocenium), can bind the counterion A^n– to a cationic group. In this process of binding, the affinity of the ions to the electrodes influences the electrochemical kinetics of adsorption. Ions that are strongly bound often facilitate electron transfer and shift the formal potentials, which makes the use of tunable redox polymers well suited for selective electrosorption and sensing.^978,983,984 Redox polymers, especially pendant-group metallopolymers, can also modulate electrode potentials, which prevents excursions in potential that would otherwise lead to side reactions. The suppression of parasitic side reactions (e.g., water splitting) improves performance and enables better control of water chemistry, as explained in Figure 28.^102,902,985 In the following sections, we describe recent advances in redox-active polymers for the separation of common pollutants like ions and uncharged contaminants.

Figure 28.

Suppression of side reactions using an asymmetric redox electrode configuration. Adapted with permission from ref (902). Copyright 2017 Royal Society of Chemistry.

4.3.2. Molecular Selectivity of Redox-Active Polymers

The main advantage of molecularly selective technologies is the ability to target a minority species in the presence of excess competing molecules. Redox-active polymers have facilitated the realization of this capability by enabling selective electrosorption and reversible release at fixed electric potentials. Because of their favorable electrochemical kinetics, electroactive polymers have been considered for various applications in analytical sensing.^986−989 The mechanisms of selective ion sorption onto conducting polymers have been studied since the advent of ion exchange voltammetry.^990−992 For example, electroactive polymer films can be used as selective sensors by preconcentrating them with a certain counterion from the contacting solution phase.^993,994 Conducting polymers (e.g., PANI, PPy),^995,996 redox polymers,^{982,997−999} and polymers entrapping redox units^1000−1002 have also been used for ion doping. These materials represent the archetypal systems of modified electrodes used during the development of electroanalytical chemistry.⁹⁵¹ The mechanisms of selectivity vary depending on the chemical structure of the redox monomer, the hydrophobicity of the polymer, and the electrostatic properties of the ions themselves. In general, the combination of controlled electric potentials and tailored molecular selectivity makes electroactive polymers a promising platform for energy efficient purification of water, recovery of valuable resources, and separation of fine chemicals.

Metallopolymers are one class of polymers that display electrochemically switchable properties, possess favorable electron transfer kinetics, and enable a diverse array of synthetic pathways for structural modification and tuning.^1003−1005 Recently, these materials were used as heterogeneous coatings on porous carbon to improve the capacity for energy storage. In particular, the noncovalent conjugation of polyvinylferrocene (denoted as PVF) with CNTs was demonstrated by both organic solution deposition and electrodeposition.^1006,1007 The resulting systems displayed high capacitances (1450 F g^–1) and energy densities (79.5 W h kg^–1).¹⁰⁰⁷ Through similar electrochemical methods, the charged moieties of metallopolymers can be used to capture and release anions efficiently. Films of PVF are oxidized in water over a range of positive potentials near 0.3 V versus Ag/AgCl,¹⁰⁰⁸ which is characteristic of stable electron transfer kinetics. Depending on the heterogeneity of the film, however, the applied voltage needed to fully charge all immobilized ferrocene units can be higher (in the range of 0.6–0.8 V versus Ag/AgCl), as was indicated by the large peak separation of different redox films in various electrolytes.

Redox-based electrosorption is cyclical and operates by electrochemical swings to capture target ions. During adsorption, electrodes become positively charged due to the formation of oxidized sites. Captured ions are then released by reversing the polarity due to the reduction of the redox adsorbent, as demonstrated in Figure 29a. The redox of PVF, for example, has been used to selectively extract organic anions such as carboxylates, phosphonates, and sulfonates from water in the presence of excess ions and without the use of chemical additives.^{96,902,978,1009} Because many organic micropollutants like pesticides (e.g., metolachlor-based pesticide), microplastics, and pharmaceuticals (e.g., ethinyloestradiol contraceptive, propranolol hydrochloride beta blocker) are negatively charged,^1010−1012 electrosorption of negatively charged species is carried out by charging a redox polymer at the anode. In such a process, the functionalized redox electrode is immersed in an electrolyte comprising the minority organic anion (in concentrations ranging from 0.1 to 3 mM) as well as excess competing ions like ClO₄^– and Cl^– (in concentrations exceeding 100 mM). The transformation of ferrocene to ferrocenium results in a positively charged surface, which then electrostatically draws anions in solution to that surface. Systems based on PVF were also found to be stable over numerous cycles, and they generally display molecular selectivity, efficient use of energy, and a high capacity for uptake and storage of charge.⁹⁰² One example, shown in Figure 30a, is a system that paired a PVF anode with a polyanthraquinone (PAQ) cathode, which was effective compared to other systems at separating carboxylate, as quantified by an electrosorption capacity as high as 157 mg of anion per gram of adsorbent.¹⁰⁰⁹Figure 30b reveals a relationship between the amount of oxidized species and the applied potential, and it shows that the utilization efficiency of ferrocene was lower than 100% during charging, which may be attributed to either active sites being inaccessible or redox units being kinetically trapped. In municipal wastewater, trace amounts of micropollutants usually exist in the range from micrograms to below a nanogram per liter.¹⁰¹³ Future work demonstrating the effectiveness of redox-based electrochemical separations toward trace micropollutants will thus provide insight into industrial feasibility.

Figure 29.

Electrochemical control of redox-active polymers for selective electrosorption. (a) Illustration of the concept of redox-based capture and release of a target anion. Reproduced with permission from ref (978). Copyright 2016 Wiley-VCH. (b) Selective separation of chromium at a PVF electrode; the schematic shows both charging and discharging steps as well as the reactions that occur at each counter electrode. (c) Results of adsorption uptake of chromium on PVF working electrodes as a function of potential. Reproduced with permission from ref (1014). Copyright 2018 Nature Research. (d) Calculated selectivity of individual anions (0.5 mM each) relative to a competing anion (20 mM of ClO₄^–) after 3 h of adsorption on PVF-CNT at 0.8 V versus Ag/AgCl. Reproduced with permission from ref (1015). Copyright 2020 Wiley-VCH.

Figure 30.

Operating principles of pseudocapacitive separation of ions. (a) Asymmetric PVF/PAQ system for electroregulated, selective recovery of lithiated carboxylates. (b) Variations in charge and ferrocene utilization efficiency (FUE) as functions of the applied potential. (c) Electrosorption capacity of this system compared to other supercapacitors reported in the literature (denoted by the light-colored symbols). Reproduced with permission from ref (1009). Copyright 2016 American Chemical Society.

As illustrated in Figure 29, redox-active electrodes made of PVF were recently used to selectively extract toxic heavy metal oxyanions from water, namely chromate (CrO₄^2–), arsenate (AsO₄), and vanadate (VO₄^3–).^{102,1014,1015} The concentrations of chromium and arsenic were as low as 100 μg L^–1, and the experiments were performed in the presence of secondary wastewater components and excess competing ions at 200 times the concentration of the target ions. This process displayed high ion uptake (>100 mg of metal per gram of adsorbent; see Figure 29c), fast kinetics (<5 min of equilibrium uptake) at moderate potentials (≈ 0.8 V versus Ag/AgCl), and nearly complete electrochemical reversibility. The redox-based mechanisms of electrosorption for PVF were inferred from its electrochemical behavior, where an increase in the uptake of anions was correlated with an increase in redox charging.¹⁰¹⁴ In situ measurements using transmission electron microscopy (TEM) supported the hypothesis that the insertion of ions into redox films was the main mechanism for selectivity, and there was no evidence of plating or phase change of the transition metal during electrosorption.¹⁰¹⁴ Further analysis revealed that Cr(VI) (hexavalent) was reduced to the less toxic Cr(III) (trivalent) during the release step due to thermodynamic speciation at the electrified surface, favored by regulation of solvent pH at the counter electrode.¹⁰¹⁴ This Faradaic reaction worked to the benefit of water purification because Cr(VI) (often produced in industrial processes) is carcinogenic, unlike Cr(III) which is an essential element in the human diet.^1016−1018

When using metallopolymers for electrosorption, one of the main mechanisms of selectivity is charge transfer. Quantum calculations of electronic structure have pointed to a strong correlation between charge transfer at the redox moieties and the binding energy of the participating anion.^978,1014 For example, the binding energies of Cl^–, ClO₄^–, CrO₄, and HAsO₄^2– with ferrocenium were calculated to be 2.90, 3.47, 5.53, and 4.73 kcal/mol, respectively. These computations were performed using density functional theory (DFT) with corrections for entropic and solvation effects. Analysis of the binding of ferrocenium to anions, with supporting evidence from nuclear magnetic resonance (NMR) spectroscopy, has also underscored the role of cyclopentadienyl ligands on intermolecular binding. Once oxidized, ferrocenium (both the metal and the ligand) becomes a strong acceptor of charge that is selective toward anions. This selectivity was observed in both aqueous and organic solvents (e.g., acetonitrile), particularly during the removal of carboxylates in the presence of competing anions like hexafluorophosphate.^{978,1009,1019,1020} The introduction of selective chemical interactions superimposing electrostatics enables selective separations by functionalized redox-metallopolymer electrodes.

More recently, radical-based redox polymers were also found to be highly effective at molecular separations for water purification. Kim et al. showed that TEMPO-based copolymers are effective at selective adsorption of per- and polyfluoroalkyl substances (PFAS) based on their affinity toward these polymers.¹⁰²¹ PFAS contaminants are highly persistent and difficult to both capture and degrade due to their unique physicochemical properties. In the work by Kim et al., redox copolymers with amine functional groups and redox-active nitroxyl radical groups were used to achieve high selectivity and uptake of various PFAS compounds, with separation factors exceeding 500 relative to Cl^–. Furthermore, the use of BDD as the counter electrode reduced energy consumption and enabled degradation of PFAS, which was selectively adsorbed on the redox polymer electrode, in the electrode regeneration step. By tuning the ratio of amines to radical groups, the electrochemical regeneration and selectivity toward PFAS could be further improved.¹⁰²¹

4.3.3. Redox Separation of Uncharged Pollutants

Redox systems have recently been used to capture and release uncharged contaminants from water via hydrophilic-to-hydrophobic transitions. Many environmental micropollutants are uncharged pharmaceutical and personal care products (PPCPs) that are hazardous to humans and animals.^1022−1025 PPCPs have unique physicochemical properties and are often present in water in trace quantities, which complicates the removal of these contaminants by conventional methods such as adsorption, precipitation, biological treatment, and membrane filtration.^{1024,1026−1028} As shown in Figure 31, Faradaic electrodes have been used to selectively capture uncharged micropollutants based on the electrochemically tunable affinity of these redox-active electrodes to the contaminants.

Figure 31.

Faradaic systems with electrochemically tunable affinity for controlled capture of uncharged contaminants. (a) Redox of a ferrocene copolymer for separation of butanol from water. Reproduced with permission from ref (1029). Copyright 2011 American Chemical Society. (b) Comparison of the mechanisms of selectivity in CDI, RMSS, and ETAS. (c) Multistage electrochemical concentration of uncharged species using the ETAS process. Reproduced with permission from ref (103). Copyright 2018 Royal Society of Chemistry.

Prior work on chemical switching in redox polymers, shown in Figure 31a, revealed the possibility of extracting an uncharged compound such as butanol from water based on its affinity to a copolymer of hydroxybutyl methacrylate (HBMA) and vinylferrocene (VF).¹⁰²⁹ The selectivity of the redox gel to butanol in the reduced state was reported in terms of a separation factor equal to the ratio of the equilibrium distribution coefficients (see eq 15) of butanol and water, which was shown to exceed 5. This selectivity was attributed to the hydrophilic-to-hydrophobic transition and preferential swelling of the solvent within the gel.¹⁰²⁹ The concepts presented in this system served as the basis for electrochemically tunable affinity separations (ETAS), which leverage the redox switchable hydrophobicity of polymers to achieve targeted separations. Figure 31b compares the principles of electrostatic CDI, redox-mediated selective separations (RMSS), and ETAS.¹⁰³ In ETAS, a nanostructured blend of PVF and PPy was used to remove and concentrate a collection of uncharged organic pollutants, including dyes like methyl orange, endocrine disruptors like ethinylestradiol, and toxic chemicals like dichlorophenol and bisphenol. The PVF/PPy electrodes were synthesized by simultaneous electropolymerization of pyrrole and electrodeposition of PVF, which yielded a homogeneous coating of the PVF/PPy system onto a carbon cloth, as shown in Figure 32.^103,1030 These nanostructured hybrid materials also possess supercapacitive properties (>500 F g^–1) superior to those of their individual components (PVF with 79 F g^–1 and PPy with 27.3 F g^–1), without the use of carbon nanomaterials.¹⁰³⁰ Differential selectivity in ETAS was quantified using the distribution coefficient K_D:

15

where Q_e is the adsorption capacity (in mg g^–1) and c_e is the concentration of the target contaminant at equilibrium. The distribution coefficient is a function of the applied electric potential (see Figure 32f) and the physicochemical properties of the contaminants.¹⁰³ As shown in Figure 31c, ETAS was implemented in a multistage cyclic batch process to concentrate target uncharged species by several orders of magnitude. The principle of hydrophobic transitions was taken a step further by using complementary asymmetric electrodes.¹⁰⁴

Figure 32.

Imaging and analytical characterization of PVF/PPy coatings on a carbon cloth. (a) Schematic illustration of the ETAS adsorbent made of PVF/PPy coated on a flexible carbon substrate. Images of the PVF/PPy system using (b) SEM, (c) TEM, (d) energy dispersive X-ray spectroscopy (EDS) N mapping, and (e) EDS Fe mapping. (f) Adsorption isotherms of PVF/PPy at different potentials. Open circles represent experimental data and solid lines are fits based on the Freundlich adsorption model. Reproduced with permission from ref (103). Copyright 2018 Royal Society of Chemistry.

4.3.4. Electrode Design for Faradaic Electrosorption

As explained in section 4.1.5, parasitic reactions compromise the performance of electrochemical systems.^778,1031 These unwanted side reactions lower current efficiency and negatively impact the chemistry of the solution, for example, by producing hydroxides which lower selectivity toward anions. Asymmetric configurations of the electrodes have thus been proposed to mitigate parasitic reactions and improve efficiency. Innovations in this area encompass both Faradaic and non-Faradaic systems and involve the use of asymmetric membranes in CDI configurations,¹⁰³² differential functionalization of the electrodes,^1033,1034 or implementation of asymmetric redox chemistries.^96,902,985 In one such example, the performance of CDI was improved by placing IEMs at the electrodes to improve selective adsorption of counterions and to prevent co-ions from escaping into the bulk, as shown in Figure 33.^668,1032

Within the context of Faradaic systems, symmetric and asymmetric configurations have been proposed to improve charge storage in pseudocapacitors^1035−1037 and batteries.^1038−1040 Similar principles hold with regard to electroseparations: that is, the better the capacity for energy storage, the more ions will be separated and retained. Moreover, systems in which electron transfer occurs at a fixed, well-defined redox potential can help control the voltage window to limit side reactions and inhibit competing Faradaic reactions that result in current leakage.⁹⁰² These desirable properties can be achieved using asymmetric electrodes made of redox-active metallopolymers, as illustrated in Figure 34. Faradaic intercalation in crystalline systems can also be used to improve the efficiency of electrosorption. For example, symmetric electrodes of CuHCF have been used to reduce the amount of energy needed for desalination (down to 0.02 kW h m^–3 for a feed with an initial concentration of 25 mM and a final concentration of 17 mM).⁹³⁷ This system was operated at a much lower cell voltage (i.e., 0.6 V) compared to a standard CDI cell, which reduced parasitic side reactions.⁹³⁷

Figure 34.

Asymmetric Faradaic systems for selective electrochemical separations. (a) Schematic of a Faradaic system with redox-active polymer electrodes, namely a ferrocene metallopolymer (PVF) for the anode and a cobaltocenium metallopolymer for the cathode. (b) Characterization of the electrochemical system and corresponding redox reactions by cyclic voltammetry. Reproduced with permission from ref (902). Copyright 2017 Royal Society of Chemistry.

Asymmetric configurations have the advantage of increasing energy storage at each of the electrodes by promoting the appropriate electron transfer reactions, namely oxidation at the anode during charging and reduction at the cathode. In the case of electrosorption, asymmetric Faradaic designs have been used to capture anions at the anode and cations at the cathode.^{86,96,154,711} One example of redox-active systems that enabled selective electrochemical separations was made of ferrocene (PVF) and cobaltocenium metallopolymers on opposite electrodes, respectively.⁹⁰² The two electrodes operated within the window of water stability and, more importantly, their redox reactions allowed for selective capture of anions from solution and subsequent control of the voltage window with 96% current efficiency. By carrying out these redox reactions at low voltages (i.e., 0.4 and −0.6 V versus Ag/AgCl for the anode and cathode, respectively), the pH of the system was stable and there was significant improvement in the efficiency with which anions were removed from dilute solutions. Furthermore, the selective properties of the PVF system were preserved upon capture of anions by preventing the formation of hydroxides.⁹⁰² In subsequent studies, it was shown that the organometallic cobalt center could be tuned using a tetraphenyl ligand,¹⁰⁴¹ which enabled simultaneous recovery of cations and anions.⁹⁰²

More recently, hexacyanoferrate (HCF) electrodes were combined with redox-active polymers to achieve selective electrochemical separations that can be carried out with extremely small swings in voltage (<0.1 V) by redox matching at the electrodes.⁹⁸⁵ As shown in Figure 35a, both anions and cations were selectively captured by combining an electrode made of PVF with another made of HCF. The redox potential of the cathode was shown to be tunable based on the content of mixed metal (see Figure 35b), which enabled structural control of the window of operating voltages down to ∼0.1 V at a current density of 1 A m^–2 for a two-cell system. This feature also improved anion selectivity and separation efficiency, as demonstrated in Figure 35c,d, which highlights the potential of hybrid redox polymers with crystal electrodes for efficient separations.⁹⁸⁵

Figure 35.

Asymmetric redox system using iron active centers for selective separations. (a) Schematic of the asymmetric Faradaic cell, in which a ferrocene metallopolymer is oxidized at the anode (blue) and an HCF crystal is reduced at the cathode (red). (b) Cyclic voltammetry shows that the HCF electrode can be structurally tuned to control the redox potential. (c) Cell potential during electrosorption and (d) recovery of molybdenum anions for different electrode chemistries. Reproduced with permission from ref (985). Copyright 2020 Wiley-VCH.

In addition to their high efficiency, asymmetric redox systems combine separations and reactions to chemically transform harmful contaminants into benign species. By combining an electrosorption working electrode with an electrocatalytic counter electrode, Kim et al. found that As(III) (trivalent) can be captured and transformed into As(V) (pentavalent), as shown in Figure 36a.¹⁰² The former is a naturally and artificially occurring contaminant in groundwater that is highly toxic, even though it is often present at low concentrations.^1042−1044 Capturing this contaminant in the presence of competing species and transforming it into As(V) is a major challenge in water purification. Because of the affinity of PVF for charge transfer anions, PVF-CNT composite electrodes were used to capture AsO₄^3– and AsO₃ during oxidation of the working electrode.^978,1014 The counter electrode was coated with poly-TEMPO methacrylate (PTMA), which comprises stable nitroxide radicals that undergo oxidation by one-electron transfer. This system efficiently reduced the concentration of a contaminated solution from 100 ppb to below 10 ppb, the recommended maximum total exposure.¹⁰⁴⁵ Moreover, separation factors of over 49 relative to competing Cl^– were achieved during the capture of these oxyanions. Upon release, the PVF electrode liberated As(III), and the counter electrode transformed the anions into As(V) in both batch and flow configurations. By taking advantage of these redox processes, contaminants were simultaneously separated and converted, as described in Figure 36. This dual process reduced energy consumption to 0.45 kW h mol^–1 of converted As and increased energy efficiency by an order of magnitude compared to systems that decouple separations from reactions.¹⁰² This framework is similar to the use of PVF with a platinum counter electrode to capture and transform Cr(VI) into the less harmful Cr(III). In the latter process, however, the redox transformation occurs at a single PVF electrode which adsorbs Cr(VI) during oxidative charging and catalyzes its reduction during discharging, while water splitting occurs at the counter electrode to regulate pH.¹⁰¹⁴ A study recently revealed the role of oxygen in ferrocene-mediated conversion of arsenic, where production of peroxide aided the regeneration of active ferrocenium units.¹⁰⁴⁶ The study also showed how PVF systems can be used for efficient remediation of groundwater contaminated with arsenic.

Figure 36.

Simultaneous capture and conversion of As(III) into As(V) using redox-active electrodes. (a) Schematic of the asymmetric electrochemical system with electrodes made of PVF (electroadsorbent) and PTMA (electrocatalyst). (b) Flow diagram explaining the speciation of arsenic in this system. Reproduced with permission from ref (102). Copyright 2020 Wiley-VCH.

Asymmetric systems have been designed with complementary hydrophilic-to-hydrophobic transitions to increase the capacity of adsorption of uncharged organic contaminants.¹⁰⁴ As an example, one electrode was coated with the PVF/PPy hybrid introduced in section 4.3 and the other with PPy containing the amphiphilic surfactant dioctyl sulfosuccinate (DOSS, known commercially as AOT). When paired, these two electrodes exhibit hydrophilic-to-hydrophobic transitions that are complementary and occur at opposite electrodes. That is, the PVF/PPy electrode becomes hydrophilic when oxidized and hydrophobic when reduced, whereas the PPy/AOT electrode becomes hydrophobic when oxidized and hydrophilic when reduced. As shown in Figure 37, this design enables capture of hydrophobic molecules like Sudan Orange G (SOG), which is released from both loaded electrodes during charging (oxidation at PVF/PPy and reduction at PPy/AOT).¹⁰⁴ The PPy/AOT electrode was developed to achieve superhydrophobic properties and strong π–π interactions with uncharged aromatic species.¹⁰⁴⁷ During electrostatic charging, the surfactant dopants are reoriented to modulate hydrophobicity, an interaction that was explored in detail using DFT and MD simulations.¹⁰⁴⁷ While discharging, the asymmetric system of PVF/PPy and PPy/AOT was found to be nearly 20 times more energy efficient than systems based on activated carbon (12 J g^–1 of adsorbent for the asymmetric redox system and 235 J g^–1 of adsorbent for activated carbon).

Figure 37.

Asymmetric electrochemical system with tunable hydrophobicity. Complementary redox-active electrodes made of (a) PVF/PPy and (c) PPy/AOT nanostructures; scale bars are 10 μm. (b) Schematic of the asymmetric system with electrochemically modulated affinity toward uncharged organic molecules. (d) Adsorption isotherms of SOG for various applied potentials. Filled markers represent experimental data, and solid lines are fits based on the Freundlich adsorption model. Reproduced with permission from ref (104). Copyright 2019 American Chemical Society.

Finally, these asymmetric systems have been redesigned into flow configurations, in which CNTs coated with PVF and PPy (doped with the anionic surfactant dodecylbenzenesulfonate, DBS) were used as the electrodes to selectively remove organic contaminants.¹⁰⁴⁸ This system integrated in situ ultraviolet, conductivity, and pH sensors for online monitoring, as shown in Figure 38, and it displayed a high capacity and rate of salt adsorption compared to CDI systems in the literature.¹⁰⁴⁹ On the basis of dynamic measurements of adsorption and release, the flow platform was also selective toward benzoate, a compound representative of toxic carboxylates, in the presence of competing electrolytes, which were 50 times more concentrated than the target anion. This work demonstrated the feasibility of using asymmetric redox designs as on-site systems for water purification.

Figure 38.

Selective adsorption of organic anions in a flow cell with asymmetric redox-active electrodes. (a) Schematic of the continuous electrosorption system with integrated sensing. (b) Cyclic voltammetry of carbon substrates functionalized with PVF-CNT, PPy-DBS-CNT, and CNT. Reproduced with permission from ref (1048). Copyright 2020 Elsevier.

5. Inverse Methods of Energy Conversion

Generating clean, renewable energy is just as important as producing purified water for environmental sustainability. In this section, we introduce what are known as “inverse methods” of energy conversion and draw parallels between these systems and the methods of water purification discussed above. The electrochemical inverse methods that we describe here include reverse ED, capacitive mixing (CapMix), and battery mixing (BattMix). These technologies are similar in that they retrieve energy from the mixing of two streams between which there is a gradient in concentration (or more precisely in chemical potential), such as river water and seawater. They differ, however, in how they are designed and operated, but as we will show in this section, reverse ED, CapMix, and BattMix are in fact just ED, CDI, and Faradaic electrosorption, respectively, operated in reverse. For completeness, we also review electrokinetic conversion of mechanical energy to electrical energy based on harvesting the streaming current produced by pressure-driven flow in charged microchannels or porous media.

Although the scope of this review is limited to electrochemical methods, we note that there exists a physical, membrane-based inverse method known as pressure-retarded osmosis (PRO), which converts the free energy of mixing into useful electrical power.^1050−1055 In this process, two streams of different concentrations are separated by a semipermeable membrane that allows only water to pass by osmosis. The imbalance in chemical potential across the membrane promotes transport of water from diluate to concentrate. This osmotic flow can be stopped if enough hydraulic pressure is applied to the concentrate. The applied pressure needed to maintain osmotic equilibrium is called osmotic pressure and is determined entirely by the composition of the solution. PRO thus occurs as long as the applied pressure is less than the osmotic pressure, and because water flows against this gradient in hydraulic pressure, the free energy of mixing can be converted into mechanical work.

5.1. Inverse Electrokinetic Methods

5.1.1. Reverse Electrodialysis

In reverse ED, the system is fed dilute and concentrated streams that are partitioned by alternating AEMs and CEMs. This stack is sandwiched between metal electrodes that are connected through an external load to which power is delivered, as shown in Figure 39.^1056,1057 Transport of cations across the CEMs and anions across the AEMs produces a flux of electric charge between the electrodes. The magnitude of the resulting electric potential is related to the difference in chemical potentials of the salt in adjacent streams, and the voltages across each membrane are additive.^1056−1058 Natural runoffs and rivers in coastal areas are abundant sources of both dilute feed that is usually left to mix with seawater and energy that is produced without capture.¹⁰⁵⁹ It was estimated that 2.3 MJ of work could theoretically be extracted from each cubic meter of river water that flows into the sea.^{1056,1059,1060} With the typical discharges of all rivers combined, the net power produced by this unconventional source of energy was estimated to be between 2.4 and 2.6 TW.^1056,1061 Of course, not all of this energy could be harvested by reverse ED, but the question is exactly how much can be extracted? Forgacs conservatively predicted a yield of 0.35 MJ per cubic meter of river water,¹⁰⁶² Audinos arrived at a similar yield and calculated an energy recovery of 21% (excluding losses due to pumping and power inversion),¹⁰⁶³ and Jagur-Grodzinski and Kramer experimentally determined a yield between 0.25 and 0.6 MJ per cubic meter of river water.¹⁰⁶⁴ Długołȩcki et al. then derived an equation to calculate the maximum power output based on the properties of the system and its components:¹⁰⁶⁵

16

where N is the number of membrane pairs, A is effective membrane area, α_avg is average membrane permselectivity, a is thermodynamic activity, is membrane resistance (in Ω m²), w is compartment width, κ is electrolyte conductivity, and subscripts “c” and “d” represent the concentrate and diluate, respectively.

Figure 39.

Operating principles and empirical results of reverse ED. (a) Dilute and concentrated streams are passed through a stack of alternating AEMs and CEMs. Differences in the chemical potentials of adjacent streams generate an electric potential across each membrane, and the total voltage is the sum of the potential differences across each membrane. In most reverse ED systems, reversible redox couples (e.g., Fe²⁺/Fe³⁺, [Fe(CN)₆]^4–/[Fe(CN)₆]^3–) in a supporting electrolyte (e.g., NaCl–HCl) are used as electrode streams to convert ion flux into electrical current.^1066−1068 (b) Power density versus discharge salinity for various pairs of low salinity (LS) and high salinity (HS) feeds. Reproduced with permission from ref (1069). Copyright 2018 Elsevier.

Early studies of reverse ED showed that little energy is recovered relative to what is theoretically available. Post et al. later investigated how much energy can be recovered in reverse ED by distinguishing between internal and external losses of energy and by focusing exclusively on the former¹⁰⁵⁹ (external losses arise due to pumping and power inversion and can only be quantified in a full-scale, optimized system). In the literature, it is explained that internal losses arise due to ionic shortcut currents,¹⁰⁷⁰ internal resistances (e.g., friction losses, pumping requirements, electrochemical overpotentials),^{1064,1071,1072} concentration polarization,^1056,1073 and osmotic transport.¹⁰⁷⁴ Internal losses must therefore be reduced to improve energy efficiency, which can be achieved by developing improved electrode systems (to lower electrochemical overpotentials),¹⁰⁷⁵ introducing air bubbles to the feed (to decrease ionic shortcut currents),^1059,1070 improving stack hydrodynamics and spacer design (to reduce concentration polarization),¹⁰⁷³ and designing spiral wound modules (to decrease membrane area).^1076,1077 Using some of these principles, Post and Veerman et al. designed systems that, in the best case, recovered more than 80% of the work made available by mixing artificial river water and seawater.^1059,1070 Several other studies¹⁰⁶⁹ in the past decade also focused on optimizing the performance of reverse ED systems by modeling and developing new IEMs,^1078−1081 spacers,¹⁰⁸² and electrodes.¹⁰⁸³ At the same time, researchers focused on improving the overall operation of the stack,^1084,1085 understanding the impact of fouling and mitigating it (see section 3.6),^1086−1089 and combining reverse ED with desalination technologies.^1090−1095

The fastest growing areas of research over the past five years, however, are related to applications for energy storage,^1096,1097 developing micro- and nanofluidic reverse ED systems,^1098,1099 and pilot studies for commercial use.^1100−1103 Pilot studies are essential to bridge the gap between experimentation at the laboratory scale and implementation at the industrial scale. So far, pilot plants for reverse ED exist in only a few countries, including The Netherlands and Italy. The first of these plants was commissioned in 2014 in Afsluitdijk, a major dam in The Netherlands, and it features a dyke that is 32 km long which separates the IJssel Lake and the Wadden Sea.¹⁰⁶⁹ This plant produces electrical energy (the amount of which has not been made available to the public) from controlled mixing of fresh water with seawater. The pilot plant in Italy was commissioned near the Ettore e Infersa Saltworks in Trapani, and it mixes saturated brine from the saltworks with brackish water from a nearby shoreline well to produce an average of 0.8 W m^–2 of membrane in total.^{1069,1100,1101} Future work in this area will largely focus on improving power density, performing complete cost analyses, and building plants with greater capacity.^1057,1074

5.1.2. Electrokinetic Energy Conversion

Although this review focuses on the interplay of electrical and chemical energy, we discuss here how linear electrokinetic phenomena (see section 3.3) can also be exploited to directly convert mechanical energy to electrical energy via the forced motion of liquid electrolytes in charged microchannels and pores. In 1964, Osterle introduced the basic idea of mechanical electrokinetic energy conversion by harvesting the streaming current produced by pressure-driven flow in a charged capillary.¹¹⁰⁴ Morrison and Osterle then estimated a maximum efficiency of 0.9% for ultrafine water–glass capillaries.¹¹⁰⁵ Gross and Osterle comprehensively analyzed energy conversion by linear electrokinetic phenomena (recently extended by Peters et al.⁵⁰⁴) and estimated a maximum electroosmotic conversion efficiency of 4%.⁵⁰⁵ This approach had attracted relatively little attention until its rediscovery in the context of microfluidics.¹¹⁰⁶ The possibility of measuring and controlling the streaming potential in single nanochannels renewed interest in advancing this classical method through nanoscale engineering.^1107−1110 Indeed, van der Heyden et al. predicted theoretical conversion efficiencies of up to 12%¹¹⁰⁸ and experimentally demonstrated 3% efficiency in 75 nm-thick silica glass nanochannels using dilute KCl electrolytes.¹¹⁰⁹ Considering conduction through the Stern layer can further reduce the theoretically predicted efficiency.¹¹¹¹ Recent advances with soft, flexible microchannels have allowed for practical applications of electrokinetic energy conversion in wearable and portable self-powered devices.^1112,1113

Significant developments have also followed from theoretical predictions of massive enhancement of electrokinetic phenomena for charged surfaces with hydrodynamic slip.^1114−1116 Ren and Stein predicted that hydrodynamic slip could increase nanofluidic energy conversion up to 40% for hydrophobic surfaces with slip lengths on the order of tens of nanometers and up to 70% in CNTs or graphitic systems based on reported apparent slip lengths.¹¹¹⁷ These theoretical values have proven difficult to achieve in electrokinetic energy conversion with homogeneous solid–liquid interfaces, and there are lingering signs of knowledge gaps in our understanding of electrokinetic phenomena under nanoconfinement.⁴⁹⁰ For example, the apparent slip observed for CNTs is absent in crystallographically similar boron nitride nanotubes with different electronic structures.⁴⁹⁶ As a possible alternative to homogeneous surfaces, Bagha et al.¹¹¹⁸ predicted that micropatterned, superhydrophobic surfaces with high slip and charged liquid–gas interfaces could increase efficiency at larger length scales, and this concept remains a subject of research.^1119−1121

5.1.3. Reverse Electrowetting

Electrocapillary phenomena involve interactions between liquid electrolyte interfaces and charged solid surfaces or electrodes. The classical example is electrowetting, where an applied voltage changes the surface tension of a droplet via energy stored in EDL capacitors and thus manipulates the shape of an electrolyte droplet.^1122,1123 Krupenkin and Taylor first proposed the inverse effect of reverse electrowetting to convert mechanical energy to electrical energy, where immiscible electrolyte droplets were compressed by the motion of parallel-plate electrodes.¹¹²⁴ Kolomeisky and Kornyshev suggested an alternative approach based on the reversible penetration of a liquid electrolyte into a solvophobic porous electrode, which could be used in a “double layer shoe” to harvest electricity from walking motion.¹¹²⁵ Analogous to the use of moving wire electrodes in CDI for desalination,⁸⁵⁵ capacitive rotors have been proposed to generate alternating current from mechanical motion by exploiting a time-varying EDL capacitance.¹¹²⁶

5.2. Inverse Electrosorption Methods

5.2.1. CapMix

In 2009, Brogioli proposed a method (later called “capacitive double layer expansion,” or CDLE) to extract energy from the concentration gradient between two feeds, as shown in Figure 40.¹¹²⁷ Unlike reverse ED and PRO, CDLE was founded on the concept of an electrochemical capacitor (obtained by immersing two electrodes in an electrolyte),^{151,1128,1129} which is the basis of water purification in CDI technologies.¹¹⁰ CDLE can therefore be described as CDI operated in reverse: current is applied to the capacitor when filled with salt water, co-ions are repelled and counterions are attracted at each electrode, and EDLs are formed to store electric charge. After the EDLs are charged (an input of energy) the cell is flushed with fresh water to increase the electric potential, and in turn the electrostatic energy, of the system. The capacitor is then discharged (an output of energy) after which the cell is flushed with salt water to decrease the electric potential for a new cycle. This four-step cycle is demonstrated in Figure 40, and it reveals the possibility of extracting surplus energy due to a concentration gradient in the capacitor.

Figure 40.

Schematic and operating procedure of CDLE, the original CapMix system. (a) The device comprises two electrodes made of porous activated carbon, and this pair of electrodes behaves as a capacitor that can be charged and discharged. Two reservoirs contain solutions with different concentrations that are pumped to the cell where they are mixed. (b) Graphical explanation of the four-step operating procedure of CDLE. In phase A, the electrical circuit is closed and the cell is charged; in phase B, the circuit is opened and the cell is flushed with fresh water to increase the electric potential; in phase C, the circuit is closed and the cell is discharged; and in phase D, the circuit is opened and the cell is flushed with salt water to decrease the electric potential. Reproduced with permission from ref (1127). Copyright 2009 American Physical Society.

Brogioli explained the dynamics of EDLs in CDLE using the GCS theory^{501−503,1130} and deduced that this process can achieve an energy output of 0.44 kW h per cubic meter of fresh water. By cycling the system at a rate of 1 Hz, the power output of CDLE would be comparable to that of membrane technologies for energy production. In 2011, Brogioli et al. assessed the performance of CDLE in a larger prototype and observed an efficiency of up to 20 times what was reported originally in ref (1127).¹¹³¹ To analyze the dynamics of CDLE and improve subsequent designs of these prototypes, Rica et al. mathematically modeled the electrodiffusion of ions as well as their adsorption and desorption in porous electrodes.^1132,1133 These models, which accounted for adsorption of both charge and salt, were based on a macroscopic formulation of ion diffusion and Ohm’s law in porous media.

The original design by Brogioli had intrinsic technical deficiencies in that it was sensitive to impurities like DO and exhibited self-discharge.¹¹³⁴ Several research groups improved the efficiency of energy conversion by introducing IEMs to the supercapacitor flow cell.^1135−1139 This system, which came to be known as capacitive Donnan potential (CDP), combined elements of reverse ED and Brogioli’s capacitive method to both produce an electric potential across an IEM and store charge in the EDLs of porous electrodes.¹¹⁴⁰ The method of Sales et al. was based on a fixed external resistance, and their work was followed by a study in which Liu et al. supplemented the charging cycle using an external power supply to ultimately draw more power.¹¹⁴¹ Shortly afterward, Brogioli et al. discovered that activated carbon behaves as a polarizable electrode on short time scales but reverts to its spontaneous (self) potential on long time scales, which limits the maximum power output of CapMix systems that use activated carbon as electrodes.^1142,1143 The authors attributed this effect to leakage of stored charge caused by unwanted electrochemical reactions.

To overcome the limitations of traditional carbon electrodes, researchers examined new geometries and arrangements for both the electrodes and IEMs, which enabled continuous operation^666,1144 and increased power output.^1145,1146 These advancements were accompanied by the use of novel materials for the electrodes, such as porous activated carbon coated with polyelectrolytes, to reduce the leakage current.^1147−1151 These so-called “soft electrodes” generate electrical energy in response to changes in both the capacitance of the EDL and the Donnan potential of the polyelectrolyte coat during exchange of electrolytes.

5.2.2. BattMix

In 2011, La Mantia et al. introduced a novel electrochemical system called the “mixing entropy battery” that extracted energy from a concentration gradient but then stored the energy in the bulk crystal structure of the electrodes.⁸⁸⁹ The novelty of this BattMix system was in the use of a redox-active material, namely NMO, for the cathode (with Ag/AgCl for the anode), which improved the performance and efficiency of energy harvesting compared to supercapacitor electrodes made of activated carbon.¹¹⁵² The BattMix cell is first charged by applying an electric field that electrochemically transports Na⁺ and Cl^– from the corresponding electrode into a dilute electrolyte. The cell is then discharged by exchanging this electrolyte for a concentrated one, which drives Na⁺ to intercalate into the MnO₂ cathode and Ag to be oxidized to Ag⁺ at the anode. Because the energy produced during discharging is greater than that expended during charging, net-positive energy can be harvested from this controlled mixing of two electrolytes. Following La Mantia et al., the scientific community developed and tested other redox-active materials (see section 4.2), such as PBAs,^1153−1157 inorganic compounds of metals,^1158−1160 and copper ammonia redox couples,^1161,1162 as well as conventional battery electrodes, such as zinc and silver.^1163,1164

Like capacitive electrodes in CapMix, redox-active electrodes in BattMix are electrical accumulators that can store and release charge, and these two methods are broadly classified as accumulator mixing technologies.¹¹⁶⁵ Marino et al. examined the electrochemical kinetics of these technologies and determined that supercapacitors (CapMix) can deliver a larger specific power than batteries (BattMix).¹¹⁶⁶ The authors also observed that the kinetics of capacitive electrodes are controlled by the diffusion of ions in the electrolyte,¹¹³³ whereas the kinetics of redox-active electrodes are controlled by the diffusion of intercalated sodium.¹¹⁶⁷ To limit the shortcomings of each class of materials, Lee et al. designed a hybrid accumulator mixing system that comprised a battery electrode (NMO), a capacitive electrode (activated carbon), and an AEM.¹¹⁶⁸ This system extracted an amount of energy per unit membrane area three times greater than that achieved by previous CapMix designs, including those with IEMs. Tan and Zhu also built a hybrid accumulator mixing system using CuHCF for the cathode and Bi/BiOCl for the anode.¹¹⁴⁹ Instead of including an IEM, however, Tan and Zhu coated the electrodes with a polyelectrolyte material to boost power output. Future work in the area of electrochemical energy harvesting includes creating better and more cost-effective materials, improving device configuration and assembly, and carrying out pilot studies.¹¹⁶⁹

6. Performance Comparisons and Process Intensification

Now that we have presented and discussed the basic principles of emerging electrochemical methods for water purification, we compare the performance of these technologies by quantifying energy demand, energy efficiency, and performance trade-offs. A fair and comprehensive comparison of these metrics is generally difficult, and only a few studies have been published which mainly compare one electrochemical method (typically ED or CDI) with RO. In this section, we expand the existing work on performance comparisons by quantifying similarities and differences between the electrochemical methods discussed in this review. From these comparisons, we highlight the kinds of applications where each technology could provide the greatest benefit. Finally, we conclude this section by discussing pathways and challenges associated with scale up as well as opportunities for process intensification by combining electrochemical methods to achieve unique capabilities.

6.1. Energy Demand

Electrochemical methods have the potential to lower the cost of processing brackish water (<15 g of total dissolved solids per liter) as well as feeds that are dilute (<1 g of total dissolved solids per liter) but contaminated with hazardous substances.^123,1170 Unlike conventional desalination methods (e.g., distillation, RO), electrochemical processes are well suited for these tasks because their demand for electrical energy is often proportional to the amount of salt removed (for instance by electrokinetics or electrosorption) and not the volume of water treated. To demonstrate how this distinction influences energy demand, Hemmatifar¹¹⁷¹ compared RO with a generic electrochemical process (EC) based on the analyses in refs (1170 and 1172). RO consumes energy, Ê_RO (per unit volume of permeate), primarily in the form of mechanical work needed to overcome of the osmotic pressure of the feed:

17

where ΔP and Δπ are the differences in hydrodynamic and osmotic pressures, respectively, and γ is water recovery. The hydrodynamic pressure difference is related to productivity, (volumetric throughput normalized by active surface area, in units of L h^–1 m^–2), and membrane permeability, L_p (in units of m s^–1 Pa^–1), by Fick’s law: . The osmotic pressure difference is related to the difference in salt concentration, Δc, across the membrane by the van’t Hoff equation: Δπ = iRTΔc, where i (≈ 2) is the van’t Hoff factor.

In contrast to RO, electrochemical processes consume electrical energy to transport ions and activate electrochemical reactions. These processes are subject to losses in the form of overpotentials, which represent energy lost as heat to the surroundings. Overpotentials in electrochemical systems are grouped into three categories, namely activation (the potential difference above the equilibrium value needed to produce a current in a redox event), concentration (the potential difference across a diffusion layer near the electrode), and resistance (the potential difference caused by resistances in conducting ions and electrons).¹¹⁷³ Hemmatifar showed that the energy demand, Ê_EC (per unit volume of diluate), can be approximated based on ionic and Ohmic (resistive) losses:⁹⁰

18

where I is current (assumed to be constant under galvanostatic operation) and is the equivalent resistance of ionic and Ohmic losses in series. This approximation holds unless the separation is slow and parasitic reactions dominate. Electrical current is related to the change in concentration as I ≈ FQΔc/Λ, where Λ is cycle efficiency. The quantity Λ equals the ratio of salt removed to electric charge fed (both in moles), and it is used to empirically account for coupled inefficiencies due to charge transfer, EDLs, and fluid flow.^1171,1174Figure 2 presents estimates of the volumetric energy consumed by RO and EC using the equations above and the parameters in ref (1171).

These results confirm that although the energy required by RO is a linear function of concentration when the feed is salty (and Δπ is large compared to ΔP), it becomes nearly independent of concentration in the dilute limit. Thus, for feeds that would be classified as either brackish water or fresh water, Ê_RO is bounded by the mechanical work needed to sustain the hydrodynamic pressure, which is independent of concentration. The energy demand of EC, on the other hand, scales approximately as (Δc)² over the entire range of concentrations, which implies that electrochemical processes should in theory be significantly less energy intensive than RO for desalination of dilute feeds. Although the estimates in this analysis are based on a specific set of parameters (e.g., water recovery, productivity, cycle efficiency), the general trend in energy consumption can be generalized to any electrochemical method that removes solute from solvent. Electrochemical processes thus offer a unique opportunity to efficiently purify brackish water and other dilute but contaminated feeds.

6.2. Energy Efficiency

Although we have shown that electrochemical methods should theoretically be efficient in the dilute limit, the energy demand in practice has so far been tens or hundreds of times the thermodynamic minimum (see Figures 2 and 42). These values correspond to thermodynamic efficiencies, η_thermo (ratio of the free energy of separation to the input of electrical work), of only a few percent, which reveals the extent of dissipation (or energy loss) in state-of-the-art electrochemical systems. This gap between the thermodynamic limit and real efficiencies, however, is expected for emerging methods and presents an opportunity for the improvement of electrochemical technologies. In the early days of RO, for example, thermal distillation used to be the state-of-the-art method to desalinate seawater. But from the 1970s to 2008, the consumption of electrical power by RO systems was reduced by almost 90%, from approximately 16 to 1.8 kW h m^–3, which is near the theoretical minimum energy of 1.1 kW h m^–3 needed to desalinate seawater with a water recovery of 50%.⁶⁸ These improvements resulted in the widespread implementation of RO, which today is considered the leading process for seawater desalination based on installed capacity and annual growth.¹²

Figure 42.

Thermodynamic energy efficiency as a function of productivity for various electrochemical methods performing laboratory-scale desalination of brackish and dilute water. Representative data are presented for EDI (squares),^{623,1184,1185} ED (diamonds),^{342,1186−1189} shock ED (circles),^87,584 and CDI with either carbon electrodes (stars)^636,757 or intercalation electrodes (triangles).^715,907,937 The color and size of each data marker represent feed concentration and the ratio c₀/c_D (see eq 19), respectively. The dashed line represents an (approximate) empirical efficiency limit and is given by , where is in units of L h^–1 m^–2.

Dissipation in electrochemical systems leads to inefficiency and may be caused by the dynamics of electrokinetics and electrosorption, ionic and Ohmic losses, parasitic losses due to charge transfer, hydrodynamic dispersion and fluid mixing, or mechanical work required to pump the electrolyte. We emphasize that all electrochemical processes exhibit a complex trade-off between kinetics (measured in terms of productivity) and energetics.^{90,738,757,759} As predicted by the simple model in eq 18, a fast process (one that is operated at high current) consumes more energy but achieves the same level of separation as a slow process,⁷⁵⁹ although this fact holds for all desalination processes because more entropy is generated when operating farther from equilibrium.^72,1175 Similarly, a slow process may be inefficient because of the dominance of parasitic side reactions,^90,757 and so it is critical to understand these trade-offs and quantify the kinetics and energetics of the separation method under consideration.⁷³⁸

With these general guidelines established, researchers have sought to develop efficient electrochemical processes by engineering improved systems, materials, and operating conditions. Systems to improve energy efficiency were designed, such as cyclic recovery of stored energy, reduction of series resistances,^753,757 and use of membranes in electrosorption methods.^750,1176 Materials to suppress parasitic reactions were developed, including redox-active electrodes⁹⁶ and intercalation materials,^116,127,937 and operating parameters were tuned, like flow rate, current, and voltage.^{750,755−757,760,1177,1178} The use of IEMs directly adjacent to the electrodes in electrosorption was first reported by Andelman and Walker,¹¹⁷⁹ and this approach was since pursued by many researchers.^{634,750,1176,1180,1181} The selectivity introduced by these membranes increases charge efficiency by allowing for voltage reversal and increasing the capacity of adsorption.¹¹⁸¹ One disadvantage of using IEMs, however, is that they can incur additional capital costs that exceed those of all other components in an electrochemical device.⁷³⁸

To suppress parasitic side reactions, which often severely compromise efficiency,⁹⁰ researchers have developed and used redox-active electrodes that guide the transfer of electrons toward redox groups.⁹⁶ For example, organometallic and metallopolymeric electrodes were used in an asymmetric Faradaic CDI system and led to no appreciable changes in the pH of the treated solution, which indicated that water reactions were indeed suppressed.⁹⁰² The authors also demonstrated stability of pH when the system was operated in batch mode before complete oxidation of the electrodes. A second class of materials that has been developed to control parasitic side reactions are intercalation electrodes.⁹³⁷ For instance, Kim et al. developed a battery desalination system using identical sodium intercalation electrodes, namely CuHCF, with one or more IEMs to partition successive channels of diluate and concentrate. A triple-stack device with five IEMs showed a 10-fold reduction in energy consumption compared to an MCDI device operated under the same conditions. Again, these improvements in performance must be weighed against the capital cost incurred by using additional membranes.

The third standard approach to improve energy efficiency is to optimize the operating conditions of the process. One example is reported by Hemmatifar et al., who showed that careful selection of the operating conditions in CDI can increase the thermodynamic efficiency up to 9%.⁷⁵⁷ These high efficiencies were achieved by operating the system at constant current, recovering energy during discharging, balancing the rates of charge transfer and fluid transport (which represent the rates of adsorption and ion advection, respectively), limiting the window of voltages to between 0.4 and 1 V (higher voltages trigger side reactions and smaller ones diminish charge efficiency), and lowering series resistances by using current collectors made of titanium mesh separated by thin (30 μm) spacers. In another study, Ramachandran et al. proposed the use of alternating electrical current and successfully removed much of the salt fed without compromising energy efficiency.¹¹⁷⁸ The alternating current was sustained at the intrinsic resonant frequency of the system (equivalent to the time constant of an RC circuit), which was shown to be inversely proportional to the geometric mean of residence time and charging time. In a similar way to electrical current, the flow profile can be better controlled to improve performance, particularly in systems that are cyclic like CDI and Faradaic electrosorption. The main issue that arises in systems with uncontrolled flow profiles is hydrodynamic dispersion, which inadvertently mixes diluate (produced during adsorption) and concentrate (produced during desorption). A recent study showed that hydrodynamic dispersion (a phenomenon that increases mass diffusivity by an amount proportional to the square of the average fluid velocity) can be inhibited by reducing the flow rate during the discharging step.⁷⁶⁰ Using this approach, water recovery was increased to about 90% (compared to 50% with the flow rate fixed) without compromising desalination, productivity, or energy efficiency. In fact, the increase in water recovery enhanced the thermodynamic efficiency by up to a factor of three. Management of electrical input and fluid flow can thus improve existing electrochemical systems at minimal cost.

6.3. Performance Trade-offs

Our discussion in the previous section would be incomplete without quantifying the performance of electrochemical methods because there are underlying trade-offs between kinetics and energetics in any separation process. For this reason, the desalination community uses productivity and energy consumption (per unit volume of diluate) as the two core measures to both quantify performance^{126,1182,1183} and estimate the capital and operating costs of a plant.^738,759 In this section, we examine the relationship between these two parameters and how they collectively influence thermodynamic efficiency.

Constraining the extent of desalination, productivity and (volumetric) energy consumption reveal complex coupling, as shown in Figure 41. This plot condenses data obtained from 30 CDI experiments reported in ref (1171) into a single graph of energy versus productivity. As noted by Hawks et al.⁷³⁸ and Wang et al.,⁷⁵⁹ these quantities depend on conditions like water recovery, feed concentration, and average reduction in concentration. The data in Figure 41 are interpolated to obtain a curve that corresponds to the following conditions: water recovery is 50%, feed concentration is 20 mM, and product concentration is 15.5 mM. The free energy of separation, ΔĜ (per unit volume of diluate), is known and given by

19

where the summation is taken over all ions in solution, is the feed concentration of ion i, and and are the concentrations of ion i in the diluate and concentrate (brine), respectively. For a univalent binary electrolyte of a single salt, eq 19 can be rewritten as⁷⁵⁷

20

The free energy of separation can then be used to determine the thermodynamic efficiency of a given desalination process:

21

where Ê is the energy consumed per unit volume of treated water.

Figure 41.

Graphical explanation of the relationship between productivity, energy consumption, and thermodynamic efficiency. Values were obtained by interpolation of data from 30 CDI experiments reported in ref (1171). In this plot, water recovery (γ), feed concentration (c₀), and average reduction in concentration (Δc) are prescribed. The dashed straight line and dashed curved line show qualitative lower limits imposed by resistive and Faradaic losses, respectively.

Figure 41 shows the relationship between productivity and energy consumption in a general electrosorption process. This nonmonotonic behavior can be explained by the fact that energy consumption is governed primarily by resistive losses at high current and can be approximated as

22

where A is total active surface area and . If cycle efficiency and total resistance are constant, energy consumption and productivity would be directly proportional (see refs (738 and 759) for similar conclusions). The dashed straight line in Figure 41 shows a qualitative lower limit imposed by resistive losses that dominates when productivity is high. At low current, on the hand, energy consumption is determined mainly by Faradaic losses⁹⁰ and can be approximated as

23

where the charging current was replaced by the Faradaic leakage current, I_F. This simple model, represented by the dashed curved line in Figure 41, shows that energy consumption is inversely proportional to productivity. With the extent of desalination fixed, thermodynamic efficiency can be improved by carefully balancing the counteracting resistive and Faradaic losses.⁹⁰

6.4. Desalination

In this section, we quantify the energy consumed by various electrochemical methods for desalination based on data reported in the literature. Figure 42 shows that the energy needed to desalinate brackish water by these methods ranges mostly between 10 and 100 times the theoretical minimum requirement. In contrast, optimized SWRO systems come within a factor of two (excluding pre- and post-treatment steps) of the thermodynamic limit.⁶⁸ As a result, much research is focused on improving electrochemical systems to reliably desalinate brackish water and remove trace contaminants from dilute feeds with energy inputs closer to the thermodynamic limit. Accomplishment of this task requires innovation and optimization of advanced technologies for ion separations. Electrochemical methods are well positioned to be at the forefront of this effort because, as shown in Figures 2 and 42, they can efficiently desalinate brackish water and dilute feeds under a broad range of operating conditions.

Figure 42 plots the thermodynamic efficiency (eq 21) versus productivity for several electrochemical systems used for desalination. Here, we use the definition of productivity given by Hawks et al.:⁷³⁸

24

where V_D is the volume of diluate, A is the projected membrane or electrode area of a single cell or a cell pair, and n is the number of cells or cell pairs (an alternative definition, which is not used here, is based on total membrane area and sets n = 1; this definition differs from eq 24 by a factor of two). The methods considered in Figure 42 include EDI (squares), ED (diamonds), shock ED (circles), and CDI with either carbon (stars) or intercalation (triangles) electrodes. The color of each marker represents feed concentration, and marker size corresponds to the ratio c₀/c_D, which quantifies salt rejection. The data in Figure 42 reveal a trade-off between energy consumption and productivity, where systems with higher productivity generally exhibit lower efficiency. We also observe a quantitative trend where all data points fall below a straight line, which we argue is an empirical efficiency limit at this time for desalination by the electrochemical methods considered in this review. Devices near this line are among the most efficient electrochemical systems currently used for desalination. In the future, further development and optimization will lead to devices that can surpass this empirical limit.

On the basis of Figure 42, CDI with membranes and intercalation electrodes as well as ED are the most efficient electrochemical systems for desalination, with maximum values of 34.2% for CDI and 24.6% for ED . At the same time, these systems have primarily been used to remove small quantities of salt (c₀/c_D equals up to 5.6 for carbon CDI, 1.5 for intercalation CDI, and 44 for ED). In contrast, methods like EDI and shock ED display lower energy efficiency but have mainly been used to deionize water (c₀/c_D equals up to 500 for EDI and 10⁴ for shock ED). For nearly all the systems presented in Figure 42, water recovery ranges between 31% and 86% (in this context, water recovery is defined as the ratio of diluate volume to total feed volume, including electrode streams). One ED system, however, achieved only 4% water recovery (η_thermo = 4.91% and ) due to the high flow rates of the electrode streams, which were not recirculated.¹¹⁸⁷

6.5. Scale Up and Optimization

6.5.1. Electrodialysis

As explained in section 3.1, the cost of desalination by ED depends on the concentration of ions in the feed. In electromembrane processes, ion concentration determines current density and membrane area, both of which contribute to capital and operating costs and influence process efficiency. For example, a more conductive membrane results in faster ion transport, lower stack resistance, and less energy consumption. Faster ion transport also reduces the membrane area required by the device. Meanwhile, to separate the various components of a feed, the system needs a selective membrane. Numerous studies have focused on preparing IEMs with improved electrochemical properties, for example by employing functional nanomaterials and advanced surface modifications.^{343,1190−1192} These efforts have shown noticeable improvements in the electrochemical properties of IEMs and ultimately in the performance of ED systems.¹²⁴ Today, ED stacks can operate at industrial throughputs and are commercially available from many companies, including PCCell, Suez Water Technologies and Solutions, Pure Water Group, and Hydro Volta. These ED systems have been deployed for a variety of applications, including brine desalination, wastewater treatment, nutrient recovery (e.g., nitrogen, phosphorus), whey and sugar demineralization, juice deacidification, and wine stabilization.^{125,416,434,1193−1198}

In most of these applications, ED systems are subject to organic fouling, as discussed in section 3.6.¹¹⁹⁹ Strong hydrophobic interactions between the membrane and organic species can lead to irreversible adsorption and an increase in membrane resistance.^1198,1199 Surface modifications with polyelectrolytes have thus attracted much interest in the design of fouling-resistant IEMs.^{344,626,1200,1201} These modifications significantly increase the hydrophilicity of membrane surfaces and are simple to perform. Today, IEMs are manufactured by several companies, including DuPont (Nafion), Chemours (Nafion), ASTOM (Neosepta), AGC Chemicals (Selemion), and FUMATECH BWT (Fumasep), and much research has been devoted to improving the durability and lowering the cost of these membranes.^{417,422,1202,1203}

In practice, an ED stack may comprise hundreds of membrane pairs to achieve a sufficient membrane area when operating at high throughput.⁴¹⁶ A large number of membrane pairs and compartments, however, increases the resistance of the stack, and so ED is usually built with thin spacers and gaskets (thicknesses are commonly in the range of 0.5–2.0 mm) to lower this resistance.¹²⁰⁴ The narrow channels and compartments created by these thin spacers also induce a high cross-flow velocity, which elevates the pressure difference between the diluate and concentrate compartments. This pressure difference must be controlled to prevent excessive stress on the membranes.¹²⁰⁵ Another effective way to decrease stack resistance and improve performance is to use thinner membranes.^1065,1206 Reducing the thickness of the membranes, however, must be accompanied by improving their mechanical properties, as thinner membranes can become mechanically delicate. Profiled (or patterned) membranes are some of the latest developed IEMs that are used to boost the performance of electromembrane processes. With their nonflat, patterned surfaces, profiled membranes enhance fluid mixing at interfaces and in turn improve ion transport.¹²⁰⁷ In the future, ED may even require no spacer, which would further reduce the stack resistance and fabrication cost of ED.

6.5.2. Electrodeionization

One aspect of EDI that remains an important research topic is resin configuration.⁴⁸² Devices with mixed-bed configurations are commonly used to boost the removal of ions present at low concentrations by exploiting water dissociation. Random contact between cation and anion exchange resins in this configuration, however, may negatively impact ion transport due to reverse junction leakage.¹²⁰⁸ Separated and layered beds are alternative configurations that improve current efficiency by providing more pathways for ion transport,¹²⁰⁸ although the stack may need additional compartments or special spacer designs to maintain the resin in layers.

The loose resin beads in EDI can accumulate at the end of compartments or near the electrolyte outlets, which increases the pressure drop across the device.¹²⁰⁹ Accumulation of resin beads can also lead to an uneven distribution of current and suboptimal ion transport. One solution to this problem is to introduce ion exchange resin wafers, which are made of resin beads bound and immobilized into a porous matrix.^{83,456,465,1210−1212} These wafers ensure a mechanically stable distribution of the resin and prevent accumulation of resin beads. The binder material used in resin wafers could be either insulating⁸³ or conducting,⁴⁶⁶ and water splitting could be promoted by incorporating a catalyst in Janus bipolar resin wafers.⁴⁶⁵

Additional design features have been introduced to EDI systems to further improve their performance. One example, known as fractional EDI, involves the division of an EDI cell into a sequence of stages to optimize the distribution of electrical current.^481,1213 A fractional EDI cell comprises two or more separated pairs of electrodes to divide the power supply into amounts appropriate for each pair. In this way, the energy consumption can be lowered, and selective separations based on the charge of the impurities can be achieved.¹²¹³ Moreover, this design is better at limiting the formation of mineral scale because concentrates from different fractions are separated, which prevents interaction between multivalent ions from one fraction and the OH^– generated in another.

6.5.3. Shock Electrodialysis

The principles discussed in sections 6.5.1 and 6.5.2 are also relevant to the engineering of shock ED, although scaled-up commercial systems have yet to be demonstrated. As in ED or EDI, stacks of repeated layers may serve to lower the energy cost of driving electrical current at the electrodes relative to that of driving separations in all the layers,¹²¹⁴ although it may be possible to achieve similar efficiency at lower cost and complexity by using fewer but thicker porous layers at the centimeter scale.^87,95 Performance of shock ED may also be improved by varying the charge and microstructure of the porous material.^528,582,1215 Similar hierarchical porous media as in EDI, such as immobilized beds of ion exchange resin,^83,465,466 could be used to boost electromigration (compared to surface conduction in a glass frit) and promote shock formation and ion separations in the depletion zone. It may also be possible to lower capital costs and improve performance by replacing the IEMs used to initiate shock waves with thin, nanoporous ceramic or polymer layers analogous to the nanochannel junctions used in microfluidic ICP.^93,550,554

6.5.4. Capacitive Deionization

Today, CDI systems are available from several water technology companies. The first to develop a commercial CDI system was Voltea (founded in 2006), and it was followed by other companies including Ur-Water, Atlantis Technologies, Idropan dell’Orto Depuratori, EST Water Technologies, Siontech, and InnoDI Water Technologies.^1216,1217 As discussed in section 4.1.3, energy consumption by CDI has been extensively studied, and recent research demonstrated the possibility of combining photovoltaics and batteries to power a pilot CDI plant for remote applications.⁷⁹² Because pumping also contributes to the total energy demand, there has been growing interest in process design to reduce pumping costs.⁷⁹² For example, Nordstrand et al. recently designed a parallel arrangement of cells with symmetrically distributed flows to maintain a low pressure drop across the system.⁶⁴⁵ Another challenge during scale up of CDI is to make the process continuous, which has been the focus of FCDI^{802,818,819,1218} and multichannel MCDI^{929,948,1219,1220} systems. The latter enables continuous production of both fresh and brine streams by periodically switching the products of the middle and side channels.¹²¹⁹ Technoeconomic analyses of CDI also show the importance of achieving long system lifetimes, as capital costs (e.g., electrodes, membranes, frames, current collectors) are expected to outweigh operating costs (e.g., electricity, materials, labor).^{772,1221,1222} In particular, it is necessary to understand and mitigate electrode degradation^652,782 (see section 4.1.5) and cell fouling (see section 4.1.6). Several other publications on CDI optimization and pilot systems provide additional design rules and principles for scale up.^{645,771,792,818,1217,1218,1220,1223−1226}

6.5.5. Faradaic Electrosorption

Redox-based separation technologies remain an emerging area of research, and there are only a few preliminary studies on scale up and technoeconomic analysis. In one example, Joo et al. demonstrated a pilot electrochemical system using LiMn₂O₄ to selectively recover lithium from the desalination concentrate produced by RO and membrane distillation.¹²²⁷ This continuous system comprised 14 pairs of λ-MnO₂ and Ag electrodes arranged in parallel, and it was sequentially submerged in solutions for capturing, washing, and cleaning to recover the lithium. The system processed the desalination concentrate at a flow rate of 250 L h^–1 and produced an up-concentrated lithium solution with 88% purity and an enrichment factor of 1800. Metzger et al. conducted a comprehensive technoeconomic analysis on MCDI, HCDI, and intercalation CDI.⁹⁰⁸ This analysis showed that, compared to MCDI, intercalation CDI can achieve higher removal capacity and is more cost-effective and energy efficient.^908,1228 To treat equal volumes of water with similar performance, for example, an intercalation CDI module would cost only 27% of an MCDI module and would be four times smaller in size.

It has been shown, for certain applications, that electrochemical systems with redox reactions can be more cost-effective than conventional desalination methods. Kim et al. designed a system with redox-active electrodes to valorize proteins from whey waste and performed a preliminary technoeconomic analysis.¹²²⁹ To produce equal volumes of whey protein, the redox desalination system consumed up to 72% less energy than ED by driving redox reactions rather than water splitting. Moreover, the use of fewer IEMs in the redox system relative to ED lowered capital costs by 62%.¹²²⁹ Further progress in redox separation technologies requires a focus on device scale up, uncomplicated and economical synthesis of electrodes, long-term stability, and innovative stacking methods.¹²³⁰

6.6. Process Intensification

Many of the methods discussed in this review are still under development, and it is anticipated that innovations in materials science, device engineering, and process design will continue to advance these technologies. In addition to these advancements, process intensification can be undertaken to achieve more holistic (or global) improvements in electrochemical methods for water purification. Process intensification is any integration of unit operations, functions, and phenomena that leads to a smaller, cleaner, safer, and more efficient technology.^1231,1232 Emerging electrochemical methods provide new opportunities for process intensification by combining these systems with either other emerging methods or established technologies that are complementary in function.^{574,1233,1234} In an electrochemical method, the electric field enables selectivity and interfacial control, promotes fast reaction rates, and reduces energy consumption.^1233,1235 Advances in electrocatalysis have revealed that existing thermal and chemical methods can in fact be replaced by novel electrochemical processes.^1236,1237 Because of the importance of product purification, waste processing, and materials recycling, separations based on electrokinetics and electrosorption can play a key role in process intensification. Examples of process intensification for electroseparations include integration of reaction and separation using redox interfaces (as shown in Figure 29)^102,1014 as well as selective recovery of target species from waste streams using shock ED.⁵⁷⁴

The dual nature of electrochemical systems, highlighted by cells with asymmetrically configured electrodes, provides a new conceptual platform for process intensification.^102,902,1014 Although some Faradaic reactions are unwanted and can compromise efficiency, others can drive separations and promote catalysis.^96,652 In section 4.3, we explained that molecular design of Faradaic processes enables electrosorptive removal of heavy metal oxyanions and simultaneous redox transformation at the counter electrode. In one such example, As(III) was captured and transformed into As(V) using PVF and PTMA as electrodes,¹⁰² and in another, Cr(VI) was captured and transformed into the less harmful Cr(III) using PVF and platinum as electrodes.¹⁰¹⁴ Both of these systems involved electron transfer, which activated the redox units, and Faradaic reactions, which transformed the target metals into more benign products. Asymmetric combinations of intercalation systems and BDD electrodes have also been proposed for removal of both lithium and organic pollutants from industrial wastewater, as shown in Figure 43.¹²³⁸ Li⁺ was recovered (in excess of 98%) using an MnO₂ electrode with high selectivity toward Li⁺ relative to Na⁺, and more than 65% of the organic pollutant was decomposed at the stainless steel counter electrode. Kim et al. observed a trade-off, however, between the utilization efficiency of LiMn₂O₄ and the rate at which Li⁺ could be recovered.¹²³⁸ Moving forward, we expect a growing set of concepts that combine separations and reactions for water purification.

Figure 43.

Schematic of an electrochemical system with LiMn₂O₄ and BDD as electrodes for removal of both lithium and organic pollutants from industrial wastewater. Reproduced with permission from ref (1238). Copyright 2018 Royal Society of Chemistry.

Electrokinetic methods can also be used for molecular separations to effectively remove harmful contaminants or recover valuable targets.^478,574 For example, Li⁺ can be selectively captured from a multicomponent mixture and recycled (or reused elsewhere) by integrating CDI with intercalation materials^88,652 as a second step following shock ED.⁵⁷⁴ As shown in Figure 44, process intensification of this kind can in principle be achieved in two steps. In the first step, shock ED is used to concentrate waste in the brine, from which Li⁺ is selectively captured in the CDI unit by intercalation into an appropriate electrode such as LiFePO₄ or LiMn₂O₄.^{688,918,1239−1241} During this process, all cations are driven toward the intercalation electrode, but Li⁺ will be predominantly inserted into its crystal lattice because the vacancies in FePO₄ are well fitted for this small monovalent cation. Moreover, the anions are inserted into a porous carbon electrode, so fluid leaving the device will be depleted of Li⁺ and its counterion(s). In the second step, Li⁺ and its counterion(s) are released from the electrodes by reversing the direction of the applied electric field, and the fresh stream produced by shock ED is passed through the CDI unit to collect these ions.

Figure 44.

Process intensification of shock ED (left) by integrating CDI (right) to recycle Li⁺ in two steps. The first step is selective capture of Li⁺ in the intercalation electrode of a CDI unit from the brine discharged by shock ED. The second step is release of Li⁺ into the fresh stream produced by shock ED by reversing the direction of the applied field. Cations other than Li⁺ and H⁺ are labeled C⁺, anions are labeled A^–, and neutral species (unaffected by the electric fields) are labeled N; streams are colored based on the relative concentration of ions. Reproduced with permission from ref (574). Copyright 2019 American Chemical Society.

Process intensification of electrochemical methods not only enables targeted separations but also dramatically reduces energy consumption. By comparing the energetics of sequential and coupled Faradaic processes, Kim et al. lowered the energy needed to remove arsenic from 2.2 to 0.45 kW h mol^–1, as shown in Figure 45.¹⁰² The basis of this energy integration was to reduce parasitic side reactions and improve current efficiency by using asymmetric redox systems. Relative to systems with nonselective electrodes (e.g., CDI with porous carbon electrodes), energy efficiency was improved by nearly an order of magnitude. Similarly, in recent work on PFAS capture by electrochemical mediation with TEMPO-based polymers, an asymmetric system of TEMPO copolymers with a BDD counter electrode achieved reactive separation of PFAS.¹⁰²¹ During release of the PFAS by the redox polymer, the BDD electrode simultaneously broke down PFAS with little energy consumption and a defluorination efficiency over 50%. Electroseparations can therefore enable processes that would otherwise be prohibitively expensive or complex, such as purification of dilute feeds, recovery of metals, valorization of wastes, and isomeric separations of biochemicals and pharmaceutical compounds.^{105,154,1234,1242} As shown in this review, electrokinetics and electrochemistry can improve energy efficiency and reduce secondary waste, which enables cleaner and more sustainable processes for water purification.

Figure 45.

Analysis of energy integration and process intensification using redox electrochemistry with coated electrodes for reactive separations. Reproduced with permission from ref (102). Copyright 2020 Wiley-VCH.

7. Conclusions and Outlook

Electrochemical methods for water purification use applied electric fields to remove contaminants by either degrading or converting them through redox reactions, driving their separations in a bulk electrolyte, electrostatically trapping them in an EDL (where they may undergo electrochemical reactions), or intercalating them in solid electrodes. The ability to remove contaminants directly from water, as opposed to removing the water from the contaminants, is the property that enables most advantages of these methods. Nondestructive electrochemical methods, which are based on electrokinetic (ED, EDI, ICP, shock ED) and electrosorption (CDI, Faradaic electrosorption) processes, tend to be more energy efficient compared to physical methods when used for molecular separations or purification of dilute feeds. These systems have other attractive features including compactness, molecular selectivity, versatility, decreased generation of secondary waste, and the ability to combine reactions and separations. The selectivity and versatility of electrochemical systems enables unique combinations of technologies for exotic separations and recovery of high-value elements. In particular, the dual nature of these systems provides a new conceptual platform for process intensification of existing methods. Significant advances, however, are still needed both at the fundamental level (e.g., to impart higher molecular selectivity to electrodes for emerging applications) and in process engineering through the development of new hybrid systems for higher energy efficiency. Comprehensive work on technoeconomic analysis is also needed for these emerging methods to provide a comparative evaluation against other treatment technologies and identify limitations, areas of opportunity, and protocols for optimal operation. Here, we highlight emerging research directions, challenges, and opportunities in electrochemical methods for water purification, ion separations, and energy conversion.

7.1. Materials Design for Multicomponent Separations

Despite advances in redox-active materials, which have enabled new selectivity and higher uptake, there remain significant challenges in both fundamental studies and new areas of application.

7.1.1. Multifunctional Redox Materials

To enhance selectivity and overcome existing limitations, one promising approach has been to combine several distinct chemical groups. While redox homopolymers are efficient for applications that demand ion selectivity,⁹⁶ redox copolymers extend these capabilities by combining orthogonal properties. For example, redox copolymers have enhanced electrochemical regeneration, chemical binding of rare-earth elements,¹²⁴³ and control of hydrophobicity, affinity, and electrostatics within redox adsorbents for PFAS molecules and other micropollutants.^{1021,1244,1245} Nanostructured combination of two distinct redox polymers has also enabled the reversible electrochemical binding and control of neutral molecules through redox-tunable hydrophobicity.¹⁰³ Similar concepts of hybrid materials may be generalized to redox crystals and various multicomponent separations, in which reversibility or selectivity could be enhanced through hybridization to achieve properties beyond the base chemical structure. Moving forward, we envision these multifunctional materials will provide a versatile platform for extraction of multiple desired elements simultaneously.

7.1.2. Computational Design and Operando Electrochemical Tools for Understanding Interactions

To guide the bottom-up design of materials, computational studies play a key role in the selection of chemistry and tuning of morphology. Tools such as MD⁷¹⁶ and electronic structure calculations¹²⁴⁶ have been critical in understanding selectivity in ultramicroporous carbon and redox-active materials, respectively. We envision molecular simulation tools will increase in utility due to the need for a greater understanding of selective interactions and their underlying mechanisms. In addition to computational tools, we expect that the integration of operando electrochemical tools such as vibrational spectroscopy will elucidate interfacial structure and binding mechanisms. These molecular tools, along with macroscopic transport modeling, can help advance the performance of electrochemical separation methods at multiple scales.

7.1.3. Advances in Membrane Materials

Advances in membrane materials can improve the efficiency of electrokinetic processes. Membrane fouling and poor stability under certain operating conditions remain longstanding challenges in electrokinetic systems. Besides, redox materials in electrokinetic systems could pass the membranes and contaminate the treated water. The introduction of more durable, cost-effective, and selective membranes is thus needed for more efficient and reliable electrokinetic systems. Structured nanomaterials are expected to play an important role in improving membrane performance.^1247,1248 Moreover, ion selectivity can be modulated through fine control of the structure, pore size, water permeability, and functional groups of the membranes.^1248,1249

7.1.4. Multicomponent Mixtures for Separations

Many of the recent studies on selective removal and resource recovery have been performed using idealized simulated solutions. Real industrial and municipal effluents, however, involve complex speciation of ions and organic compounds, which affects both the selectivity and the lifetime of electrodes and membranes. Municipal effluents are often complex multicomponent mixtures that comprise organic substances, inorganic ions, and sometimes biological species,^145,1250 many of which could form unwanted deposits. Groundwater often contains significant organic matter, and the solution can exhibit a range of ionic strengths and pHs. As a result, there is a need to evaluate the performance of functional materials when exposed to real samples¹²⁵¹ to quantify reliability and stability (e.g., electrochemical, chemical, mechanical). At a fundamental level, these studies may reveal the need for materials modification such as the use of antifouling coatings, which can provide a tunable balance between uptake, selectivity, and stability.

Nutrient recovery and metal recovery are two areas of emerging interest in the development of molecularly selective materials. There have been significant advances in the design of selective materials for nutrient recovery, especially using nanostructured porous carbons.⁷¹⁶ Improving selectivity further requires a deeper understanding of solvation, sterics, and electronic structure interactions due to the similarity in electric charge between nutrient species (particularly NO₃^–) and competing ions. At the same time, many electrosorption systems are still highly dependent on pH during separations,^1252,1253 and because many oxyanions have complex speciations that depend on pH, there have been approaches that leverage electrochemical swings in pH for the recovery of nutrients such as PO₄.^806,1253

The recovery of critical elements from mining, recycling, and industrial wastewater streams is another area of emerging importance for selective separations. Many of these transition metals are present in small amounts and are surrounded by excess competing species of similar valence and structure, which makes ion-selective electrosorption a “needle-in-a-haystack” challenge.^1246,1254 Functionalized carbons and redox-active electrodes have made significant strides toward efficient metal recovery and heavy metal removal.^{646,1255,1256} However, there are key challenges still to be overcome, such as achieving higher selectivity between different transition metals within multicomponent mixtures. Rare-earth elements, for example, are often found in the presence of each other as well as other transition metals,¹²⁵⁷ which makes their separation a difficult and costly task. Many competing metals are electroactive at moderate potentials (e.g., copper, lead), and they can often interfere with selective electrosorption processes. Therefore, pretreatment and tuning of electrochemical potentials will be needed to selectively purify multicomponent mixtures.

7.2. Intensifying Water Treatment Processes Through Hybrid Approaches

Electrochemical methods can assist in process intensification by decreasing waste, lowering solvent use, and reducing the number of unit operations. In particular, areas for integration include the direct combination of electrochemical methods with renewable energy sources as well as the development of single-cell reactive separations.

7.2.1. Integrating Renewable Energy Sources and Electrochemical Methods

Coupled with global environmental crises, the development of sustainable manufacturing processes has become a major goal for industries. In this regard, combining electrochemical processes with renewable energy sources (e.g., solar panels) can reduce energy consumption and carbon intensity.¹²⁵⁸ Recently, electrochemical methods for water purification have been hybridized with dye-sensitized solar cells for direct conversion of light to electricity.^1259,1260 The photoanode uses desalinated chloride to generate reactive chlorine species, which can treat wastewater, while the cathode produces molecular hydrogen.¹²⁶¹ The combination of photoelectrochemistry and redox-flow desalination enables continuous treatment, even in the absence of an external energy source. These systems demonstrate opportunities for the development of processes that integrate energy conversion and electrochemical separations.

7.2.2. Coupling Reaction and Separation

For persistent contaminants, degradation of these species is as important as their removal from water. The coupling of reaction and separation can be a powerful approach for providing modular water treatment, as in the case of capture and breakdown of PFAS.^985,1262 The integration of membrane technologies and electrochemical advanced oxidation processes has gained much attention for treatment of organic pollutants in wastewater. In these hybrid processes, a conductive membrane serves as a flow-through anode, which filters the wastewater and drives oxidation of the organic contaminants and pharmaceutical residues.¹²⁶³ Moreover, integration of redox processes and magnetic nanoparticles enables selective separation of various organic and inorganic micropollutants.¹²⁴⁵ We envision that emerging electrosorption and electrokinetic systems coupled with advanced oxidation processes can provide efficient process intensification for wastewater treatment.

7.3. Translation to Practical Applications

While there have been extensive studies of carbon electrode stability,^1264,1265 many of the emerging Faradaic materials have only been examined using idealized solutions. Demonstration of stability and reliability using real samples with relevant electrolyte concentrations is essential. For desalination, electrode stability can be improved by doping the carbon and creating hybrid materials.^1265,1266 It is also important to design materials that suppress parasitic reactions and function at lower potentials, especially when the side reactions involve chloride or oxygen, which often produce unwanted byproducts.⁶⁵² Proof of economic feasibility is also critical for commercialization of electrochemical systems. Recently, there has been an increase in the number of publications on life cycle assessment and technoeconomic analysis.^{1221,1267,1268} These technoeconomic analyses show that the costs of electrodes and membranes are key contributors to the overall capital and operating costs.^908,1221 In practice, technoeconomic analysis will provide a comprehensive framework for assessing feasible electrochemical approaches for water treatment.

In summary, electrochemical systems for water purification and resource recovery are already commercialized in many industries to treat water contaminated with various kinds of waste. Whether or not the emerging purification methods meet their expectations in the areas of water desalination, remediation, and separations, it seems likely that several of these methods will eventually be employed. In the near term, we expect that a deeper knowledge of the transport phenomena and electrochemical kinetics that govern these methods will facilitate their engineering and optimization. In the long term, work directed toward understanding and improving the design of these processes at scale may guide how commercial prototypes should be built for a given application.

Glossary

Abbreviations

AEM

anion exchange membrane

H₂Q

hydroquinone

AOT

brand name of DOSS

HA

hydroxyapatite

BattMix

battery mixing

HBMA

hydroxybutyl methacrylate

BDD

boron doped diamond

HCDI

hybrid capacitive deionization

BDI

battery deionization

HCF

hexacyanoferrate

BMCSL

Boublik–Mansoori–Carnahan–Starling–Leland

HS

high salinity

BSK

Bazant–Storey–Kornyshev

ICP

ion concentration polarization

BWRO

brackish water reverse osmosis

IEM

ion exchange membrane

CapMix

capacitive mixing

LDH

layered double hydroxide

CDI

capacitive deionization

LS

low salinity

CDLE

capacitive double layer expansion

MCDI

membrane capacitive deionization

CDP

capacitive Donnan potential

MD

molecular dynamics

CEM

cation exchange membrane

NASICON

sodium superionic conductors

CIET

coupled ion–electron transfer

NMO

sodium manganese oxide

CNT

carbon nanotube

NMR

nuclear magnetic resonance

CTAB

cetyltrimethylammonium bromide

PANI

polyaniline

CVD

chemical vapor deposition

PAQ

polyanthraquinone

DBS

dodecylbenzenesulfonate

PBA

Prussian blue analogue

DC

direct current

PFAS

polyfluoroalkyl substances

DFT

density functional theory

PNPS

Poisson–Nernst–Planck–Stokes

DO

dissolved oxygen

PPCP

pharmaceutical and personal care product

DOSS

dioctyl sulfosuccinate

PPy

polypyrrole

DSA

dimensionally stable anode

PRO

pressure-retarded osmosis

EC

electrochemical process

PTMA

poly-TEMPO methacrylate

ED

electrodialysis

PVF

polyvinylferrocene

EDI

electrodeionization

RC

resistor–capacitor

EDL

electric double layer

RCDI

rocking-chair capacitive deionization

EDS

energy dispersive X-ray spectroscopy

RGO

reduced graphene oxide

EDTA

ethylenediaminetetraacetic acid

RMSS

redox-mediated selective separations

ENAS

energy normalized adsorbed salt

RO

reverse osmosis

ETAS

electrochemically tunable affinity separations

SEC

specific energy consumption

FB

flow-between

SEM

scanning electron microscopy

FBV

Frumkin–Butler–Volmer

SHE

standard hydrogen electrode

FCDI

flow electrode capacitive deionization

SOG

Sudan Orange G

FO

forward osmosis

SWRO

seawater reverse osmosis

FTE

flow-through electrode

TEM

transmission electron microscopy

FUE

ferrocene utilization efficiency

TEMPO

(CH₂)₃(CMe₂)₂NO

GCMC

grand canonical Monte Carlo

VF

vinylferrocene

GCS

Gouy–Chapman–Stern

Biographies

Mohammad A. Alkhadra graduated from UC San Diego in 2018 with both a B.S. and an M.S. in Chemical Engineering, after which he moved to Cambridge, MA, to begin his Ph.D. at MIT in the same major. At MIT, Mo focuses on ion separations using shock ED under the supervision of Prof. Martin Z. Bazant.

Xiao Su received his B.S. in Chemical Engineering from the University of Waterloo in 2011 and his Ph.D. in Chemical Engineering from MIT in 2017. He joined the faculty in the Department of Chemical and Biomolecular Engineering at the University of Illinois at Urbana–Champaign (UIUC) in 2019, where he focuses on electrochemical separations and process intensification. His research has been recognized by an NSF CAREER award in 2019 and the ACS Victor K. LaMer Award in 2020.

Matthew E. Suss heads the Cleantech Innovations Laboratory in the Faculty of Mechanical Engineering and Wolfson Department of Chemical Engineering at Technion—Israel Institute of Technology and is affiliated with the Grand Technion Energy Program and Grand Water Research Institute. Prof. Suss serves as a member of the Israel National Research Center for Electrochemical Propulsion (INREP) and as an Israeli delegate to the European Federation of Chemical Engineering (EFCE) working group on electrochemical engineering. Prof. Suss has won numerous national and international research awards, including the prestigious Alon Fellowship and an ARCHES award.

Huanhuan Tian joined the Department of Chemical Engineering at MIT in 2018 and is pursuing her Ph.D. under the supervision of Prof. Martin Z. Bazant. She received her B.S. and M.S. with honors in 2018 from the Department of Engineering Mechanics at Tsinghua University in China. Huanhuan is interested in issues related to energy and the environment, and her current research focuses on mathematical modeling of shock ED for selective ion separations.

Eric N. Guyes is a Ph.D. candidate at Technion under the supervision of Prof. Matthew E. Suss. He arrived at Technion in 2013 as a Fulbright Scholar and later joined Prof. Suss’s group in 2014. His doctoral research focuses on the mechanisms and applications of selective ion removal in CDI for water desalination and treatment applications.

Amit N. Shocron is a Ph.D. candidate at Technion under the supervision of Prof. Matthew E. Suss. He received his B.S. and M.S. in Mechanical Engineering also at Technion. His research focuses on CDI for ion selective water treatment and water desalination.

Kameron M. Conforti received his Ph.D. in Chemical Engineering at MIT in 2019, after which he worked as a Postdoctoral Associate in the Department of Mechanical Engineering at MIT. He is currently a Senior Research Engineer in the Competency Research Lab at Saint-Gobain Research North America, where he specializes in modeling heat and mass transfer.

J. Pedro de Souza is a Ph.D. candidate in the Department of Chemical Engineering at MIT. He received his B.S. in chemical engineering at the University of Texas at Austin in 2016. He works with Prof. Martin Z. Bazant on theoretical modeling of the equilibrium structure and transport of ions at charged interfaces.

Nayeong Kim received her B.S. in Chemical and Biological Engineering from the University of Wisconsin–Madison in 2017 and her M.S. in the same major from Seoul National University in 2020 under the supervision of Prof. Jeyong Yoon. She is pursuing her Ph.D. at UIUC under the supervision of Prof. Xiao Su, where she focuses on environmental electrochemistry and integrated electrochemical separations.

Michele Tedesco has a Ph.D. in Chemical Engineering and is currently a scientific project manager at Wetsus. His research focuses on IEMs and related processes, including ED and reverse/bipolar ED, from fundamental modeling to experimental investigation.

Khoiruddin Khoiruddin is an assistant professor in the Department of Chemical Engineering at Institut Teknologi Bandung (ITB), Indonesia. He received his Ph.D. in Chemical Engineering from ITB in 2018. Formerly, he was a process engineer at the membrane company GDP Filter Indonesia from 2009 to 2018. His research interests include design of IEMs and applications of electromembrane processes.

I Gede Wenten is a professor in chemical engineering at ITB, Indonesia. He received his Ph.D. in Chemical Engineering from the Technical University of Denmark (DTU) in 1995. He has extensive experience in membrane technology, both in the industrial and academic settings, with a career spanning over three decades. His research interests include design, preparation, and fouling control of membranes.

Juan G. Santiago received his Ph.D. in Mechanical Engineering from UIUC in 1995. His research includes the development of microsystems for on-chip chemical and biochemical analysis, methods for DNA quantification and hybridization, and CDI. He is a Fellow of the American Physical Society, a Fellow of the American Society of Mechanical Engineering, and a Fellow of the American Institute for Medical and Biological Engineering. He serves and has served as an editor of several journals and cofounded several companies in microfluidics.

T. Alan Hatton is the Ralph Landau Professor and Director of the David H. Koch School of Chemical Engineering Practice at MIT. He earned his B.S. and M.S. from the University of Natal, Durban, and his Ph.D. from the University of Wisconsin–Madison in 1981. As part of the MIT Energy Initiative, Prof. Hatton codirects the Center for Carbon Capture, Utilization, and Storage. His research focuses on the development of electrochemical processes to facilitate chemical separations and to mediate the transformation of capture waste into useful commodity chemicals.

Martin Z. Bazant is the Edwin G. Roos (1944) Chair Professor of Chemical Engineering and Mathematics at MIT. He has made seminal theoretical contributions in electrokinetics and electrochemistry with applications in energy storage, water purification, and microfluidics. Prof. Bazant’s awards include the 2015 Alexander Kuznetsov Prize in Theoretical Electrochemistry, the 2018 Andreas Acrivos Award for Professional Progress in Chemical Engineering (AIChE), and the 2019 MITx Prize for Teaching and Learning in Massive Open Online Courses. He consults extensively for industry and serves as the Chief Scientific Advisor for Saint-Gobain Research North America in Massachusetts.

The authors declare no competing financial interest.