Introduction

Chlorine (Cl2) gas is an essential chemical with an annual production of over 70 million tons for a wide range of applications in water treatment, organic synthesis, and polyvinyl chloride production1,2,3. It is mainly produced through electrochemical chlorine evolution reaction (CER)4 in the chlor-alkali industry5,6,7, where the oxygen evolution reaction (OER) can compete with CER in aqueous electrolytes owing to their comparable onset potentials and similar types of catalytic active sites8,9,10, thereby it is a great challenge to ensure the highly selective production of Cl2 gas. The state-of-the-art industrial electrolyzers for CER generally use relatively acid-resistant mixed metal oxides based on high-cost precious metal (Ru or Ir)11,12, which are highly active for OER as well. To ensure the high efficiency in the CER system and improve the Cl2 selectivity, a high molar concentration of Cl- ion (5.0 M) and acidic pH conditions are required13,14,15. However, for practical applications, Ru-based CER catalysts are still insufficiently stable under long-term operation due to the formation of soluble ruthenium chloride16. Hence, some additives or dopants, such as the SnOx, were used to improve the activity as well as stability of mixed metal oxides-based CER catalysts over the past decades17,18,19.

The CER occurs under harsh reaction conditions according to the Pourbaix diagram of an aqueous saline electrolyte20,21,22, especially at pH <3 with a highly polarized potential range, unavoidably leading to the dissolution of most metal oxide materials. Recently, low-cost non-precious metal oxides, such as highly crystalline Co3O4 and transition-metal antimonates23,24,25,26, have been tested as CER catalysts, which display decent stability since their dense, thick, and crystalline natures somehow prevent their complete dissolution in corrosive and acidic electrolytes. By contrast, an amorphous nonprecious metal oxide film can dissolve in such electrolytes in tens of minutes27. As a result, the intrinsic electrochemical surface amorphization upon anodic polarization is very detrimental to the catalytic stability but technically inevitable. Despite the beneficial effect of enriching the surface-active sites through amorphization, it remains a huge challenge to exploit amorphous non-precious metal oxide catalysts, like the alkaline OER catalysts, toward highly active, selective, and sustainable CER.

In this work, we present a self-adaptive amorphous CoOxCly catalyst, which was in situ deposited on a bare F-doped tin oxide (FTO) electrode at pH ≤ 2 in saline solutions containing Co2+ and Cl- ions. The final state of the deposited film catalyst depends on a given electrochemical condition, including the applied potential, solution pH, and electrolyte concentration28. We show that CER takes place heterogeneously on the in situ deposited CoOxCly catalyst. We find that an anodic polarization at ≥ 1.67 V (versus RHE if not otherwise noted) leads to a moderate deposition rate of CoOxCly and the Cl- in electrolytes aids the co-deposition of Co and Cl into the catalyst film. Experimental results and density functional theory (DFT) calculations show that the introduction of Cl- can share the oxidative equivalent under highly anodic potential and inhibit the oxidation of Co to a higher level for OER, leading to increased selectivity for CER. Online differential electrochemical mass spectrometry (DEMS) and rotating ring-disk electrode (RRDE) verified the efficient CER with ~100% selectivity in 0.5 M Cl- brine solution. A certain time of catalyst deposition at 1.67 V leads to the CER activity increasing to around 10 mA cm−2 and afterwards a slight increase over 500 h.

Results and Discussion

Screening out metal cations for CER in acidic brine

We first screened several transition metal cations, such as nickel (Ni), manganese (Mn), iron (Fe), copper (Cu), and cobalt (Co) at pH 2 for CER. We demonstrated that adding Cl- ions into Mn, Ni, and Cu-containing brine solutions did not obviously shift the onset potential relative to the Cl--free electrolytes with the same ionic activity. Interestingly, a substantial decrease in the onset potential and an evident increase in current density were observed in the saline solution containing 0.001 M Co2+ or Fe3+ and 0.5 M Cl- relative to that containing 0.001 M Co2+ or Fe3+ and 0.5 M ClO4- (Supplementary Fig. 1). However, although the Fe-catalyst displayed initial CER activity, the subsequent electrodeposition was not favorable, regardless of either increasing Fe3+ or Cl- concentration thus being precluded (Supplementary Note 1 and Supplementary Figs. 2–6). In contrast to the large onset difference between CER and OER over deposited Co-catalyst, the dimensionally stable anode (DSA) and RuO2 catalysts loaded on FTO showed close onset potentials in 0.5 M NaCl electrolyte for CER relative to the Cl--free electrolyte for OER (Supplementary Fig. 7). Meanwhile, we precluded the anion influence from Co precursors on cyclic voltammograms (Supplementary Fig. 8). In addition, we found that the current densities increase with the concentration of either Co2+ or Cl- (Supplementary Fig. 9). The X-ray fluorescence (XRF) Co Kα mapping showed that the amount of deposited Co sites increases with the Cl- concentration, suggesting the interplay between Co2+ and Cl- (Supplementary Fig. 10).

The linear scan voltammetry (LSV) curve after 2 h of electrodeposition at 1.67 V in Cl--containing electrolyte was used as an example to exhibit a substantial increase in current density and a downshift of onset potential over 400 mV at 10 mA cm−2 relative to that of Cl--free electrolyte (Fig. 1a and Supplementary Fig. 11). The in situ deposition of CoOxCly at 1.67 V was accompanied by Cl2 production (Supplementary Fig. 12), as well as reduced charge transfer resistance (Supplementary Fig. 13). As expected, a stepwise increase in potentiostatic time brought in an increased mass loading of the Co sites, which was verified by the time-dependent Co Kα XRF test (Supplementary Fig. 14). Accordingly, the continuous increase of active Co sites led to a gradual attenuation of overpotential from ~0.4 V after the first potential scan to finally ~0.1 V (Fig. 1b). During this process, the catalyst film presents an epitaxial-growth-like behavior, of which the applied potential at the electrode/electrolyte interface can oxidize the Co2+ for the deposition of outermost catalyst layer until the interfacial potential drop across the catalyst29,30 cannot support the subsequent oxidation-deposition of Co2+.

Fig. 1: The CER performance of amorphous CoOxCly catalysts.
figure 1

a The LSV curve of CoOxCly and CoOx catalysts in 0.5 M NaCl and NaClO4 at pH 2, respectively. The CoOxCly and CoOx catalysts were obtained after 2 h electrodeposition at 1.67 V in the electrolyte containing 0.1 M Co2+ and 0.5 M Cl- or 0.5 M ClO- at pH 2, respectively. 1.67 V was selected based on the CER onset potential of the first potential scan. b The overpotential trend at 10 mA cm−2 with increasing the electrodeposition time from the first potential scan to 500 h at 1.67 V in the electrolyte containing 0.1 M Co2+ and 0.5 M Cl- at pH 2. The overpotential was iR corrected with 90% ohmic resistance. The electrolyte resistance was 46.0 Ω. c The Cl2 selectivity of CoOxCly catalyst under different Cl- concentrations and anodic potentials during the deposition process at pH 2. The data were recorded based on a rotating disk electrode method at 1600 RPM. The error bars represent the standard deviation for triplicate measurements. d The LSV curves and corresponding DEMS signals during CER at pH 2 at 10 mV s−1. e The performance comparison between different CER catalysts (the detailed data were summarized in Supplementary Table 1).

Full size image

Electrochemical CER selectivity and stability

Achieving high selectivity for CER is crucial due to the competing OER31,32,33. Thus, we evaluated the CER selectivity based on the rotating ring-disk electrode (RRDE), where the generated Cl2 on the disk catalyst electrode was quantified through electrochemical reduction on the ring electrode (Supplementary Fig. 15 and Supplementary Note 2). Through comparing the onset potentials for both O2 and Cl2 reduction on the Pt ring electrode based on the established method34, 0.95 V has been applied for Cl2 reduction and quantification (Supplementary Fig. 16 and Supplementary Note 3). The CER selectivity was evaluated by the chronoamperometry method at indicated potentials using RRDE35 that was conducted in an Ar-saturated electrolyte containing 0.1 M Co2+ with various concentrations of Cl- at pH 2 at 1600 RPM (Supplementary Figs. 17–21). Significantly, in acidic saline containing 0.1 M Co2+ and 0.5 M Cl-, a CER selectivity of ~100% at 1.67 V was achieved (Fig. 1c). It is worth noting that the Cl2 selectivity already reached 90 ± 1.5% at 1.67 V when the Cl- concentration is as low as 0.01 M. Importantly, in 0.5 M Cl-, the CER selectivity can easily achieve 100% at a wide range of potentials. These results are in line with the substantially reduced onset potential for CER in comparison to that for OER (Fig. 1a). Compared to the CoOxCly catalyst, the selectivity of DSA (~80%) and RuO2 (~88%) catalysts were distinctly low even in 0.5 NaCl electrolyte (Supplementary Fig. 22).

To further prove the CER selectivity, online DEMS was applied to discriminate the CER and OER36,37,38. The DMES experiment was implemented in a three-electrode H-cell in acidic saline solutions containing 0.1 M Co2+ and 0.5 M Cl- (Supplementary Fig. 23). Significantly, as shown in Fig. 1d, the DEMS data clearly exhibit that the 70Cl2 gas was the only product in 0.5 M Cl- electrolyte during the CER process, and no other signals were detected, such as 1H16O35Cl, 35Cl32O2 or 35Cl16O (Supplementary Fig. 24). In contrast, in the absence of Cl-, the 32O2 was the exclusive product (Supplementary Fig. 25).

In addition, it is well accepted that, for non-noble metal oxides-based CER catalysts, the activity decay is also a major issue in acidic electrolytes1,7. A highly crystalline and thick Co3O4 catalyst film prepared through calcination was used as a control39, and we found that its current density quickly declined from 5.0 to 0.87 mA cm−2 (Supplementary Fig. 26). This activity decay resulted from the leaching/dissolution of Co2+ from the surface of Co3O4. Similarly, the noble metal oxides-based DSA and RuO2 catalysts exhibited a fast descent in current density. In contrast, the current density of the CoOxCly catalyst deposited at 1.67 V increases to 10 mA cm−2 and then stabilizes at about 15 mA cm−2 over 500 h test, which leads to a downshift of potential ~ 300 mV at 10 mA cm−2 relative to the initial CV activity (Supplementary Fig. 27). Meanwhile, the electrochemical stability of the CoOxCly at 250 mA cm−2geo could sustain over 20 h (Supplementary Fig. 28). This catalyst system could even outperform the representative precious and non-precious metal oxide catalysts (Fig. 1e and Supplementary Table 1).

Cl־-induced deposition of amorphous CoOxCly

To understand the catalyst deposition coupled CER process, we characterized the composition and structure of the deposited catalyst film. Compared to the CoOx film, the Co Kα and Cl Kα mapping of CoOxCly catalyst electrode by XRF show that the Co and Cl elements were distributed throughout the entire catalyst film, indicating the co-deposition of Cl- together with Co2+ (Fig. 2a and Supplementary Fig. 29). The energy-dispersive X-ray spectroscopy (EDX) mappings confirmed the presence of the relatively even distribution of Co, Cl, and O for the formation of CoOxCly film even after 500 h stability test (Fig. 2b and Supplementary Fig. 30). The scanning electron microscope (SEM) images demonstrated the comparable dense surfaces of both CoOxCly and CoOx films (Supplementary Figs. 31, 32). The cross-section of CoOxCly film characterized by scanning transmission electron microscopy (STEM) showed a film thickness of about 600 nm after 40 h of electrodeposition at 1.67 V (Fig. 2c and Supplementary Fig. 33). The EDX mapping of the cross-section further exhibited the relatively homogenous distribution of Co and Cl elements (Supplementary Fig. 34). In addition, we characterized the structure of the electrodeposited catalyst films. X-ray diffraction (XRD) demonstrated that CoOxCly film was absent of characteristic diffraction peaks (Supplementary Fig. 35). The crystal structure of both CoOx and CoOxCly catalysts was further imaged by the HRTEM. The lattice fringe together with the selected area electron diffraction (SAED) pattern showed that CoOx film was composed of CoOx nanocrystals (Supplementary Fig. 36). In contrast, CoOxCly film exhibited an amorphous structure without lattice fringe and SAED rings (Fig. 2d). Even after 500 h of operation, the CoOxCly film still remained the amorphous structure (Supplementary Fig. 37). The amorphous nature was probably ascribed to the Cl- co-deposited with Co2+ in the catalyst as it may be induced by reconstruction or distortion relative to CoOx40.

Fig. 2: The composition and structure of the CoOxCly catalyst.
figure 2

a The XRF Co Kα and Cl Kα mapping of the CoOxCly film electrodeposited in the electrolyte containing 0.1 M Co2+ and 0.5 M Cl- at 1.67 V for 10 h. The scale bar is 1 mm. b The EDX mapping of the CoOxCly film after 40 h electrodeposition at 1.67 V. c The cross-section TEM images of the CoOxCly film after 40 h electrodeposition at 1.67 V. d The HRTEM images and corresponding SAED patterns (inset) of the CoOxCly film after 40 h electrodeposition at 1.67 V. e The Cl 2p X-ray photoelectron spectroscopy (XPS) spectrum of the CoOxCly film. This peak could be well fitted by 197.6 eV for Cl 2p3/2 and 199.4 eV for Cl 2p1/2, respectively41,42.

Full size image

Suppression of Co oxidation state by Cl- in the CoOxCly catalyst

Further, compared with the CoOx film in the absence of Cl- (Supplementary Fig. 38), a peak at 197.6 eV for Cl 2p in CoOxCly film was observed, which was attributed to the Co-Cl bond (Fig. 2e)25,41,42. Analogous to Co (Supplementary Fig. 8 and 39a), the content of Cl increased with the electrodeposition time (Supplementary Fig. 39b and Supplementary Table 2). In addition, the chemical states of the as-prepared CoOxCly film were analyzed by XPS to study the role of Cl- inclusion. The Co 2p spectra can be well deconvoluted by the Co2+ and Co3+ components according to the reported binding energies43,44 (Fig. 3a). The Co3+/Co2+ ratio in the CoOx film is higher than that in the CoOxCly film (Fig. 3b). Moreover, the Raman spectrum of CoOxCly film with peaks at 180.7, 456.2, 502.4, 591.5, and 660.1 cm−1 belong to the 3 F2g, 1 Eg, and 1 A1g vibrational modes of cobalt oxides (Fig. 3c)45,46. The blueshift of the vibration bands relative to those of CoOx film indicated the variation of Co local coordination owing to Cl- introduction, consistent with its relatively lower Co oxidation state (Fig. 3a)40.

Fig. 3: The role of Cl- on the Co oxidation state for the CoOxCly catalyst.
figure 3

a The Co 2p XPS spectra of the CoOxCly and CoOx film after 2 h electrodeposition at 1.67 V, respectively. b The corresponding Co3+/Co2+ ratio of the CoOxCly and CoOx film. c The Raman spectra of the CoOxCly and CoOx films after 10 h electrodeposition at 1.67 V. d Co K-edge XANES at different applied potentials from the open-circuit condition to 2.02 V for CoOxCly film. e Co K-edge energy at different applied potentials for CoOx and CoOxCly film. The absorption energy (E0) is obtained from the first maximum in the first-order derivative as the electronic vacancy. The error bars represent the standard deviation for triplicate measurements.

Full size image

In situ X-ray absorption fine structure (XAFS) spectroscopy was further applied to elucidate the valence state of Co under CER conditions (Supplementary Fig. 40). The Co K-edge X-ray absorption near edge structure spectra (XANES) of CoOxCly and CoOx film were collected under anodic polarization in the 0.5 M NaCl and NaClO4 electrolytes, respectively47. As shown in Fig. 3d and Supplementary Fig. 41, the anodic polarization induced a shift of the Co K-edge to a higher energy region for both the CoOxCly and CoOx film suggesting an increase in the Co oxidation state44,48,49. However, the CoOxCly film underwent a larger shift (~0.83 eV) than that of the CoOx film (~0.41 eV) upon the potential increase from OCP to 2.02 V (Fig. 3e). Clearly, the CoOxCly film showed lower Co K-edge absorption energy relative to that of CoOx film at all the test potentials. This suggested a relatively lower Co oxidation state in CoOxCly film. Even at OCP, the Co valence state remains lower in the CoOxCly relative to CoOx, thus upon increasing the potential to 1.67 V, an obvious increase in the Co valence state allows CER to initiate40. Further, we analyzed the Fourier-transformed of k3-weighted Co K-edge extended X-ray absorption fine structure (EXAFS) spectra of both in CoOx and CoOxCly catalysts, which showed two major peaks associated with Co-O (~1.90 Å) and Co-Co (~2.85 Å) coordination (Supplementary Figs. 42, 43). An additional peak located around ~2.35 Å in the CoOxCly catalyst was observed, indicating the presence of Co-Cl coordination (Supplementary Fig. 43). Compared to the CoOx catalyst, the CoOxCly catalyst exhibited a lower fitted Co-O coordination number during the anodic polarization (Supplementary Tables 3, 4) and displayed a larger increase in Co-O coordination number relative to the CoOx catalyst, in coincidence with its significant degree of Co oxidation state variation50,51.

To get a deeper understanding of the CER process together with the effect of Cl- introduction, in situ spectroelectrochemical UV-Vis was applied to track the deposition processes of CoOxCly and CoOx catalysts (as illustrated in Fig. 4a and Supplementary Fig. 44). Figure 4b demonstrated the variation of spectral absorption of the CoOx film deposited in the absence of Cl- at 1.67 V. The absorption intensity increased rapidly within 30 min of deposition and several characteristic peaks appeared at A1-A4 corresponding to the charge transfer transitions52,53. The absorption centered at A2-A4 was assigned to the higher oxidation state of Co coordinated with highly oxidized oxygen species in the oxygen evolution process54,55,56. Instead, A2-A4 absorbance bands were absent in CoOxCly film, again suggesting a relatively lower oxidation state of Co than that in CoOx film (Fig. 4c). Moreover, the absorbance intensity of CoOxCly film is almost two times that of the CoOx film. Such a phenomenon could be ascribed to the Cl- accelerated electrodeposition and CER process. To clarify this phenomenon, we tracked the absorption of electrodeposited CoOxCly by stepwise increasing the Cl- concentration from 0 to 2.0 M (Supplementary Fig. 45). As expected, the A1 absorption intensity distinctly increased with the concentration of Cl- (Fig. 4d). When Cl- concentration was at 0.01 M, the A1 band intensity was relatively weaker while the A2-A4 appeared, indicating the existence of OER active phase analogous to that in CoOx. These results together with in situ XANES analyses illuminate that the Cl- introduction promotes the electrodeposition process and improved the Cl2 selectivity by suppressing the formation of highly oxidized Co sites required for OER.

Fig. 4: In situ UV-Vis spectra for tracking the electrodeposition process of the CoOxCly and CoOx film on FTO.
figure 4

a The schematics illustrate the in situ spectroelectrochemical UV-Vis tracking of the catalyst deposition process (left). The schematic diagram for CER on the CoOxCly electrode (right). b The UV-Vis absorption spectra of the deposited CoOx film in the electrolyte containing 0.1 M Co2+ and 0.5 M ClO4- at 1.67 V. c The UV-Vis absorption spectra of the deposited CoOxCly film in the electrolyte containing 0.1 M Co2+ and 0.5 M Cl- at 1.67 V. d The UV-Vis absorption spectra of the deposited CoOxCly film upon increasing the Cl- concentration from 0.01 to 2.0 M. Inset shows the absorption spectra around the A2-A4 range for the CoOxCly electrodeposited in 0.01 M Cl- electrolyte.

Full size image

The role of Cl־ on the CER process by DFT calculations

Subsequently, we applied the density functional theory (DFT) calculation to further clarify the effect of Cl- on Co valence states. Based on the above experiment results, we selected cobalt oxide with both octahedral and tetrahedral Co (Cooctahedral /Cotetrahedral = 3) (Supplementary Fig. 46 and Supplementary Table 5), and constructed six possible explicit solution models based surfaces without and with Cl- in the catalyst (Supplementary Fig. 47). The negative formation energies suggested that Cl- was spontaneously introduced into the catalyst (Supplementary Note 4, Supplementary Table 6, and Supplementary Fig. 48). We chose two models with lowest formation energy (Cl-doped-d and Cl-doped-e) to study the Co valence state in comparison with that of the Cl-free model (Supplementary Fig. 49). Because the surface atoms have lower coordination numbers than bulk atoms, surface Co sites are covered with water ligands. The density of electronic states (DOS) (Supplementary Fig. 50) shows that tetrahedra Co induces the defect states in the bandgap, attributed to the surface oxygens coordinative unsaturation and higher oxidation states of Co (Supplementary Fig. 51 and Supplementary Table 7). The defect states usually lead to significant optical absorption at the infrared and ultraviolet-visible regions57, which may correspond to the A2-A4 absorbance of CoOx film. When the surface oxygen atoms are substituted by Cl-, the defect states disappeared, in line with the absence of A2-A4 absorbance bands of CoOxCly film. Moreover, the hole states near the Fermi level appear to suggest Cl- in the catalyst enhanced electrical conductivity with reduced charge transfer resistance (Supplementary Fig. 13). As shown in Supplementary Figs. 52, 53 and Supplementary Tables 8, 9, the calculated Bader charge and bond length suggested that Cl- introduction could increase the Co electronic charge and decrease the valence state, consistent with in situ UV-Vis and XANES results.

To gain an in-depth understanding of the high Cl2 selectivity of CoOxCly, we further studied the influence of Cl- introduction on OER and CER. We first constructed the OER and CER reaction free-energy diagrams as a function of electrode potential (U) at pH 2 and 0.5 M Cl- based on the thermodynamically stable structures of various adsorbates10,58 (i.e. Cl*, ClO*, H2O*, HO*, O* and HOO*) on Co sites (Supplementary Note 5). Without Cl- in the catalyst, Cl* is favorable at U < 1.4 V, which transferred to Cl-O* at a higher potential (Fig. 5a). On the other hand, in the presence of Cl in the catalyst, OH* became mover favorable at U < 1.4 V (Fig. 5b and Supplementary Fig. 54). Hence, we evaluated the influence of pH and Cl- concentrations on the reaction pathways (Supplementary Fig. 55). The results indicated that even with very low Cl- concentrations, CER is predominant at pH 2 on both Cl-free and Cl-containing catalysts.

Fig. 5: DFT calculations to understand the role of Cl־ on the reaction mechanism.
figure 5

a The surface phase diagram of the Cl-free catalyst model. b The surface phase diagram of the Cl-containing catalyst model. c The reaction free energy diagram of OER on the Co sites of the Cl-free and Cl-containing catalysts, respectively. d The reaction free energy diagrams of CER on the Co sites of the Cl-free and Cl-containing catalysts, respectively. Insets: corresponding computational models for each CER step. e The schematic illustration of the role of Cl- introduction on the CER reaction active sites.

Full size image

To further understand the high selectivity of Cl2 to O2, we compared the reaction free-energy diagrams between OER and CER (Fig. 5c, d and Supplementary Note 6). For CER, we consider the mechanism that proceeded through the Volmer (O*/* + Cl- → ClO*/Cl* + e-) and the subsequent Heyrovsky step (ClO*/Cl* + Cl- → Cl2 + e-), where the Cl- directly adsorbed on the active site. On the Cl-free catalyst, the theoretical overpotentials for the OER and CER were 0.63 V and 0.39 V, respectively (Fig. 5c, d, Supplementary Figs. 56–59, and Supplementary Tables 10, 11). On the Cl-containing catalyst, the theoretical overpotential for both the OER and CER were reduced to 0.41 V and 0.2 V, respectively (Fig. 5c, d, Supplementary Figs. 60–63, and Supplementary Tables 12, 13). Even though the theoretical potential of 1.36 V for CER is 0.13 V higher than the theoretical potential of 1.23 V for OER, such results still suggest that CER is more favorable than OER on the catalysts with and without Cl. However, the presence of Cl- in the catalysts can accelerate the catalysis reaction. In addition, there is a transition of the active site from Co* to the Co-O* for CER with the Cl- introduction into the catalyst (Fig. 5e). The same trend was observed for other Cl-containing models (Supplementary Figs. 64–67, and Supplementary Tables 14, 15).

In this work, we present a highly selective and robust CER system based on an amorphous CoOxCly electrocatalyst electrodeposited in an acidic electrolyte containing both Co2+ and Cl- ions. Notably, this work presents the first case in that Cl- ions work as a promoter for catalyst deposition and CER. Experimental results demonstrate that ~100% CER selectivity has been acquired and in situ spectroscopic studies together with DFT theoretical calculations show that the introduction of Cl- into the catalyst suppresses the formation of highly oxidized Co sites required for water oxidation while allowing for the active site transition from Co* to Co-O* for CER with enhanced CER selectivity via the Volmer-Heyrovsky pathway. This work provides a new avenue to stabilize the non-precious metal oxides for anodically electrocatalytic reactions in strong acids and we would suggest in situ characterize the catalyst system with cautions before we categorize a soluble species or molecule as a homogeneous catalyst.

Methods

Chemicals

All chemicals used in this work were commercially available. Cobalt nitrate hexahydrate (Co(NO3)2·6H2O, 99.99%), cobalt chloride hexahydrate (CoCl2·6H2O, 99.99%), ferric nitrate nonahydrate (Fe(NO3)3·9H2O, 99.99%), manganese nitrate tetrahydrate (Mn(NO3)2·4H2O, 99.99%), potassium ferricyanide trihydrate (K3Fe(CN)6·3H2O, 99.99%), urea (CO(NH2)2, 99.999%), sodium chloride (NaCl, 99.99%), ammonium fluoride (NH4F, 99.99%), perchloric acid (HClO4, 70–72%), and anhydrous ethanol (C2H6O, 99.8%) were purchased from Sigma-Aldrich. Cobalt sulfate heptahydrate (CoSO4·7H2O, 99.99%), nickel nitrate hexahydrate (Ni(NO3)2·6H2O, 99.99%), copper nitrate trihydrate (Cu(NO3)2·3H2O, 99.99%), sodium perchlorate monohydrate (NaClO4 · H2O, 99.99%), and cobalt acetate tetrahydrate (Co(acetate)2·4H2O, 99.9%) were purchased from Aladdin. Hydrochloric acid (HCl, 31%), nitric acid (HNO3, 65–68%), sulfuric acid (H2SO4, 95–98%), and acetone (C3H6O, 99.9%) were purchased from KESHI. Commercial ruthenium oxide (RuO2, Ru > 75%) and iridium oxide (IrO2, Ir > 85%) were purchased from Adamas. In all experiments, ultrapure deionized water (Milli-Q, 18.2 MΩ cm-1) was used for the electrolyte preparation. The clean fluorine-doped tin oxide (FTO) with resistance <15 ohm/sq was used as a substrate for electrodeposition. All chemicals were used without further purification.

In situ electrodeposition of CoOxCly catalyst

The electrolyte for the working electrode side was prepared by mixing 0.1 M Co2+ salts and 0.5 M NaCl in 40 mL of deionized water. The pH of the electrolyte was adjusted to 2 using HClO4. Typically, the CoOxCly film was in situ electrodeposited on the FTO in the Co salt-NaCl electrolyte at 1.67 V at 25 °C. The electrochemical tests were conducted in a custom-designed PEEK cell. The Pt counter electrode was separated by an N117 Nafion membrane with a thickness of 183 um and size of 4 × 4 cm and was immersed in 0.5 M NaCl (pH 2) as well, to avoid the corrosion and the cathodic deposition of Co-based film. Before the experiment, the membrane was inserted in 5% hydrogen peroxide soaking for 60 min, deionized water for 30 min, 5% diluted sulfuric acid for 60 min, and deionized water for 30 min at around 80 °C.

Synthesis of Co3O4 film

Typically, the Co3O4 electrode was prepared by a simple hydrothermal method following a previous report39. Specifically, 0.233 g Co(NO3)2· 6H2O, 0.07 g of NH4F, and 0.240 g of CO(NH2)2 were dissolved in 100 mL deionized water under stirring for 30 min. Then, the obtained solution was transferred to a 50 mL Teflon-lined autoclave with an FTO inside as substrate. The reaction was conducted at 120 °C for 10 h and then cooled to room temperature. The obtained FTO substrates were washed with deionized water, dried at 60 °C overnight, and annealed in air at 350 °C for 2 h with a heating rate of 2 °C min-1.

Synthesis of RuO2 and DSA catalysts

To prepare the RuO2 film electrode, 5 mg commercial RuO2 was dissolved into 0.95 mL ethanol by adding 0.05 mL 5 wt% nafion117 solution under ultrasonic treatment for 60 min. 10 uL of this suspension was dropped onto the clean FTO (the efficient contact area is 0.283 cm2) for the formation of a catalyst film. The DSA catalyst film was prepared based on the commercial protocol on Ti (the ratio of Ir/Ru is 2).

Electrochemical characterizations

The CER tests were conducted based on a standard three-electrode with a computer-controlled electrochemical workstation (Bio-Logic SP−200). FTO (efficient contact area: 0.283 cm2), platinum mesh, and Hg/Hg2SO4 electrode were used as the working electrode, counter electrode, and reference electrode, respectively. A custom-designed PEEK cell with a quartz window was used as the electrochemical cell. All experiments were carried out in a temperature-controllable box, and the temperature was kept at 25 oC if not specially indicated. Before electrochemical experiments, the reference electrodes were calibrated by a standard reversible hydrogen (RHE) electrode (PHYCHEMI) in the corresponding electrolytes after 60 min Ar bubbling. For the working electrode, magnetic stirring at 1500 RPM was applied to remove chlorine bubbles during the electrochemical tests. All the measured potentials were calibrated with a reversible hydrogen electrode (RHE) as the reference. The pH values of the electrolyte were measured through an Ohaus Starter 2100 pH meter with temperature calibration. For the electrochemical tests at different NaCl concentrations, the NaClO4 · H2O was added for compensation to maintain a constant ionic strength8. The ohmic resistance (R) for iR corrections was determined by electrochemical impedance spectroscopy (EIS), and the recorded ohmic resistance was around 46.0 Ω, as shown in Supplementary Fig. 13. The overpotential in Fig. 1b and potential in Supplementary Fig. 28 were iR corrected with 90% ohmic resistance, while the other data were not iR-corrected. Electrochemical impedance spectroscopy (EIS) was measured at 1.67 V with a frequency scan range from 100 kHz to 10 mHz, and the amplitude of the sinusoidal wave was 10 mV.

In situ X-ray Absorption Fine Structure (XAFS) measurements

The XAFS measurements were recorded at the BL11B beamline of the Shanghai Synchrotron Radiation Facility (SSRF). The beam current of the storage ring was 220 mA in a top-up mode and the Si (111) double-crystal monochromator was applied to monochromatize the incident photons, with an energy resolution ΔE/E ~ 2×10-4. The collected spot size of the sample was ~200 μm × 250 μm (H × V). And the Co foil was used to calibrate the position of the absorption edge (E0). All XAFS spectra were collected in fluorescence mode. The E0 was obtained from the largest peak in the 1st derivative XANES according to the previous work47,59. The CoOxCly and CoOx films were polarized at different anodic potentials in a three-electrode system, respectively. After the corresponding current density reaches a steady state, the XAFS data was collected at the indicated potentials. All XAFS data were processed by the ATHENA and ARTEMIS modules implemented in the IFEFFIT software package.

In situ UV-Vis

The in situ UV-Vis spectroscopy was implemented on a QE Pro UV-Visible spectrometer (Ocean Optics) equipped with an HL-2000 light source and a DH-2000-BAL source. A fiber-optic cable (100 μm fiber core diameter) was used to direct the light from the light source through the sample to the detector. The UV-Vis spectra were recorded in the different electrolytes as a function of electrolysis times at 1.67 V. The FTO (area: 0.283 cm2), Pt electrode, and Hg/Hg2SO4 were used as the working electrode, the counter electrode, and the reference electrode, respectively. The spectrum of the unpolarized cleaned FTO immersed in the same electrolyte was recorded for the background subtraction of the UV-Vis spectra of the deposited CoOxCly and CoOx catalysts.

Online differential electrochemical mass spectrometry (DEMS)

The online DEMS experiments were implemented with the chamber containing the quadrupole mass spectrometer (MS) with a Faraday-SEM detector array (Hiden Analytical, HPR-40), which allows the isolation of the ion source from the electrochemical cell by forming a small pre-chamber37,38. For the DEMS measurements, a custom gas-tight electrochemical H-cell with a conventional three-electrode arrangement was employed for the electrochemical experiments. In this experiment, a Nafion membrane was used to separate the cathode and anode. The acquisition probe is vertically attached to the working electrode where the volatile species pass through a porous Teflon membrane into the vacuum inlet of the mass spectrometer. For the DEMS measurements, the multiplier mode was adopted. The LSV and mass spectrometric cyclic voltammograms (MSCVs) were acquired in electrolytes containing 0.1 M Co2+ with 0.5 M NaClO4 or 0.5 M NaCl, respectively. The MSCVs for the concerned species (70Cl2: m/z = 70, 1H16O35Cl: m/z = 52, 35Cl32O2: m/z = 67, 35Cl16O: m/z = 51 and 32O2: m/z = 32) were recorded. Before the tests, the electrolyte was purged with Ar (99.999%) over 1 h to remove the dissolved oxygen.

Materials characterizations

The X-ray fluorescence method based on Co and Cl content analyses was conducted on a Bruker M4 Tornado instrument with Rh target for X-ray generation. The catalyst morphology of the samples was characterized by a SU8010 field emission scanning electron microscope (SEM). The cross-section of the catalyst was characterized by a transmission electron microscope (JEOL JEM2100F) operating at 200 keV in both TEM and STEM modes. The electron beam spot size is 1.0 nm. An EDX detector (Oxford Instruments) was equipped for the element analysis. The TEM sample for cross-section observation was prepared by cutting the electrodeposited FTO electrode into two pieces, of which the cross-section was firstly glued by the epoxy resin and then processed through Ar+ ion-milling and polishing (PIPS II, GATAN) at grazing incidence (<5°). The X-Ray Diffraction (XRD) measurement of CoOxCly catalyst was conducted on Bruker D8 ADVANCE A25X. The high-resolution TEM (HRTEM) images of CoOxCly and CoOx catalysts were taken on an FEI-Tecnai G2 microscope. Moreover, the element mapping and selected area electron diffraction (SAED) were performed on an FEI TECNAI G2 F30 field emission transmission electron microscope. The X-ray photoelectron spectroscopy analyses were carried out using a Thermo Scientific K-Alpha system to evaluate the composition and chemical states of CoOxCly and CoOx catalysts. Al Kα X-ray source with a power of 250 W was used as the X-ray source and the binding energies were calibrated by referring to the C 1 s peak (284.8 eV) of the adventitious carbon. The Raman was carried out with an XploRA PLUS Raman spectrometer (Horiba) equipped with a 532 nm He-Ne laser as the excitation source.

Rotating ring-disk electrode detection of selectivity Cl2 evolution

The Cl2 selectivity for CER was analyzed based on the RRDE system in which the disk electrode generates Cl2 while the Pt ring electrode monitors Cl2 concentration through its electrochemical reduction current. The potential of the Pt ring was fixed at 0.95 V to avoid the current response from O2 reduction. Before each measurement for Cl2 quantification, the Pt ring-disk electrode was successively polished with aqueous suspensions of 1.0 μm, 0.3 μm, and 0.05 μm alumina. It was then rinsed with deionized water, ultrasonically cleaned in ethanol for 15 s, and dried. Later on, the Pt ring was further electropolished in 0.5 M H2SO4 by 30 CV scans between -0.1 and 1.7 V at a rate of 500 mV s−1. The chronoamperometry method was applied for the Cl2 selectivity quantification. During the Cl2 selectivity measurements, the potentiostatic electrolysis was carried out for 5 min at 1600 RPM at different applied disk potentials35. Each experiment was repeated three times with an interval of 5 min in an Ar-saturated electrolyte at pH 2 containing 0.1 M Co(NO3)2 and NaCl with different concentrations.

The CER current density (jCER) on the disk electrode can be calculated by the following equation:

$${j}_{{{{{{\rm{CER}}}}}}}={{{{{\rm{|}}}}}}{j}_{{{{{{\rm{R}}}}}}}\,/\,{{{{{{\rm{N}}}}}}}_{{{{{{\rm{l}}}}}}}{{{{{\rm{|}}}}}}={{{{{\rm{|}}}}}}{{{{{{\rm{i}}}}}}}_{{{{{{\rm{R}}}}}}}/{{{{{\rm{A}}}}}}{{\times }}{{{{{{\rm{N}}}}}}}_{{{{{{\rm{l}}}}}}}{{{{{\rm{|}}}}}}$$
(1)

Where jR is the current density measured on the ring, Nl is the collection efficiency (0.37), and A is the disk electrode area (0.196 cm2).

The OER current density (jOER) can be calculated by the following equation:

$${j}_{{{{{{\rm{OER}}}}}}}={j}_{{{{{{\rm{D}}}}}}}-{j}_{{{{{{\rm{CER}}}}}}}={j}_{{{{{{\rm{D}}}}}}}-{{{{{\rm{|}}}}}}{j}_{{{{{{\rm{R}}}}}}}\,/{{{{{{\rm{N}}}}}}}_{{{{{{\rm{l}}}}}}}{{{{{\rm{|}}}}}}={j}_{{{{{{\rm{D}}}}}}}-{{{{{\rm{|}}}}}}{{{{{{\rm{i}}}}}}}_{{{{{{\rm{R}}}}}}}\,/{{{{{\rm{A}}}}}}{{{{{\rm{\times }}}}}} \times {{{{{{\rm{N}}}}}}}_{{{{{{\rm{l}}}}}}}|$$
(2)

Where jD is the current density measured on the disk.

The Cl2 selectivity was calculated by the following equation:

$${{{{{{\rm{Cl}}}}}}}_{2} \, {{{{{\rm{selectivity}}}}}}(\%) \,=\, 100\times 2\times {j}_{{{{{{\rm{CER}}}}}}}/({j}_{{{{{{\rm{D}}}}}}}+{j}_{{{{{{\rm{CER}}}}}}})\\ \,=\, 100\times 2\times {{{{{\rm{|}}}}}}{j}_{{{{{{\rm{R}}}}}}}/{{{{{{\rm{N}}}}}}}_{{{{{{\rm{l}}}}}}}{{{{{\rm{|}}}}}}/({j}_{{{{{{\rm{D}}}}}}}+{{{{{\rm{|}}}}}}{j}_{{{{{{\rm{R}}}}}}}/{{{{{{\rm{N}}}}}}}_{{{{{{\rm{I}}}}}}}{{{{{\rm{|}}}}}})$$
(3)

Computation details

The first-principles calculations based on density functional theory (DFT) were carried out within the projected augmented wave method, as implemented in the Vienna Ab-initio Simulation Package (VASP). The CoOxCly film catalyst models were made by cleaving along the (100) facet of Co3O4 which contained two tetrahedrons and six octahedrons and were solved by explicit solution models. To generate highly accurate electrochemical stability diagrams, we employed a recently developed approach, which included the use of a Hubbard U term, a van der Waals functional (optPBE), and the use of spin polarization for the calculations. The U-value applied for d-orbitals of Co is taken as 3.50 eV. A cutoff energy of 500 eV was used for the plane wave expansion at all calculations of structure relaxations, vibrational frequencies, and single point energy. Monkhorst-Pack (4 × 4 × 1) k-point grids were used for Brillouin zone integration. The equilibrium structures were obtained when the maximum atomic forces were smaller than 0.01 eV/Å and when the total energy convergence of 10−6 eV was achieved for the electronic self-consistent field loop. The thermodynamic energy correction used in the free energy calculations was listed in Supplementary Table 6. The detailed descriptions of formation energy, OER, and CER reaction free energies calculations have been discussed in Supplementary Notes 4–6.