Recent advances in the electrochemical oxidation water treatment: Spotlight on byproduct control

Electrochemical oxidation (EO) is a promising technique for decentralized wastewater treatment, owing to its modular design, high efficiency, and ease of automation and transportation. The catalytic destruction of recalcitrant, non-biodegradable pollutants (perand poly-fluoroalkyl substances (PFAS), pharmaceuticals, and personal care products (PPCPs), pesticides, etc.) is an appropriate niche for EO. EO can be more effective than homogeneous advanced oxidation processes for the degradation of recalcitrant chemicals inert to radical-mediated oxidation, because the potential of the anode can be made much higher than that of hydroxyl radicals (EOH = 2.7 V vs. NHE), forcing the direct transfer of electrons from pollutants to electrodes. Unfortunately, at such high anodic potential, chloride ions, which are ubiquitous in natural water systems, will be readily oxidized to chlorine and perchlorate. Perchlorate is a to-be-regulated byproduct, and chlorine can react with matrix organics to produce organic halogen compounds. In the past ten years, novel electrode materials and processes have been developed. However, spotlights were rarely focused on the control of byproduct formation during EO processes in a real-world context. When we use EO techniques to eliminate target contaminants with concentrations at μg/L-levels, byproducts at mg/L-levels might be produced. Is it a good trade-off? Is it possible to inhibit byproduct formation without compromising the performance of EO? In this minireview, we will summarize the recent advances and provide perspectives to address the above questions.

requires DC power supply and electrode array, and it is usually operated in open air at room temperature. The treatment capacity of EO processes is scalable, depending on the numbers and areas of electrode arrays used. EO exhibited excellent performance on the destruction of nonbiodegradable pollutants (PFAS, PPCPs, pesticides, etc.) (Carter and Farrell, 2008;Zhuo et al., 2012;Jasper et al., 2016). For example, recent advances show that one-log removal of typical PFAS (perfluorooctanesulfonic acid (PFOS) and perfluorooctanoic acid (PFOA)) and PPCPs (carbamazepine, metoprolol, and ciprofloxacin) can be achieved at energy consumptions less than 10 kWh/m 3 Le et al., 2019;Yang et al., 2019a).
By tuning the anodic potential, various oxidants (OCl -, O 3 , Cl$ , $OH, etc.) can be produced at the anodes (Fig. 1). EO is more efficient than the homogeneous advanced oxidation processes (AOPs) on the degradation of pollutants that barely react with radicals, because the potential of anodes can be made much higher than that of hydroxyl radicals (E OH = 2.7 V vs. NHE), forcing the direct transfer of electrons from target compounds to electrodes. For instance, perfluorooctanoate (C 7 F 15 COO -) is resistant to $OH oxidation. However, it can be readily destroyed by EO via direct electron transfer (DET) oxidation (Niu et al., 2013).
The generation of oxidants requires high anodic potential. Unfortunately, at such high anodic potential, chloride ions, which are ubiquitous in natural water systems, will be readily oxidized to free chlorine (HClO/ OCl -; pK a 7.5) and then further oxidized to perchlorate (ClO 4 -). Perchlorate is an endocrine disruptor posing an adverse impact on thyroid gland function (Urbansky and Schock, 1999). In 2019, the US Environmental Protection Agency (EPA) proposed a maximum contaminant level at 56 μg/L. Some US states have already regulated ClO 4 in drinking water at lower levels (2 μg/L for Massachusetts and 6 μg/L for California). Current experimental results indicate that, once it is formed, the elimination of ClO 4 is very challenging (Schaefer et al., 2007;Liu et al., 2016).
In the area of electrochemical disinfection, balancing the acute risk caused by pathogens against the chronic carcinogenic risk associated with DBPs is an endless debate. As for the treatment of recalcitrant chemical contaminants and pathogen inactivation, it is clear that DBPs (μg/L-level) and perchlorate (mg/L-level) are produced during the removal of μg/L-level target compounds (or cells/L-level pathogens) in chloride bearing surface water and waste stream (Anglada et al., 2009, Fig. 1 Oxidation power of oxidants and direct electron transfer reactions characterized in the scale of redox potentials. The bottom frame shows the mechanism of indirect oxidation, in which target compounds react with reactive chlorine species (RCS) and reactive oxygen species (ROS) generated electrochemically. The top frame demonstrates the oxidation reactions based on the direct electron transfer mechanism. 2010; Bagastyo et al., 2013;Schaefer et al., 2015;Yang et al., 2019a). These facts keep us wondering whether adopting EO treatment is a good trade-off.
Despite that breakthroughs were made on the development of novel electrode materials, it is equally important to work on the control of byproducts. This review primarily focuses on the control of THMs, HAAs, and ClO 4 as they are facing regulatory pressure. It is also important to note that THMs and HAAs could serve as indicators of exposure to the complex mixture of DBPs. The concentrations of THMs and HAAs are usually proportional to the total organic halogen concentrations in treated effluents (Pourmoghaddas and Stevens, 1995). Although this correlation has not been proved for the EO systems, we speculate that through the control of THMs, HAAs, and ClO 4 -, the overall reduction of other unknown halogen byproducts can be achieved.
It is also important to address that health concerns of DBPs and ClO 4 were initially raised by the drinking water treatment sector. EO is less likely to be applied to treat drinking water with low conductivity. In wastewater treatment, the control of byproducts seems to be unnecessary as the effluents are not required to be treated to a drinkable degree. However, the author believes it is still important to investigate the byproduct suppression strategies in the EO treatment of wastewater, and in some case, drinking water due to the following reasons: 1) Some studies demonstrated that EO could be used to treat the emerging contaminants and inactivate pathogens in drinking water and surface water at acceptable energy consumption levels (less than 10 kWh/m 3 ) (Martínez-Huitle and Brillas, 2008;Schaefer et al., 2015;Yang et al., 2019a).
2) Effluents from the EO wastewater treatment units containing (DBPs) will eventually be discharged to the surface water, which could potentially be utilized downstream as a potable water supply.
3) The author has received review comments from almost every article and proposal, requiring the evaluation of DBPs formation in the EO processes, even though the technique was proposed for wastewater treatment. These comments were raised due to the high expectations from broad scientific and industrial communities on the emerging EO technology. "The control of byproducts in the EO reactions in all matrices." The author believes this is a question of when, not if.
2 Byproduct formation during EO treatment

Electrode materials
The performance of EO and the byproduct generation rates are largely dependent on the properties of electrodes. As shown in Fig. 2(a), depending on the activity of water oxidation, electrodes can be categorized as active and nonactive electrodes (Chaplin, 2014;Martínez-Huitle et al. 2015;Chaplin, 2019). The property of electrodes can be identified by cyclic voltammetry (CV) analysis ( Fig. 2(b)). In these tests, electrodes are immersed into the inert electrolyte (e.g., Na 2 SO 4 or NaClO 4 ). The anodic potential is increased at a fixed rate, and the response of current is recorded. Assume that all the electrons withdrawn from the anode are only contributed by water oxidation (2H 2 O ! O 2 + 4H + + 4e -), the larger current recorded implies faster reaction rates. In the CV analysis of the active electrodes, the sharp increase of current usually occurs at 1.5 V RHE . These onset potentials are slightly higher than the theoretical potential for water oxidation (1.23 V RHE ). Therefore, it is concluded that active electrodes have low overpotentials for oxygen evolution. Typical Active electrodes include RuO 2 and IrO 2 . As for nonactive electrodes, such as PbO 2 , Sb-SnO 2 , boron-doped diamond (BDD), and sub-stoichiometric TiO 2x electrodes, their onset potentials usually are higher than 2 V RHE .
The different activities toward oxygen evolution stems from the surface chemistry of electrode materials. Mechanisms are well explained in previous studies (Trasatti, 1984;P. Chaplin, 2014). During electrolysis, $OH is produced from water as a surficial intermediate. The $OH strongly binds with the active electrode (M n O), and is rapidly incorporated into its lattice structure to form higher oxide (M n+1 O), then the reaction proceeds to produce O 2 . As for non-active electrode, $OH loosely binds with the electrode surface. The oxygen production is hence disfavored.
Assume that both active and non-active electrodes are operated at a constant current density of 10 mA/cm 2 ( Fig. 2(b)), the active electrode at 2 V RHE can only produce chlorine (Cl 2 /Cl -1.36 V RHE ). The non-active electrode has a potential of 3.2 V RHE , which is high enough to satisfy the thermodynamic criteria for the production of multiple oxidants such as $OH, O 3 , and H 2 O 2 ( Fig. 1).
Since oxygen evolution shares similar mechanisms with chlorine evolution (Krishtalik, 1981), active electrodes have high efficiency for chlorine production. The free chorine produced can efficiently inactivate pathogens and remove ammonium via breakpoint chlorination. As for nonactive electrodes, they demonstrate superior performance on the degradation of recalcitrant organic compounds that are inert to free chlorine but reactive toward radicals.
Electrolysis using active electrodes tends to produce DBPs due to the buildup of free chlorine to react with matrix organics in bulk solution (vide infra). In contrast, nonactive electrodes are oxidative enough to mineralize organic contaminants and destroy the as-formed DBPs (Jasper et al., 2017), but their strong oxidation capability leads to the rapid oxidation of free chlorine to ClO 4 -.
The adsorbed chlorine radical then combines with Clin the bulk solution or with neighboring surficial Cl$ to produce Cl 2 (rxn. 2 and 3), which subsequently hydrolyze to HOCl or OCldepending on pH.
The homogeneous reaction of chlorite (ClO 2 -) and free chlorine could produce ClO 3 - (Gordon and Tachiyashiki, 1991). However, the contribution of this step to ClO 4 formation in EO treatment might be minimal, as the chlorine balance analyses showed that the concentrations of ClO 2 in the bulk electrolyte are negligible (Yang et al., , 2019aMostafa et al., 2018). Our studies on the EO treatment of latrine wastewater (a mixture of feces and urine) showed that at moderate current density (10 mA/ cm 2 ), IrO 2 electrode was inert to, while BDD electrode was quite reactive to ClO 4 generation Jasper et al., 2017). Interestingly, a study focusing on EO treatment of urine reported that at a higher current density (20 mA/cm 2 ), the IrO 2 electrode was able to produce ClO 4 -. When active electrode Pt/Ti was operated at 270 mA/cm 2 , the production of ClO 4 was observed as well (Jung et al., 2010). These recent studies break the stereotype that active anode cannot produce ClO 4 -. In fact, both active and nonactive electrodes can produce ClO 4 when the thermodynamic criteria (e.g., anodic potential and current density) are met. The kinetics of the multi-step oxidation of chlorine oxyanions (rxn 4-9) at the electrode/electrolyte interface determines how fast the ClO 4 can be produced.

Formation of DBPs
The DBP formation mechanisms in EO treatment were rarely investigated specifically. Therefore, mechanisms discovered in disinfection are revealing. THMs and HAAs result from reactions between free chlorine and natural organic matters (NOMs) in the drinking water disinfection process (Bond et al., 2012). NOMs could be considered as surrogates for most of the organic contaminants of concern, as NOMs have multiple functional groups: carboxylic acid, enolic hydrogen, phenolic hydrogen, quinine, alcoholic hydroxyl, ether, ketone, aldehyde, ester, lactone, amide and amine (Thurman, 2012). Free chlorine reacts with NOMs through oxidation, addition, and electrophilic substitution. In general, aromatic precursors have THM and HAA formation potentials (2-1892 μg/mg C) two to three orders of magnitude higher than aliphatic precursors (Bond et al., 2012). High pH favors the formation of DBPs by promoting the hydrolysis of halogenated leaving group (Fig. 3). Long contact time and high chlorine/organic ratio promote the formation of THMs and HAAs (Sérodes et al., 2003;Sun et al., 2009). Learning from the above, we know that both aliphatic and aromatic compounds, no matter they are target pollutants or matrix organics, can react with chlorine to produce DBPs. But the formation of DBPs is not that easy. It requires multiple steps (long contact time), appropriate pH, and high chlorine dose. Therefore, it is possible to control DBPs by adjusting the operational parameters and reaction conditions of EO treatment (vide infra).
In homogeneous chlorination disinfection operation, free chlorine at high concentration is spiked in water initially. As for EO treatment, chlorine is gradually generated. DBP formation is usually insignificant at the early-stage of electrolysis because free chlorine will be instantly consumed by excessive organics and ammonium. Organics are not yet broken down to small molecule precursors with abundant halogen leaving groups. Sharp increase of DBPs was reported when the majority of organics was removed (>90% removal of COD) accompanied by the rapid increase of free chlorine concentration (Zöllig et al., 2015;Jasper et al., 2017). These results indicate that the DBP formation during electrolysis is still dominated by the homogeneous reactions in the bulk solution. Most of the studies reported that active electrodes tend to produce more DBPs than non-active electrodes because the former produces more free chlorine. However, a few conflicting results were reported. Bagastyo et al. (2012Bagastyo et al. ( , 2013 studied the formation of THMs and HAAs during the electrolysis of RO concentrate using IrO 2 and BDD electrodes. In this study, the BDD electrode produces more HAAs and THMs than IrO 2 . The authors speculated that BDD oxidized chloride to chlorine radicals (Cl$) and dichloride radical anion (Cl 2 $ -). The chlorine radicals might be able to react with organics to form DBPs.

Control strategies
3.1 Fabricate anode materials Pralay et al. use perfluoro-decyl trichlorosilane to fabricate the surface of BDD with the goal of inhibiting DET steps (rxn. 5, 7, and 9) via steric hindrance and hydrophobic effects (Gayen and Chaplin, 2017). The fluorinated polymeric coating prevents the direct contact between ClO 3 and BDD surface but allows $OH to diffuse into the bulk electrolyte. Another study reported that covering the BDD surface with perfluorooctanoic acid (PFOA) could inhibit the formation of ClO 4 - (Jawando et al., 2015). This result implies that the EO process could produce less ClO 4 during the treatment of per-and poly-fluorinated chemical waste at high concentrations. It is important to note that the above studies operated BDD at low current densities (ca. 1 mA/cm 2 ). At higher current densities, the stability of the surface coating will be a critical issue, as BDD electrodes can oxidize the poly-and perfluoroalkyl coatings. Our study found that significant formation of ClO 4 was still observed when BDD was operated at 10 mA/cm 2 to treat 10 mg/L PFOA in the presence of 3 mmol/L Cl - (Yang et al., 2019a). Forming less ClO 4 during treatment of high PFAS concentration may be important in the PFAS remediation process. However, it cannot be applied as a general strategy to mitigate ClO 4 formation. Another option is to use alternative electrodes for BDD. Active electrodes such as Pt/Ti, IrO 2 /Ti, and RuO 2 /Ti have lower ClO 4 formation potentials, but they are also less efficient for organic oxidation. Recently, sub-stoichiometric TiO 2-x NTA and Ti 4 O 7 phase anode were developed with comparable performance with BDD (Zaky and Chaplin, 2013;Zaky and Chaplin, 2014;Yang et al., 2018). These TiO 2-x electrodes produced less ClO 4 than BDD Wang et al., 2020). Though the molecular-scale mechanism is still unclear, we suspected that $OH-mediated oxidation, rather than DET, is the major contaminant removal mechanism on TiO 2-x electrodes with abundant surficial titanol groups (Kesselman et al., 1997;Bejan et al., 2012), while BDD has higher reactivity for DET mediated oxidation, which favors the ClO 4 formation.

Enhance cathodic dehalogenation
Typical EO cells use stainless steel or Ti metal as cathodes, on which hydrogen evolution reaction (HER) occurs via proton reduction. It is viable to replace the HER cathodes with catalytic electrodes that can reductively remove perchlorate and DBPs. Perchlorate can be electrochemically reduced by Rh, Pt, Sn, Cu, and Ni (Horányi and Bakos, 1992;Wasberg and Horányi, 1995;Wang et al., 2007). Reactions can be described by the Langmuir-Hinshelwood model: ClO 4 is adsorbed on to cathode surface, transfer one oxygen atom to the cathode, and then react with neighboring reactive atomic hydrogen ($H), which is an intermediate of HER. Alloy could have a higher ClO 4 reduction activity. Ni-Pt and Co-Pt outperformed pure Pt due to the enhanced production of $H (Rusanova et al., 2006;Mahmudov et al., 2008). However, long electrolysis duration (t 1/2 = 4-5 h), acidic media, and elevated temperature (>50°C) are required for electrochemical ClO 4 reduction. The sluggish cathodic ClO 4 reduction kinetics is incomparable to the fast formation rate on the anode. Cathode materials like Fe, Pd-Fe, Pd, Cu, graphene, and graphite exhibit high activity for THM and HAA reduction (Li and Farrell, 2000;Radjenović et al., 2012;Mao et al., 2016;Mao et al., 2018). The electrochemical reductive removal of THMs and HAAs is faster than ClO 4 reduction as THMs and HAAs are directly reduced by $H, and no oxygen transfer step is involved. Complete removal HAAs could be achieved within an hour. The removal of THMs is even faster due to the volatilization effect.
To the best of our knowledge, there is no study showing that cathodic dehalogenation can completely eliminate the DBPs and perchlorate produced by anode to achieve the zero-DBP discharge. Aside from the low $H production efficiency, the sluggish reduction kinetics could largely be assigned to the fact that DBPs and perchlorate are neutral or negatively charged at circumneutral pH, making their adsorption on cathode difficult. Recently Chaplin et al. developed a carbon-Ti 4 O 7 reactive electrochemical membrane (REM) cathode (Almassi et al., 2020). The REM was operated in a flow-through mode, in which water was forced to pass through the microporous cathode. The firstorder rate constant for dibromoacetic acid reduction is 9.16 min -1 , which is 57 to 1110 times higher than those obtained in batch mode (Mao et al., 2016(Mao et al., , 2018. Changing the operational mode from flow-by to flow-through dramatically reduces the thickness of the diffusional boundary layer from~100 μm for plate-type electrodes to~1 μm for REM electrodes (Chaplin, 2019), leading to a more efficient contact between target compounds and the electrode surface. Given the above, it is promising to develop tandem REM anode + REM cathode modules to remove contaminants and eliminate byproducts simultaneously.

Quench byproduct precursors
Introducing free chlorine quencher during electrolysis could suppress the formation of DBPs and ClO 4 -. Hydrogen peroxide (H 2 O 2 ) is an ideal quencher because it is effective, inexpensive, and widely used in site remediation projects. H 2 O 2 readily reacts with free chlorine to form Cland H 2 O (Connick, 1947). Our recent study (Yang et al., 2019a) found that adding 50 mmol/L H 2 O 2 can effectively inhibit ClO 4 formation in EO treatment of PFOS using BDD anode (Fig. 4). According to rxn. 4-9, ClO 4 formation could be suppressed by scavenging free chlorine and $OH. Through computational kinetic modeling, it is found that although H 2 O 2 reacts with free chlorine more slowly than with $OH (~10 4 vs. 10 7 M -1 s -1 ), the former reaction contributes the most to ClO 4 inhibition. H 2 O 2 can be in situ generated by cathode via oxygen reduction reaction (O 2 + 2H + + 2e -! H 2 O 2 ). For example, in the electro-Fenton process, H 2 O 2 is generated at carbonaceous cathodes and then react with Fe 2+ to produce $OH as an oxidant (Brillas et al., 2009). Interestingly, in the electro-Fenton process, ClO 3 -, ClO 4 and THMs were not detected (Cotillas et al., 2015). Note that this study used low current densities (0.12-2.5 mA/cm 2 ). Thus, the intrinsic formation rates of DBPs and ClO 4 might be small. Electro-peroxone process is another EO technique that involves cathodic H 2 O 2 production (Yuan et al., 2013). In this process, O 3 and O 2 are purged to the electrolyte simultaneously. Oxygen will be reduced to H 2 O 2 and then react with O 3 to produce radicals (rxn. 10).  (Lin et al., 2016). A recent study of (Yao et al., 2019) shows that H 2 O 2 could suppress the formation of TCM and HAAs during the electroperoxone treatment of chloride-containing water. These results imply that H 2 O 2 can inhibit the reactions between chlorine and organic precursors in the bulk solution.
It is important to note that the quenching effect of H 2 O 2 may be valid within a current density window. Studies showed that H 2 O 2 could suppress ClO 4 formation when the BDD anode was operated at current densities equal or below 10 mA/cm 2 (Lin et al., 2016;Yang et al., 2019a). However, ClO 4 cannot be suppressed by H 2 O 2 when current densities are above 16 mA/cm 2 (Lin et al., 2016). It seems that 10 mA/cm 2 is a critical point beyond which the benefit of H 2 O 2 addition could be attenuated.
The addition of H 2 O 2 could incur different impacts on the EO treatment efficiency, depending on the properties of target pollutants. As shown in Fig. 4, the degradation of PFOS is primarily contributed by DET oxidation. It is found that H 2 O 2 barely affects PFOS removal due to its weak affinity to the BDD surface. However, the inhibitory effect of H 2 O 2 is observed in the EO treatment of benzoic acid-a probe readily reacts with $OH. These results imply that if the target compounds are primarily removed via radical-mediated oxidation, then the addition of H 2 O 2 could be detrimental due to the radical quenching effect (rxn. 11 and 12) (Buxton et al., 1988;Yu, 2004).
Ammonium is another common chlorine scavenger. It reacts with free chlorine to form chloramines. In the chlorination disinfection process, ammonium/ammonia is intentionally introduced to produce chloramine as a disinfectant with less formation potentials of THMs and HAAs (Qi et al., 2004;Hong et al., 2013). The same strategy can be adopted in EO treatment. In the EO treatment of wastewater containing NH 4 + and Cl -(e.g., latrine wastewater), chloramines are the dominant reactive chlorine species before breakpoint chlorination Jasper et al., 2017). Negligible ClO 4 -, THM, and HAA formation was observed when chloramines prevailed (Jasper et al., 2017;Zhang et al., 2018). The high concentration of NH 4 + (ca. 30 mmol/L) in the latrine wastewater enables the efficient quenching of free chlorine. For the EO treatment of wastewater with less abundant NH 4 + , whether NH 4 + should be added intentionally should be evaluated on a case-by-case basis.

Optimize operational parameters
Acidic pH disfavors the formation of THMs and HAAs in EO treatment (Bagastyo et al., 2012;Yang et al., 2019b), likely through the inhibition of hydrolytic conversion of precursors to THMs and HAAs (Fig. 3) (Chen, 2011). The acidification of water can be readily realized without acid addition in a membrane electrolysis cell (Yang et al., 2019b). In this cell configuration, the anode is separated from cathode by an ion-exchange membrane. Wastewater in the anodic chamber will be rapidly acidified to pH < 2, mmol/L Na 2 SO 4 + 3 mmol/L NaCl electrolytes in the presence or absence of H 2 O 2 (50 mmol/L). BDD anode was coupled with stainless steel cathode and operated at 10 mA/cm 2 . Data was collected from reference (Yang et al., 2019a).
while electrolyte in the cathodic chamber will be alkalized to pH>13. Membrane assisted electrolysis was found to accelerate COD and ammonium removal and produce less THMs and HAAs than membrane-free electrolysis (Yang et al., 2019b). The acidic effluent can be neutralized by passing through the cathodic chamber for safe discharge. Attention should be paid to the control of reaction endpoints. That means overtreatment should be avoided. Take latrine wastewater treatment for example. The complete removal of COD and NH 4 + requires an electrolysis duration of 2-4 h . Under the same condition, the pathogen disinfection (5 log removal of E. coli) and the removal of trace pharmaceuticals can be readily achieved within 1 h Jasper et al., 2016). If disinfection is the primary design goal, then the EO treatment should be terminated at 1 h to avoid reaching break-point chlorination. Jasper et al. (2017) found that when breakpoint chlorination occurs, the concentrations of free chlorine rapidly increased, and such transition is reflected as the sharp increase of oxidation-reduction potential (ORP). From an engineering point of view, optimum reaction endpoints can be controlled by sensors (ORP probe, NH 4 + selective electrode, free chlorine online detector, etc.).

Process integration
Perchlorate is resistant to homogenous reduction reactions. At neutral pH, it even barely reacts with the hydrated electron, the strongest reductant known (Vellanki et al., 2013). Fortunately, perchlorate can be reduced by biological reactions. Microorganisms collected from saturated aquifer could gain the ability of ClO 4 reduction after 20 days of inoculation using emulsify oil as an electron donor (Schaefer et al., 2007). Schaefer et al. (2017) used sand columns bioaugmented with perchloratedegrading bacterium Azospira suillum to treat the effluent of the EO process. It was found that after 150 days of inoculation, sand columns achieved three order-of-magnitude removal of ClO 4 at a 9-day residence time. Although Biological processes have low capital and operational costs, the long residence time limits its treatment capacity. Recently, a series of bio-inspired Rhenium (Re) complex catalysts with high ClO 4 reduction efficiency were developed. (Liu et al., 2015 The structures of Re complex catalysts are designed to mimic the structure of molybdopterin in bacterial reductase. Using H 2 gas as an electron donor, a complete reduction of perchlorate can be achieved within 2 h. These breakthroughs enable the catalytic hydrogenation techniques to serve as posttreatment processes after EO treatment. Electrodialysis was used to removed chloride from wastewater and consequently reduced byproduct formation in the downstream EO treatment (Bagastyo et al., 2013).
However, electrodialysis reduces the overall conductivity of water, leading to higher energy consumption of EO treatment. Instead of removing chloride, transferring target compounds from wastewater to electrolytes with controlled composition might be more practical. For instance, PFAS in water can be adsorbed by anion exchange resins. It is a typical operation to regenerate the spent resins by NaCl solutions (Yu et al., 2009;Deng et al., 2010). The eluent is composed of high concentration Cland PFAS. One can envision that the EO treatment of PFAS in the resin regenerant could produce considerable amounts of byproducts. Alternatively, if resins are regenerated by Na 2 SO 4 solution, then the byproduct formation can be minimized. (Liang et al., 2018) 3.6 Feasibility analysis The section aims to address the advantages and limitations of the above control strategies in views of cost-effectiveness and system complexity. From high to low, the feasibility of strategies (by section number) is ranked in the order of 3.3>3.4>3.2>3.5>3.1. The addition of quenchers (3.3) and the control of reaction endpoints (3.4) do not require the modification of EO units. Byproduct suppression can be obtained instantly. The addition of H 2 O 2 will not affect the performance of DET oxidation of some compounds. Thus, the energy consumption of such EO system will not be increased. However, it will be increased for the removal of compounds relying on radical-mediated oxidation, as extended treatment duration will be required to offset the radical-scavenging effect of H 2 O 2 . Attention should also be paid to the elimination of residual H 2 O 2 before distribution (Barazesh et al., 2015). Anodic oxidation reactions can readily convert H 2 O 2 to oxygen (E 0 = -0.682 V NHE ). Additional EO post-treatment can be deployed, in which active electrodes should be used and operated at a low anodic potential (e.g., E < E 0 Cl 2 =Cl -: 1.36 V NHE ) to avoid the re-formation of chlorine and byproducts. Significant system modification is required to implement the strategy of cathodic dehalogenation (3.2) and process integration (3.5). Cathodic dehalogenation requires the change of HER active cathode to dehalogenation catalysts. This modification will not affect energy consumption as the water treatment efficiency is still determined by the anodic reactions. Though both treatment efficiency and effluent concentration can be well controlled by process integration, this strategy definitely increases the complexity of the treatment train, which will increase capital and operational costs. More studies are required to investigate the optimum technology combination and the seamless connection between units. The modification of anode material is identified as a strategy with low feasibility. Because DBPs are primarily derived from redox reactions in the bulk solution. The surface coatings for the suppression of DET are not stable.
Previously, a considerable amount of studies focusing on electrode material development preferred to evaluate the electrode performance in inert electrolytes (Na 2 SO 4 , NaClO 4, Na 2 HPO 4 , etc.). In future studies, it is critical to include the tests performed in chloride-bearing solutions, which is more environmentally relevant. In this case, the radical chemistry needs to be revisited. According to the reaction chains in Fig. 5, Electrodes that demonstrated high activity for $OH production in inert electrolytes might end up producing Cl 2 $and Cl$in real wastewater and surface water containing chloride (Park et al., 2009;. Because Cl 2 $and Cl$have lower redox potentials than $OH, the EO treatment efficiencies of compounds only react with $OH (e.g., nitrobenzene) could be compromised. One must also note that chloride will be converted to free chlorine at concentrations serval orders of magnitude higher than all of the radicals.  Though it is even less oxidative than chlorine radicals, free chlorine can significantly enhance the removal of some organics (e.g., phenol, salicylic acid) that are vulnerable to chlorination (Park et al., 2009;Cho et al., 2014).
We have demonstrated that H 2 O 2 at a concentration higher than 50 mmol/L can effectively control byproduct formation during the EO treatment of PFAS contaminated groundwater (Yang et al., 2019a). However, the addition of H 2 O 2 obviously deviate from the original design principle of EO as "chemical-free" processes. The addition of 50 mmol/L H 2 O 2 could incur $ 0.6-1.2 m -3 extra costs to the EO treatment (Yang et al., 2019a). Thus, the in situ generation of H 2 O 2 is desired. With this in mind, improvement of EO can be made by combining anodes with carbonaceous cathodes to in situ generate H 2 O 2 . Using pure oxygen, up to 20 wt.% H 2 O 2 can be generated electrochemically (Xia et al., 2019). The electro-Fenton or electro-peroxone reactions introduced above all used pure O 2 to maintain high H 2 O 2 concentration in the reactors. Strictly speaking, these techniques are still not chemicalfree processes as auxiliary oxygen tanks need to be deployed. Alternatively, air could be used as an oxygen source by gas diffusion electrode (GDE) (Barazesh et al., 2015). The GDE cell configuration enables the energyefficient production of H 2 O 2 at low costs because pressurized gas sources (pump, cylinder, etc.) are no longer required. Recent progress demonstrated that the air-GDE pair could produce H 2 O 2 ranges from 2 to 300 mmol/L (0.0068 wt.%-1 wt.%) (Luo et al., 2015;Barazesh et al., 2015Barazesh et al., , 2018, which are sufficiently higher than the effective 50 mmol/L concentration to inhibit perchlorate and DBP formation. Therefore, coupling air-GDE with novel anode materials is an exciting direction to advance EO technology. In this article, we place the spotlight on THMs and HAAs, which are only the tip of the DBP iceberg. We introduced that NH 4 + could suppress the formation of THMs and HAAs by converting free chlorine to chloramine. However, chloramine might react with matrix organics to form N-nitrosodimethylamine (NDMA), a carcinogenic chemical (McCurry et al., 2017;Selbes et al., 2018). Only a handful of studies reported the electrochemical treatment of NDMA (Chaplin et al., 2009;Almassi et al., 2019). So far, no study reported the formation of NDMA in the electrochemical treatment of wastewater containing both NH 4 + and Cl -. The report on the formation and transformation of other nitrogen-based and iodinated DBPs is scarce as well. More importantly, the removal of detectable byproducts does not indicate all toxic substances are removed. Given above, the author believes that challenges and opportunities down the road include: 1) develop advanced electrocatalysts and identify best practice to balance treatment efficiency and byproduct formation; 2) close the mass balance of total organic halogen formed in EO treatment by advanced massspectrometric techniques; 3) evaluate the toxicity of the EO treated effluents.