Enhanced chlorine evolution from dimensionally stable anode by heterojunction with Ti and Bi based mixed metal oxide layers prepared from nanoparticle slurry Journal of Catalysis

This study reports enhanced current (CE RCS ) and energy efﬁciency (EE RCS ) of reactive chlorine species (RCS) generation on Ir 7 Ta 3 O y anode by Ti/Bi mixed metal oxide heterojunction layers despite reductions in pseudo-capacitance and ﬁlm conductivity. In potentiostatic electrolysis of 50 mM NaCl solutions, dra- matic improvement (0.61 mmol cm (cid:1) 2 hr (cid:1) 1 at 2.5 V NHE) was noted by simple coating of thin ( ~ 2 l m) TiO 2 layer from ball-milled TiO 2 nanoparticle (80–100 nm) suspension, even with moderate elevation in voltammetric wave. Decoration of Bi 2 O 3 particles (1 – 2 l m) showed limited or adverse effects for RCS generation and stability. However, Bi-doped TiO 2 layers prepared from polyol-mediated or co- precipitation methods marked the highest CE RCS ( ~ 100%) and EE RCS (8.16 mmol Wh (cid:1) 1 at 2.5 V NHE) by increased mixing level and effective shift in surface charge. Surface (cid:3) OH exclusively mediated the RCS generation whose further transformation to higher oxide could be restrained by the heterojunction layer. Authors. Published by Elsevier Inc. ThisisanopenaccessarticleundertheCCBYlicense(http://


Introduction
Chlorine (Cl 2 ) is a widely utilized industrial reagent for polymer synthesis globally. In chemical industry, chlor-alkali process would be one of the most commercialized methods to electrolyze concentrated NaCl solutions for generation of Cl 2 , H 2 , and caustic soda (NaOH). Dimensionally stable anodes (DSAs), based on precious metal oxides such as RuO 2 and IrO 2 , have been most often used for these purposes owing to the supreme activity and stability [1][2][3]. Variable secondary components have been explored primarily for stability enhancement, to come up with the mixed metal oxide anodes, Ir 7 Ta 3 O x and Ru 3 Ti 7 O x as the optimal DSAs compositions for many decades [1,3,4].
Aqueous Cl 2 (commonly noted as free chlorine) is also a famous oxidant for drinking water disinfection and, recently, electrolytic reactive chlorine species (RCS) has been utilized for on-site treatment of saline wastewater as well [5]. Owing to the value consuming nature of environmental technologies, however, the scarcity of Ir and Ru components have substantially constrained a large scale application of the electrochemical chlorination for water treat-ment. In addition, the DSAs active for chlorine evolution reaction (ClER, E°(Cl 2 /Cl À ) = 1.36 V NHE) also effectively catalyze oxygen evolution reaction (OER, E°(O 2 /H 2 O) = 1.23 V NHE) [6], a parasitic side reaction with respect to the water treatment. A sufficient anodic potential bias on hydrous metal oxide (>MO x ) would initiate the formation of surface hydroxyl radical (>MO x (•OH), Eq. (1)) whose further oxidation could lead to higher oxide (>MO x+1 , via Eq. (2)) formation. It has been widely accepted that the overall adsorption energy of oxygen to metal (MÀO bond strength) is known to be the principal factor of OER activity mostly via Eq. (4) [7][8][9][10].
Specific group of metal oxides such as PbO 2 , SnO 2 , and TiO 2 could be characterized by a weak bond strength between metal and active oxygen to predominantly exist in the >MO(ÁOH) form which would prefer ClER due to stronger oxidation power than MO x+1 . In particular, TiO 2 is non-toxic and widely-used earthabundant catalysts to provide > Ti(ÁOH) under an anodic potential with high surface density. Nevertheless, relatively large kinetic barrier for the water discharge has limited the utilization of TiO 2 as an electrocatalyst. We previously proposed a heterojunction architecture with coating TiO 2 layers with or without mixing Bi (Bi x Ti 1Àx O z ) on top of Ir 7 Ta 3 O y DSA for an enhanced RCS generation in dilute aqueous solutions [11]. In these configurations, conductor-like nature of IrTaO y could serve as an ohmic contact to Ti substrate [11,12], while surface hydrous TiO 2 (>Ti(OH -)) provided elevated quasi-stationary concentration of >MO x (•OH). In spite of limited consideration for limited active sites and different kinetic parameters to predict selectivity in parallel reactions, density functional theory (DFT) calculations on model RuO 2 /TiO 2 architecture [13] further suggested that the TiO 2 heterojunction layer could shift locations in volcano plots of OER and ClER, which in-turn influences the ClER selectivity. In addition, mixing Bi in the outer layer was evinced to increase electrostatic sorption of anions (Cl À ). In our earlier contribution, nevertheless, the peroxo-route for aqueous Ti-glycolate complex and solution casting of the Bi x Ti 1Àx O z layer were rather dangerous and labor consuming.
This study reports simplified preparation methods for coating TiO 2 based heterojunction layers (on Ir 7 Ta 3 O y DSA) which was either decorated by Bi 2 O 3 micro-particles or mixed with Bi 3+ . In particular, employing nanoparticle slurry precursors was expected not only to augment the catalytic edge sites [3], but also to allow an effective passivation owing to greater viscosity than aqueous precursor [14]. In dilute (50 mM) NaCl solutions, the activity and selectivity for ClER were comparatively evaluated for the heterojunction anodes with variable outer layer loading (thickness) and mixing levels of Bi. In addition, the roles of >MO x (•OH) and >MO x+1 on the RCS generation were interrogated by correlation with formate ion degradation.

Preparation and characterization of the anodes
All chemical reagents were analytical grade and obtained from Sigma-Aldrich or Daejung chemicals to be used without further purification. Solutions were prepared with Millipore-Q water (Millipore) with a specific resistivity of 18 MX cm À1 . For pretreatment of the anode base, Ti foils (99.5% Alfa-Aesar) were cleaned with SiC paper, sonicated in organic solvent solutions (equivolumic deionized water, ethanol, and acetone) for 0.5 h, and immersed in 10 wt% oxalic acid for 50 min at 80°C. The IrTaO y layer was prepared by soaking the Ti substrate (3 Â 2 cm 2 , distance = 5 mm) into solutions with 73 mM H 2 IrCl 6 , and 27 mM TaCl 5 . 4 N HCl and equivolumic ethanol/isopropanol solutions were compared as solvents. The loaded precursor was sequentially dried at 80°C and annealed at 525°C for 10 min, which was repeated five times before final annealing at 525°C for 1 h. For preparation of outer heterojunction layers, three dif-ferent types of precursor were prepared as follows. A) P90 TiO 2 (Evonik Industries) and Bi 2 O 3 particles separately ball-milled (Ultra Apex mill) in ethanol to have particle size range of 80-100 nm and 1-2 lm, respectively. The TiO 2 nanoparticles (3.8 wt%) were dispersed in ethanol/terpineol (16:19 wt%) solution with added ethyl cellulose binder (2.2 wt%). Bi 2 O 3 microparticles were added with variable Bi to Ti ratios (0:10, 1:9, and 3:7). B) 6 mL titanium butoxide was added to 150 mL ethylene glycol to be stirred for 8 h in room temperature. The mixture was poured into 510 mL acetone and 8 mL of deionized (DI) water was added before vigorous stirring for 1 h. Resulting precipitate was collected using a centrifuge (6000 rpm for 4 min), washed with ethanol 5 times, and re-dispersed in ethanol with a given concentration of bismuth citrate. C) titanium tetraisopropoxide (1.25 mL) in ethanol (25 mL) was added to 8 mM Bi (NO 3 ) 3 solution (pH 1.5 adjusted by acetic acid) by dropwise and the mixture was stirred overnight. Resulting precipitate was collected (11000 rpm for 10 min), washed with DI water 5 times, and re-dispersed in DI water. These precursors were sprayed on the IrTaO y layer and annealed at 450°C for 30 min and the coating procedure was performed for 3 times. In this study, heterojunction anodes from precursor A were denoted as (Bi 2 O 3 ) x (TiO 2 ) 1Àx (x = 0, 0.1, and 0.3), while the outer layers with low mass loading (L) were prepared without repetition to compare the performance with analogues with high mass loading (H). The anodes from precursor B and C with nominal Bi to Ti ratio of 3:7 were named as Bi 3 Ti 7 O x -1 and Bi 3 Ti 7 O x -2, respectively, to describe greater mixing level of Bi.
The surface morphologies were observed by using a ZEISS 1550VP field emission scanning electron microscope (SEM) and elemental compositions were estimated using an Oxford X-Max SDD X-ray energy-dispersive spectrometer (EDS) system. The EDS analysis was performed either in a point-and-identification mode for 10 arbitrary sites or in mapping mode. The surface crystallography was assessed by X-ray diffraction (XRD) using an X'pert MD (Panalytical) diffractometer with CuÀK radiation.

Electroanalysis.
The electrochemical activities of (Bi 2 O 3 ) x (TiO 2 ) 1Àx and Bi 3 Ti 7 O x -1/2 heterojunction anodes were assessed using linear sweep voltammetry (LSV) and cyclic voltammetry (CV). In a single compartment cell (working volume = 60 mL), each anode was parallel matched with AISI 304 stainless steel cathode (distance = 5 mm) with active geometric surface area of 3 Â 2 cm 2 . Ag/AgCl/sat. KCl reference electrode (BASI Inc.) was located 3 mm away from the anode center. The three electrodes configuration was connected to a potentiostat (SP-50, Bio-Logic) to control anodic potential (E a ) in normal hydrogen electrode (NHE) scale (E a (NHE) = E a (Ag/ AgCl) + 0.197 V). The CV and LSV data were recorded with E a ranges of 0.2-1.0 V NHE (scan rate of 20 mV s À1 ) and of 0.8-2.0 V NHE (scan rate of 5 mV s À1 ), respectively. Before all electrochemical experiments, open circuit potentials were measured for 30 min and ohmic resistances were measured by current interruption method at 200 mA current bias.

Potentiostatic Electrolysis.
The efficiency of RCS generation was evaluated during potentiostatic electrolysis of 50 mM NaCl solutions at variable E a values (2.0, 2.5, and 3.0 V NHE). The evolution of [RCS] was measured by DPD (N, N-diethyl-p-phenylenediamine) reagents 3 times at 2 min intervals and, during this period, further oxidation of RCS to ClO 3 À or ClO 4 À could be negligible [15]. The specific rate (SR), current (CE) and energy efficiency (EE) of the RCS generation were estimated by the following equations (Eqs. (7)- (9)).
In order to assess the roles of >MO x (ÁOH) on RCS generation, formate ion was employed as the ÁOH probe compound. The potentiostatic electrolysis of 50 mM NaCOOH solutions was performed for 2 h at E a 2.0 or 2.5 V NHE. The [HCOO -] of samples were periodically quantified by ion chromatography (IC, Dionex, USA) with an anion-exchange column (Ionpac AS 19).

Physico-chemical characteristics of (Bi 2 O 3 ) x (TiO 2 ) 1Àx heterojunction layers
For the heterojunction electrodes, the underlying IrTaO y layer was synthesized using an organic solvent (ethanol/isopropanol solutions), due to more uniform surface coverage and moderately greater electrochemically active surface area (ECSA) than analogous from hydrochloric acid solvents ( Figure S1 and S2). SEM images in Figure S3 show the horizontal surface morphology of the (Bi 2 O 3 ) x (TiO 2 ) 1Àx heterojunction anodes with variable molar ratios of Ti to Bi and catalysts loading (thickness). Cross-sectional SEM images estimated the average thickness of (TiO 2 )-L and (TiO 2 )-H to be ca. 2 and 5 mm, respectively. (TiO 2 )-H and -L showed marginal numbers of crack and pinhole that are typically observed for thermally decomposed (mixed) metal oxide electrodes. Since P60 TiO 2 nanoparticles already underwent annealing, further thermal expansion of the crystalline TiO 2 particles could be limited, while the aggregation of nanoparticles would alleviate the topological distortion. For samples with added Bi 2 O 3 particles in precursor solutions, discrete islands of bismuth oxide scattered on the surface were noted. EDS mapping in Figure S3g-h clearly identified the immobilized particles (diameter~1 lm) to be Bi 2 O 3 .
In the absence of Bi mixing ((TiO 2 )-L and -H), the primary XRD peaks in Fig. 1 (individual patterns in Figure S4) commonly involved standard patterns of anatase TiO 2 (2h = 25.2, 37.7, and 47.9°) and less prominently of rutile TiO 2 (2h = 27.5, 36.1, and 54.4°), being in compatible with the composition of precursory TiO 2 particles (92:8 wt% of anatase:rutile for P60). A phase transition during the thermal annealing at 450°C would be limited since transformation into the rutile structure is known to occur at a temperature normally exceeding 600°C [16]. Negligible characteristic pattern of IrO 2 (2h = 34.7, 40, 54, and 69.3°) on surface of these anodes indicated relatively thorough covering even with single spray coating. In comparison, (  [17]. Therefore, each metal oxide phase was expected to maintain the intrinsic physicochemical properties with minimal doping effects; thus, the outer heterojunction layers could be denoted as (Bi 2 O 3 ) x (TiO 2 ) 1Àx . In the presence of Bi 2 O 3 particles, sig-nals from underlying IrO 2 (peak near 35°) and Ti substrate (peak near 40°) were invigorated. Considering that the XRD analysis was performed after the electrochemical analyses described hereinafter, these observations suggest the instability of bismuth oxide in anodic environment (vide-infra).

Voltammetric responses of IrTaO y /(Bi 2 O 3 ) x (TiO 2 ) 1Àx heterojunction anodes
Areal capacitance from CV within the potential window of water splitting has often been utilized to estimate the ECSA [18,19]. Fig. 2a (raw CV data in Figure S5) illustrates the areal capacitance measured under anodic potential between 0.2 and 1.0 V NHE in 50 mM NaCl solutions (circum-neutral pH). The highest areal capacitance (22 mF cm À2 ) of IrTaO y was reduced by about 30% upon addition of surface TiO 2 layers. Mixing Bi 2 O 3 particles on surface insignificantly affected or even further decreased the areal capacitance. Within the potential scan range, oxidative transformations of hydrous > TiOH or > BiOH (e.g., to oxyhydroxide or higher oxide) could be limited to especially due to the fully oxidized Ti 4+ . Nevertheless, considering rather thorough surface passivation by TiO 2 layers (at least for IrTaO y /TiO 2 anodes as shown in Fig. 1), the electrochemical activity for the heterojunction anodes would be ascribed to a thermal diffusion of Ir components into the upper layer (vide infra) [11].
Linear sweep voltammetry (scan range: 0.8 -2.0 V NHE, scan rate: 5 mV s À1 ) in 50 mM NaCl solutions (Fig. 2b) (Table 1). In these experimental conditions, the current generation would be ascribed to parallel OER and ClER. Electrical conductivity of semiconductor electrocatalysts should be another consideration especially for thick catalyst film. To this end, the outer heterojunction layer mostly lowered the current generation due to the inherently poor electrical conductivity of TiO 2 and Bi 2 O 3 . More pronounced current decline with included Bi 2 O 3 particles, particularly with higher loading ((Bi 2 -O 3 ) 1 (TiO 2 ) 9 -H and (Bi 2 O 3 ) 3 (TiO 2 ) 7 -H), agreed with the concurrent increase in ohmic resistance measured by a current interruption method. The current interruption would collectively evaluate the electrolyte resistance of ion migration, polarization resistance and resistance across the catalytic film, often described as resistances in series in an equivalent circuit [20]. Since the Bi 2 O 3 particles were speculated to be anchored within TiO 2 matrix without a direct contact with IrTaO y layer, (Bi 2 O 3 ) x (TiO 2 ) 1Àx anodes would suffer from invigorated resistance across the multiple junctions (IrTaO y /TiO 2 /Bi 2 O 3 ).
Nevertheless, the observed OER and ClER activities were incompletely justified with the areal capacitance and ohmic resistance, particularly for the unexpected elevated current generation from TiO 2 -L anode. A recent report presented evidences that the OER and ClER overpotential values of IrO 2 could be reduced upon a few cycles of atomic layer deposition (ALD) of thin TiO 2 layer by shifting surface charge density and overall M-O bond strength [21]. Underlying Ir 4+ was speculated to be partly oxidized to result in sub-stoichiometric TiO 2Àx outer layer, significantly elevating the electrical conductivity. Even for much thicker TiO 2 coating (greater than1 lm) in this study, Ir ions might be diffused into the TiO 2 matrix during the thermal annealing, leading to a vertical concentration gradient of Ir (in mixture with Ti). Effective ionic radius of Ir 4+ (62.5 pm) is more similar with Ti 4+ (60.5 pm) than Bi 3+ (103 pm), supporting the observed enhancement only with the TiO 2 -L layer. For the TiO 2 -H heterojunction anode, more pronounced increase in ohmic resistance led to net reduction in anodic wave, as in the case of far thicker (~30 mm) TiO 2 layers on IrTaO y in our previous report [11].

Reactive chlorine generation by IrTaO y /(Bi 2 O 3 ) x (TiO 2 ) 1Àx heterojunction anodes
Reactive chlorine generation was comparatively evaluated in 50 mM NaCl solutions (Fig. 3) at variable applied anodic potentials (2.0, 2.5, and 3.0 V NHE). Throughout this study, the corresponding cell voltage and ohmic resistance (measured by current interruption at 200 mA) ranged 4-6 V and 3.3-5.6 X, respectively. The current density in potentiostatic condition (Fig. 3a) was in general agreement with the LSV shown in Fig. 2b. EE always decreased with the greater applied potential (cell voltage) due to the increment in ohmic drop, whereas CE varied insignificantly. Despite moderate variations in ohmic drop and iR-compensated anodic potential among anodes, the insignificant dependency of CE on the applied potential would allow a qualitative comparison. The CE and EE values of the control IrTaO y electrode were averaged to 52% and 4.23 mmol Wh À1 at 2.5 V NHE, respectively, while the heterojunction layers substantially augmented the CE and EE for RCS generation (in Fig. 3b and d). Accordingly, despite the general reductions in the operational current density, heterojunction anodes marked greater specific RCS generation rates (Fig. 3c) than the control. The most remarkable enhancement was noted for TiO 2 -L with more than 70% increase in SR RCS , due to the elevated voltammetric response. Utilizing nanoparticle slurry precursor with relatively simple preparation procedure, these results outperformed our previous report on IrTaO y /Bi x Ti 1Àx O z heterojunction anodes by casting aqueous Ti-glycolate solutions prepared by a peroxo-route [11].
It would be a rational postulation that the RCS generation through charge transfer between the physi-sorbed •OH and Cl À (Eq. (5)) would be far more facile than oxygen atom transfer from higher oxide (Eq. (6)) [5]. In comparison, the >MO x+1 form would more preferentially mediate OER (Eq. (4)) than (>MO x (•OH), Eq.
(3)) [10]. Therefore, the speciation of hydrous metal oxide under an anodic bias (MO x (•OH) versus MO x+1 ) is expected to determine the selectivity between OER and ClER [22,23]. More predominant steady-state surface concentration of >MO x (•OH) would account for the overall enhancements of CE and EE for RCS generation on heterojunction architectures, as suggested previously [11,13]. In order words, primary Ti 4+ species on surface of heterojunction anodes would prevent transformation into the higher oxide, whereas the control IrTaO y would favor >MO x+1 formation by Ir 4+ /Ir 6+ transition.   [11] proposed competitive roles of Bi mixing for ClER on the solution casted IrTaO y /Bi x Ti 1Àx O z heterojunction anodes; accelerating interaction with Cl À (via a positive shift in surface charge) but hampering ClER kinetics through the higher oxide formation (via Bi 3+ /Bi 5+ redox transition). Accordingly, CE values of ClER were noted to be maximized at Bi fraction of 0.1-0.3 [11]. In this study, the former effect was insignificant, most presumably because of spatially separated TiO 2 Figure S6a illustrates the concentrations of Bi and Ti in electrolyte during an accelerated life test at 1000 mA cm À2 operation for the (Bi 2 O 3 ) 3 (-TiO 2 ) 7 -H heterojunction anode. Bi 2 O 3 particles were expected to maintain intrinsic properties so as to be vulnerable to anodic potential bias and local acidity from OER, as confirmed by a distinct accumulation of Bi in electrolyte. EDS mapping ( Figure S6b-f) after the life test also located intensified signals of Ir on remains with dissolution/detachment of Bi 2 O 3 particles. The concurrent rise of Ti concentration would be associated with detachment of TiO 2 nanoparticles along with the collapse of particle aggregation. In summary, decoration of micron-sized Bi 2 O 3 particles on the TiO 2 heterojunction brought about limited or even adverse effects both for RCS generation and stability.

Reactive chlorine generation by IrTaO y /Bi 3 Ti 7 O x heterojunction anodes
In order to address the drawbacks of Bi 2 O 3 decoration, this study evaluated two different Bi 3 Ti 7 O x heterojunction layers with more uniform doping of Bi into the TiO 2 matrix. For Bi 3 Ti 7 O x -1 anode, titanium glycolate nano-particle slurry was prepared by polyol-mediated synthesis [14], to be mixed with a bismuth citrate solution (Bi:Ti = 3:7) overnight. During this process, Bi 3+ was expected to be randomly substituted with Ti 4+ in the polyol structure by ligand exchange to form a homogeneous solid mixture during the subsequent thermal decomposition. In addition, a wellestablished co-precipitation procedure for metal ion doped nanosized TiO 2 colloidal suspension [24] was employed for preparation of Bi 3 Ti 7 O x -2 anode. Under the same spraying procedure, the mass loading of Bi 3 Ti 7 O x -1 layer was far greater (3.20 mg cm À2 ) than Bi 3 -Ti 7 O x -2 (0.342 mg cm À2 ) due to the different precursor viscosity (aqueous versus organic solvent).
The SEM images on the horizontal surface of Bi 3 Ti 7 O x anodes ( Figure S7) showed rugged morphology without specific grains of Bi 2 O 3 . The outer film contained multiple cracks that have been typically found for annealed mixed metal oxide due to the variable thermal expansion coefficients [25]. In addition, XRD profiles ( Fig. 4 and S8) for the both Bi 3 Ti 7 O x anodes showed insignificant peaks relevant to Bi 2 O 3 . These observations would substantiate homogeneous mixed metal oxide formation by the modified preparation schemes. Besides reflections from Ti substrate, intense signals from rutile TiO 2 dominated on Bi 3 Ti 7 O x -2. Insertion of Bi 3+ into the TiO 2 lattice would distort the crystalline structure to facilitate transformation of anatase to rutile at relatively low annealing temperature (450°C) [16]. At superimposable locations, by far smaller peaks were noted on Bi 3 Ti 7 O x -1, suggesting predominant amorphous phase. Evidences were presented that polyolmediated Ti-glycolate could remain amorphous up to 500°C [14].
CV presented in Figure S9a estimated areal capacitance to be 14.91 mF cm À2 and 21.96 mF cm À2 for Bi 3 Ti 7 O x -1 and À2, respectively. Moderately lower capacitance of Bi 3 Ti 7 O x -1 would be associated with the greater mass loading of outer layer and film resistance. During the LSV in 50 mM NaCl solutions ( Figure S9b), anodic current waves on Bi 3 Ti 7 O x anodes outperformed (Bi 2 O 3 ) 3 (-TiO 2 ) 7 -H anodes and the voltammogram of Bi 3 Ti 7 O x -1 was comparable with TiO 2 -H. Despite the lack of conclusive evidence, coordinately unsaturated active sites in amorphous structure (Bi 3 -Ti 7 O x -1) could moderately facilitate the kinetics of OER and ClER. Fig. 5 shows that both Bi 3 Ti 7 O x anodes marked CE RCS near unity (93.6% and 98.3%, on average). Consequently, the IrTaO y /Bi 3 Ti 7 O x heterojunction anodes marked the highest EE RCS (8.02-8.16 mmol Wh À1 at 2.5 V NHE) among the anodes interrogated in this study. These electrodes even outperformed a commercial Ir based DSA (De Nora) which showed lower values of CE RCS (40.9% on average) and EE RCS (3.1 mmol Wh À1 at 2.5 V NHE) as shown in Figure S10. These results demonstrated that even with the analogous composition, mixing level of constituents could significantly affect the electronic and electrostatic properties. In particular, relatively homogenous intercalation of Bi component might avoid the ohmic resistance across multiple junctions, while increase the point of zero charge for enhanced electro-sorption of Cl À . Furthermore, accelerated life test at 1000 mA cm À2 indicated insignificant dissolution of Bi and Ti during operation for 30 h (data not shown).
In order to evaluate the effects of ion migration upon the potential bias (increasing [Cl À ] in the anode vicinity), the RCS generation experiments were performed in 50 mM NaCl solutions with or without another electrolyte (50 mM NaClO 4 ) under a galvanostatic condition ( Figure S11). The measured values of SR RCS , CE RCS , and EE RCS in the mixed solutions were 82, 86, and 97% of those in NaCl solutions, respectively. Moderate reductions could be associated with the competing ion migration and lower potential bias in the presence of other electrolyte, which would be a consideration to understand the performance in real (waste)water matrix. On the other hand, one may argue that the elevated CE RCS by the heterojunction architecture would be ascribed to the reduced current density, since an increasing current could invigorate OER only when ClER is rate-limited by diffusion of Cl À . However, the operational current density in this study would be lower than the limiting current density for ClER, as corroborated by Figure S12. The observed CE RCS values showed negligible correlation with the current density. Therefore, the shift in ClER selectivity should be understood by speciation of surface intermediates with variable reactivity with Cl À (vide infra).

Formate ion degradation versus RCS generation to assess intermediate speciation
In this study, the hydroxyl radical (•OH) on anode surface could be a nonselective oxidant to be utilized for Cl À oxidation to RCS. Due to the short lifetime (10 À6 À 10 À3 s), free or bound •OH has been quantified using an array of probe compounds such as salicylic acid, benzoic acid, coumarin, benzoquinone, and RNO (pnitrosodimethylaniline) [26]. Formate ion could be another surface •OH prove due to superb reactivity with •OH [27], simple quantification method using IC, and relatively low molecular weight to avoid diffusion limitation.
In order to further interrogate the speciation between MO x (•-OH) and MO x+1 and their roles on RCS generation, potentiostatic (E a = 2.0 or 2.5 V NHE) formate ion degradation experiments (initial concentration = 50 mM) were performed using the series of (Bi 2 -O 3 ) x (TiO 2 ) 1Àx and Bi 3 Ti 7 O x heterojunction anodes ( Figure S13). We limited the applied anodic potential up to 2.5 V NHE due to slightly lower electrical conductivity of 50 mM NaCOOH solutions (4.3 mS cm À1 ) than 50 mM NaCl solutions (5.4 mS cm À1 ). The differences in ohmic drop between two solutions were estimated to be 0.16 V at maximum. The reaction between HCOOand > MO x (-ÁOH) could be approximated to pseudo-first-order owing to the facile kinetics (Eq. (10)).  (6)) would play a minor role for our heterojunction anodes. In spite of previous reports [28] on HCOOH degradation by IrO 2 /IrO 3 couple in acid (pH 0), oxygen atom transfer from MO x+1 to HCOOcould be excluded in the current experimental conditions. Ozone, as a presumed analog of the higher oxide, is known to undergo sluggish kinetics both with Cl À (<3 Â 10 À3 M À1 s À1 ) and HCOO -(1 Â 10 2 M À1 s À1 ) [29,30].
Provided that free and bound •OH react with HCOOat analogous bimolecular rate constants, the steady-state concentration of surface bound •OH ([>MO x (•OH)] SS ) could be estimated from k obs formate using eq. (11).
[>MO x (ÁOH)] SS As shown in Figure S14a, the [>MO x (•OH)] SS estimates were in similar order with our previous study [11]. Eq. (12) given by a pseudo-steady-state approximation further estimated the rate constant for Eq. (2) (k 2 ), assuming that >MO x (•OH) and >MO x+1 are exclusively responsible to the formate ion oxidation and OER, respectively.
where CE OER is the CE of OER calculated from the charge balance (1 -CE of formate ion degradation). Figure S14b substantiated that the outer heterojunction layers significantly lowered k 2 values of IrTaO y . Under further assumption that transition from >MO x (•OH) to >MO x+1 (Eq. (2)) is limiting the OER rate, CE RCS would be determined by the ratio of k 2 to k obs chloride . Despite the apparent inverse relation between k 2 and CE RCS noted in Fig. 6b, relatively weak correlation suggested the liberation of O 2 from >MO x+1 (Eq. (4)) could also determine the OER rate depending on the outer layer composition. For example, a strong binding of O on Bi might allow the ClER to overwhelm OER on the Bi 3 Ti 7 O x -2 anode with relatively high k 2 value.

Conclusions
This study prepared Ir 7 Ta 3 O y DSAs with outer heterojunction layers based on mixed Bi and Ti oxide to compare the RCS generation efficiency during competitive OER and ClER in dilute (50 mM) NaCl solutions. Crack free, anatase dominant TiO 2 layers prepared from ball-milled P60 nanoparticles brought about significantly enhanced SR RCS , CE RCS , and EE RCS , while TiO 2 -L even elevated the anodic current wave possibly owing to a tuned overall M-O bond strength. However, decoration of Bi 2 O 3 particles on the outer TiO 2 films ((Bi 2 O 3 ) x (TiO 2 ) 1Àx ) gave limited or adverse effects on voltammetric response, RCS generation, and stability presumably  , respectively. Parallel formate ion degradation experiments revealed that the RCS generation would be exclusively mediated by >MO x (ÁOH) whose further transition to >MO x+1 was proven to be retarded by the heterojunction layer. Our heterojunction anodes with simplified preparation procedure would allow an energy-efficient electrochemical oxidation processes for water disinfection and high salinity wastewater treatment through the enhanced chlorine evolution. Improvements in the industrial chlorine generation (chlor-alkali process) are also expected with reduced power input and operational cost. Further study should tackle the effects of outer layer TiO 2 loading (thickness) on film resistance and CE RCS , as well as stability of the Bi-doped TiO 2 layers in more systematic manners based on periodic CE RCS measurements during accelerated life test.