CoO_x clusters-decorated IrO₂ electrocatalyst activates NO₃^- mediator for benzylic C-H activation

Abstract

Electrochemical conversion of petrochemical-derived hydrocarbons to high-value oxygenates can utilize renewable energy and reduce carbon emissions. However, this involves the challenging activation of inert C(sp³)-H bonds at room temperature. Here, we introduce an electrocatalyst:mediator assembly in which CoO_x clusters-decorated IrO₂ electrocatalyst activates NO₃^- mediator to a highly reactive radical capable of abstracting a hydrogen atom from benzylic C-H. The interface between CoO_x and IrO₂ promotes NO₃^- activation by facilitating the desorption of NO₃^● radical for subsequent reaction. Our strategy is demonstrated through the selective oxidation of toluene to benzaldehyde with high Faradaic efficiency of 86( ±1)% at 25 mA/cm², a factor of >3 times higher than the bare electrocatalyst. The electrocatalyst:mediator assembly is operated stably for 100 h, with minimal decline in performance. When translated into a flow system, a Faradaic efficiency of 60( ±4)% at 200 mA/cm² was achieved.

Introduction

The partial oxidation of petrochemical-derived hydrocarbons to oxygenates is important for the plastics and textiles industries. This involves the activation of inert C(sp³)-H bonds in the hydrocarbons, which are characterized by high bond dissociation energies of 70–130 kcal/mol¹. Thus, the partial oxidations of hydrocarbons typically occur with high temperatures, pressures and harsh oxidants, such as chromium or selenium-based reagents, thereby leading to significant carbon emissions and environmental consequences². For example, the partial oxidation of toluene is important to the production of benzaldehyde, a pharmaceuticals and plastics precursor with an annual demand of 170 million kg^3,4. The reaction is conducted at high temperatures (i.e., 250–650 ^oC) and pressures due to the relative stability of C(sp³)-H bond adjacent to aromatic rings, and requires extensive cooling due to its highly exothermic nature, thus resulting in high energy requirements and CO₂ emissions⁵.

The electrification of hydrocarbon partial oxidations offers the potential for benign conditions and utilization of renewable energy, which can significantly reduce the carbon footprint⁶. These reactions can be coupled to a cathodic hydrogen evolution reaction (HER) from water, thereby increasing the chemical value that can be derived from a given amount of electricity input. An example is the electrochemical oxidation of ethylene and propylene to the respective epoxides/glycols by the direct oxidation or chlorohydrin routes^7,8,9,10. Nevertheless, the scope of electrochemical partial oxidations remains limited, especially for reactions that involve C(sp³)-H activation¹¹. While electrochemical C-H functionalization, such as amination¹² and arylation¹³ have been developed, these are mainly for the synthesis of specialty chemicals in a fully organic environment and tend to operate in a restricted current density window (i.e., 0.5–5 mA/cm²)¹⁴. These reactions rely on homogeneous catalysts such as organometallics^12,13,15 and organic mediators⁵, while non-targeted commercial electrodes, such as platinum and reticulated vitreous carbon, serve as current collectors (Fig. 1a). In the context of electro-refining petrochemicals-derived hydrocarbons to commodity chemicals, however, parameters such as Faradaic efficiency, current density and the ability to interface with cathodic HER become critical to enable technical scale application and maximize the chemical output of renewable electricity. In this case, materials-driven electrocatalysts are important to reduce reaction overpotential and provide adequate active sites to funnel electron selectivity towards the desired redox reaction^2,16, but lack the ability to selectively target inert C(sp³)-H bonds (Fig. 1b). We believe that combining the insights from organic electrosynthesis and materials-driven electrocatalysis can help to overcome the low reactivity of C(sp³)-H bonds while enabling the partial oxidation of hydrocarbons at high Faradaic efficiencies and current densities.

Fig. 1: Combining organic electrosynthesis and materials-driven electrocatalysis in an electrocatalyst:mediator assembly.

Illustration of anodic toluene oxidation by a, homogeneous catalysts with non-targeted commercial electrodes as current collectors, b, materials-driven electrocatalysis and c, the electrocatalyst:mediator assembly. d Schematic of anodic toluene oxidation coupled to hydrogen evolution from water at cathode. e Faradaic efficiencies of anodic toluene oxidation with different mediators. f LSV of the reaction system with and without NO₃^-. All error bars represent standard deviation based on three independent samples. No iR correction was conducted. Source data for Fig. 1e, f are provided as a Source Data file.

In this work, we present the concept of an electrocatalyst:mediator assembly, in which materials-driven electrocatalysis funnels electron selectivity towards the desired redox reaction, while the mediator selectively targets inert C(sp³)-H bonds (Fig. 1c). We demonstrate this through the selective oxidation of toluene to benzaldehyde; CoO_x clusters-decorated IrO₂ (CoO_x/IrO₂) electrocatalyst electrochemically oxidizes NO₃^- mediator to the NO₃^● radical, which can abstract a hydrogen atom from benzylic C-H in toluene. The electrocatalyst:mediator assembly yielded a high Faradaic efficiency of 86( ±1)% towards benzaldehyde at 25 mA/cm², which is more than 3 times higher than direct oxidation on the bare electrocatalyst. The electrocatalyst:mediator assembly is operated stably for 100 h with minimal decline in performance throughout, and can also be broadly applied to the electrochemical oxidation of other toluene derivatives. The interface between CoO_x and IrO₂ promotes NO₃^- activation by facilitating the desorption of NO₃^● radical for subsequent reaction, yielding 2x higher Faradaic efficiency compared to commercial electrodes such as glassy carbon, boron doped diamond and Pt, and 20% higher compared to IrO₂. When translated into a flow system, a Faradaic efficiency of 60( ±4)% at 200 mA/cm² was achieved.

Results

We demonstrated the concept of our electrocatalyst:mediator assembly through the partial oxidation of toluene to benzaldehyde (Fig. 1d, configuration of the H-type cell in Supplementary Fig. 1). This is conducted on the CoO_x/IrO₂ electrocatalyst on Ti support with 0.24 M toluene in an anolyte comprising 0.1 M LiBF₄ in acetonitrile (Supplementary Fig. 2a). At the cathode, hydrogen evolution from water is conducted on a Pt plate in aqueous LiBF₄ solution (Supplementary Fig. 2b). Both reactions are separated by a proton exchange membrane. The direct oxidation of toluene on the bare CoO_x/IrO₂ electrocatalyst yielded a low Faradaic efficiency of 26( ±4)% at 25 mA/cm² (Fig. 1b, potential versus reaction progress and linear sweep voltammogram can be found in Supplementary Fig. 3). With the introduction of the NO₃^- mediator, the Faradaic efficiency increased by a factor of >3x to 86%. We compared the electrochemical performance with that of other mediator candidates with known oxidative properties, such as NHPI, ClO₄^-, Br^- and Cl^- (Fig. 1e, the corresponding amount of produced benzaldehyde and LSVs can be found in Supplementary Fig. 4 and Fig. 1f, respectively)^17,18,19,20. The Faradaic efficiencies with N-Hydroxyphthalimide (NHPI) and ClO₄^- are higher than that of the bare electrocatalyst at 46( ±13) and 30( ±2)%, but far lower than that obtained with NO₃^-. No difference in Faradaic efficiency was observed whether NH₄NO₃, LiNO₃ or NaNO₃ was used, which indicates that the cation does not play a role in this reaction (Supplementary Fig. 5a). The optimal NO₃^- concentration was 0.06 M; further increments in NO₃^- concentration led to slightly lower Faradaic efficiencies (Supplementary Fig. 5b). The volume ratio of toluene to acetonitrile was also optimized (Supplementary Fig. 5c).

To understand the role of NO₃^-, we introduced 2,2,6,6-Tetramethylpiperidine 1-oxyl (TEMPO) radical scavenger into the reaction mix (Fig. 2a). This drastically reduced the Faradaic efficiency to 4.9( ±2.6)%, which suggests that the reaction involves the formation of radicals, which are quenched through the addition of TEMPO. We removed aliquots of anolyte over the course of the reaction for electron paramagnetic resonance (EPR) studies, which showed that the TEMPO signal decreased in intensity and was completely extinct at 20 min (Fig. 2b). This confirmed the formation of radicals during reaction, suggesting that NO₃^- was promoted to the highly reactive NO₃^● radical on the CoO_x/IrO₂ electrocatalyst. Cyclic voltammetry (CV) revealed an oxidation peak at about 1.7 V versus Ag/Ag⁺, which can be attributed to the conversion of NO₃^- to NO₃^● (Supplementary Fig. 6). This conversion is reversible, as evidenced by the reduction peak in reverse scan at about 1.8 V versus Ag/Ag⁺. The UV-vis spectra of the anolyte before the reaction showed a peak at 300 nm, which corresponds to the n→π* transition in acetonitrile²¹. This peak disappeared after the reaction, and a new peak was formed at 268 nm, which can be attributed to HNO₃ (Fig. 2c). This suggests that the NO₃^● radical abstracted a hydrogen atom from toluene to form a benzyl radical, which reacted with O₂ from air to form benzaldehyde. This is confirmed by the drop in Faradaic efficiency to 11( ±2)% when Ar was sparged into the anolyte (Supplementary Fig. 7). It also means that the Faradaic efficiency can be improved by increasing the presence of O₂ in the anolyte, i.e., 86( ±1)% with the sparging of air versus 65( ±10)% at ambient conditions.

Fig. 2: CoO_x/IrO₂:NO₃^- electrocatalyst:mediator assembly for the partial oxidation of toluene to benzaldehyde.

a Quenching of reaction with the addition of TEMPO radical scavenger. All error bars represent standard deviation based on three independent samples. b EPR spectra of the anolyte with and without NO₃^-, using TEMPO to track the formation of radical intermediates. c UV-vis spectra of the anolyte before and after reaction. d Half-cell potential and Faradaic efficiency of benzaldehyde over 100 h. e Composition of the CoO_x/IrO₂ anode before and after 100 h of reaction, based on EDX analysis. Source data for Fig. 2a-e are provided as a Source Data file.

Meanwhile, the Faradaic efficiency for bare CoO_x/IrO₂ electrocatalyst decreased to 3.0% with the addition of TEMPO, which suggests that the reaction pathway for direct oxidation also involves the formation of a benzyl radical (Fig. 2a). The time taken for the TEMPO signals to disappear, as observed by EPR spectroscopy, was longer at 40 min, which indicates that this reaction pathway was not as robust as with a NO₃^- mediator (Fig. 2b). This explains why the Faradaic efficiency achieved with the electrocatalyst:mediator assembly is much higher than direct oxidation on the bare electrocatalyst.

The CoO_x/IrO₂:NO₃^- electrocatalyst:mediator assembly was highly stable under the electrochemical settings required for the oxidation of toluene to benzaldehyde. We demonstrated this through 100 h of continuous operation at a current density of 25 mA/cm², during which portions of the electrolyte were periodically removed for analysis and replaced with fresh electrolyte. The system maintained a stable applied potential of 2.17( ±0.06) V versus Ag/Ag⁺ and Faradaic efficiency averaging 81( ±7)% (Fig. 2d). Post-reaction analysis of the anode through energy-dispersive X-ray spectroscopy (EDX) revealed no obvious changes in composition of the CoO_x/IrO₂ electrocatalyst (Fig. 2e, Supplementary Fig. 8), which suggests that electrocatalyst:mediator assembly is suitable for long-term operation.

The CoO_x/IrO₂:NO₃^- electrocatalyst:mediator assembly can also be broadly applied to the electrochemical oxidation of other toluene derivatives (Supplementary Fig. 9). The Faradaic efficiency is higher when the electron-donating methyl group is present in the benzene ring (i.e., 97( ±7)% for p-xylene), which stabilizes the benzyl radical intermediate. The converse is true for the electron-withdrawing chloro and carboxylic groups (i.e., 64( ±7)% for p-chlorotoluene and 52( ±6)% for p-toluic acid). We observed lower Faradaic efficiencies when bulky benzyl or ethyl groups are adjacent to the benzylic C-H (i.e., 44( ±4)% for diphenylmethane and 34( ±7)% for propylbenzene). This indicates that the oxidation reaction occurs at the electrocatalyst surface where the bulky groups present a large steric hindrance, which suggests that CoO_x/IrO₂ plays a significant role in the electrocatalyst:mediator assembly.

We thus proceeded to characterize the CoO_x/IrO₂ electrocatalyst. CoO_x/IrO₂ was deposited on a Ti mesh through dip coating in electrocatalyst ink comprising Co and Ir precursors, followed by thermal decomposition²². The CoO_x loading amount was optimized by varying the ratio of Co to Ir precursors in the electrocatalyst ink (see Supplementary Figs. 10-16 for characterization of the CoO_x/IrO₂ samples with different CoO_x loadings). The CoO_x/IrO₂ anode synthesized with Co/Ir ratio of 20 wt% showed the highest double layer capacitance (C_dl) in the reaction medium, which indicates highest electrochemical active surface area (ECSA, Supplementary Fig. 10h, CV curves of anodes synthesized from inks of Co/Ir ratio in Supplementary Fig. 10b-g) and therefore highest catalytic activity (Supplementary Fig. 10a). LSV scans performed in reaction medium (Supplementary Fig. 10i) revealed an obvious oxidation peak at around 1.6-1.7 V versus Ag/Ag⁺ for the anode synthesized from ink of Co/Ir ratio 20 wt%. This peak corresponds to the oxidation peak of NO₃^- (see Supplementary Fig. 6), which indicates greater affinity towards NO₃^-. This means that NO₃^- can be oxidized more easily to the NO₃^● radical to facilitate the electrochemical toluene oxidation reaction, as evidenced by the higher corresponding current density. Scanning electron microscopy (SEM) showed that CoO_x/IrO₂ particles were well-coated on the Ti mesh (Supplementary Fig. 13a-b). X-ray photoelectron spectroscopy (XPS) confirmed the presence of Co in +2 and +3 oxidation states, as well as Ir in +4 oxidation state (Fig. 3a-c). X-ray diffraction (XRD) showed that the CoO_x/IrO₂ electrocatalyst consisted of (110), (101), (200), (210), (220) and (301) planes of rutile IrO₂ at 28.1^o, 34.7^o, 40.1^o, 45.0^o, 58.5^o and 69.3^o respectively, as well as the (400) plane of Co₃O₄ at 44.8^o (Fig. 3d)^23,24. EDX showed that Co, Ir and O were well-distributed in the CoO_x/IrO₂ electrocatalyst (Fig. 3e). High-resolution transmission electron microscopy (HRTEM) revealed the CoO_x cluster-on-IrO₂ structure of the CoO_x/IrO₂ electrocatalyst (Fig. 3f). The small and dark particles, with an average size approximating 2 nm, are CoO_x clusters distributed over the IrO₂ substrate. The corresponding FFT pattern showed that the clusters comprised face-centered cubic lattice with spinel structure (inset, Fig. 3f). IrO₂ existed in continuous phase across a relatively large area, which was crystalline and comprised tetragonal lattice with rutile structure, which is clearly distinguished from that of the CoO_x phase. (Supplementary Fig. 13c). As no distortion was observed in IrO₂ lattice, it can be concluded that Co was neither doped into IrO₂ lattice at interstitial sites nor by replacing Ir atoms.

Fig. 3: Characterization of CoO_x/IrO₂ electrocatalyst.

XPS spectra of a, Co b, Ir and c, O in CoO_x/IrO₂. d XRD pattern of CoO_x/IrO₂. e EDX images showing the distribution of Co, Ir, and O on CoO_x/IrO₂. f HRTEM image of the CoO_x cluster of the CoO_x/IrO₂ electrocatalyst. Inset: The corresponding FFT image of the CoO_x cluster. g Co K-edge XANES spectrum and h, Fourier transformation of the k²-weighted EXAFS of CoO_x/IrO₂. i Ir L₃ edge XANES spectrum and j, Fourier transformation of k²-weighted EXAFS of IrO₂ and CoO_x/IrO₂ catalyst. Source data for Fig. 2a–d and g–j are provided as a Source Data file.

To understand the influence of CoO_x clusters on the electronic structure and local atomic geometric environment of IrO₂, we performed X-ray absorption fine structure (XAFS) experiment on the CoO_x/IrO₂ electrocatalyst and on IrO₂ synthesized through the same method. The Co K edge X-ray absorption near edge structure (XANES) spectrum of CoO_x/IrO₂ showed an identical edge energy compared to that of a commercial Co₃O₄ standard, which implies that Co possess similar oxidation states as Co₃O₄ (Fig. 3g). The Fourier transformation of extended X-ray absorption fine structure (EXAFS) revealed three peaks with positions similar to those of Co₃O₄ but with smaller amplitudes. These peaks correspond to the Co-O path, the Co-Co path between octahedral centers, and the Co-Co path between octahedral and tetrahedral centers, indicating the presence of small Co₃O₄ clusters highly distributed across the IrO₂ surface (Fig. 3h). This observation is consistent with the findings from the HRTEM analysis_. The low intensity of Co-Co scattering paths compared with those of the Co₃O₄ standard suggests higher atomic disorder due to the Co₃O₄ cluster configuration. Meanwhile, Ir L₃ edge XANES spectrum of CoO_x/IrO₂ revealed a higher white line than those of both standard rutile IrO₂ and our synthesized IrO_2, implying that the Ir in CoO_x/IrO₂ exists in >4+ oxidation state and is therefore capable of enhanced electro-oxidation performance (Fig. 3i-j)^25,26.

The importance of CoO_x/IrO₂ in the electrocatalyst:mediator assembly is demonstrated by comparing its electrochemical performance with electrocatalyst:mediator assemblies comprising other electrocatalysts (Fig. 4a, LSVs in Supplementary Fig. 17a). No reaction occurred on the Ti mesh, which indicates that it functions as a support and does not contribute to the reaction. Poor Faradaic efficiencies (5.4–34%) were obtained with commercial electrocatalysts such as glassy carbon, boron-doped diamond (BDD) and Pt. The highest Faradaic efficiency was obtained with IrO₂ at 66( ±4)%. The modification of IrO₂ with CoO_x clusters led to an increase in Faradaic efficiencies even under different current densities (10–40 mA/cm²). As little reaction was observed with Co₃O₄, this suggests that the CoO_x clusters may not be the reaction centers of the CoO_x/IrO₂ electrocatalyst. Meanwhile, modification with Au (Au/IrO₂) yielded lower Faradaic efficiencies (Fig. 4b, see Supplementary Figs. 16 and 18 for characterization of IrO₂ and Au/IrO₂ and LSV in Supplementary Fig 17b). EPR spectroscopy performed on a reaction with TEMPO scavenger showed that, when IrO₂ was used as the electrocatalyst, the TEMPO signal took a longer time to become extinct at 30 min, compared to 20 min with the CoO_x/IrO₂ electrocatalyst (Fig. 4c). When Au/IrO₂ was used as the electrocatalyst, the TEMPO signal became extinct at 35 min. This indicates that the rate of radical formation was highest with CoO_x/IrO₂, followed by IrO₂ and Au/IrO₂, which explains why Faradaic efficiency is the highest with CoO_x/IrO₂ electrocatalyst.

Fig. 4: The importance of CoO_x/IrO₂ in the electrocatalyst:mediator assembly.

Faradaic efficiencies of anodic toluene oxidation with electrocatalyst:mediator assemblies comprising a commercial electrocatalysts and b IrO₂-based electrocatalysts. All error bars represent standard deviation based on three independent samples. The reaction was run in a H-type cell, in which 0.5 mL toluene was dissolved in 19.5 mL of anolyte comprising 0.1 M LiBF₄ and 0.06 M LiNO₃ in acetonitrile solution. The catholyte is 20 mL of 0.1 M LiBF₄ MilliQ water (18.2 Ω) solution. The anolyte and catholyte were separated with a Nafion® 117 membrane. c EPR spectra of the anolyte with electrocatalyst:mediator assemblies comprising IrO₂-based electrocatalysts. d Adsorption energies of the NO₃^● radical on the IrO₂(100)-O_v, CoO_x/IrO₂(100)-O_v and Co₃O₄(100) slabs. Side views of the NO₃^● radical adsorbed on the e, Co and f, Ir atom of CoO_x/IrO₂(100)-O_v. Source data for Fig. 4a-c are provided as a Source Data file.

We further investigated the CoO_x cluster-promoted NO₃^- activation through density function theory (DFT) analysis, by modelling the adsorption of NO₃^● radical on the IrO₂ versus CoO_x/IrO₂ electrocatalysts. The IrO₂ electrocatalyst was modelled through an IrO₂(100) slab containing an oxygen vacancy, termed here as IrO₂(100)-O_v, and the NO₃^● radical adsorption energy is -2.15 eV (Fig. 4d, Supplementary Fig. 19). This suggests that it is difficult for NO₃^● radicals formed on the IrO₂ surface to desorb and abstract a hydrogen atom from toluene. A Co₃O₄(100) slab was built to simulate NO₃^● radical adsorption energy on the surface (Supplementary Fig. 20), and the NO₃^● radical adsorption energy is -1.02 eV. To simulate the interface between the CoO_x cluster and IrO₂ surface, we built a cluster-on-surface model reported in previous studies^22,27, referred to as CoO_x/IrO₂(100)-O_v (Fig. 4e-f, Supplementary Fig. 21). The synergistic effect between the CoO_x cluster and IrO₂ surface is such that the NO₃^● radical adsorption energy is -2.18 eV on the Ir_interfacial site and -0.81 eV on the Co_interfacial site. This suggests that it is easier for the NO₃^● radicals formed on the interfacial Co site to desorb for subsequent reaction, thereby regenerating the active site for another NO₃^- activation and furbishing the electrochemical system with higher concentration of NO₃^● radicals.

We also employed DFT to investigate the role of the NO₃^● radical in enabling C–H bond cleavage (the cartesian coordinates of intermediates and transition states are in Supplementary data 1). We simulated initial states where toluene binds, in different configurations, to an IrO₂(100)-O_v slab (Supplementary Fig. 22). In the A0 initial state where the methyl group in toluene binds with an Ir atom, the methyl group points toward the substrate and allows the hydrogen atom to adsorb onto the surface to stabilize itself upon C-H bond cleavage. In this configuration, the C-H bond is elongated to 1.21 Å compared to the typical 1.10 Å in the methyl group of toluene, thereby weakening the bond and facilitating its breakage (Fig. 5a, Supplementary Fig. 23). The released hydrogen atom is transferred to an oxygen atom on the IrO₂(100)-O_v surface via the transition state A1-TS, resulting in the formation of an adsorbed benzylic radical and OH group in A2. The overall reaction step from A0 to A2 is characterized by an energy barrier of 0.74 eV and releases -1.07 eV of energy (Fig. 5c).

Fig. 5: DFT simulation of the C-H bond cleavage step on IrO₂ electrocatalyst.

Proposed reaction pathways of C-H bond cleavage a directly on the IrO₂(100)-O_v surface and b mediated by a NO₃^● radical. Calculated energy profiles of the proposed reaction pathways c, without and d, with the NO₃^● radical. e Side view and f top view of the B0 state.

On the other hand, the NO₃^● radical facilitates the cleavage of the C–H bond by abstracting a hydrogen atom from toluene via the transition state B1-TS, thereby generating an adsorbed benzylic radical in B2 (Fig. 5b, e-f and Supplementary Fig. 24). The overall reaction step from B0 to B2 has an energy barrier of 0.26 eV and releases -1.31 eV of energy (Fig. 5d), making it more favorable than the first pathway. This means that the pathway with the NO₃^● radical (B0-B2) has an energy barrier that is 0.48 eV lower and releases -0.24 eV more reaction energy, compared to the pathway without NO₃^● radical (A0-A2). Thus C-H bond cleavage is kinetically and thermodynamically more favorable in the presence of the NO₃^● radical.

To compare with the electrochemical toluene oxidations reported in literature^15,20,28,29, we conducted a batch reaction on 0.12 M toluene in an anolyte comprising 0.1 M LiBF₄ in acetonitrile, under a current density of 25 mA/cm² over a reaction time of 800 mins. This is because Faradaic efficiency can decrease towards the end of the batch reaction as the toluene reactant gets depleted; our average Faradaic efficiency over the course of the reaction is about 56.5%. We obtained a yield of 75% and overall conversion rate of 0.15 mmol per cm² catalyst per hour of reaction, which corresponds to an overall benzaldehyde generation rate of 0.13 mmol per cm² catalyst per hour of reaction (Fig. 6a). The average Faradaic efficiency of the partial ethylbenzene-to-acetophenone conversion is about 67.2%, and we obtained a yield of 89% and overall conversion rate of 0.17 mmol per cm² catalyst per hour of reaction, which corresponds to an overall acetophenone generation rate of 0.16 mmol per cm² catalyst per hour of reaction (Fig. 6b)^15,20,28,29.

Fig. 6: Translating reaction into flow system towards high current densities.

Comparison of the partial current density, Faradaic efficiency, yield, selectivity, overall conversion rate and product generation rate against state-of-art a electrochemical toluene-to-benzaldehyde conversions^20,28, and b electrochemical ethylbenzene-to-acetophenone conversions^15,29. c Schematic of flow cell setup that conducts anodic toluene oxidation and cathodic HER. d Faradaic efficiencies of partial toluene-to-benzaldehyde conversion at different current densities with and without the addition of HFIP. The reaction in flow was conducted in a 3-compartment flow cell system. 340 µL toluene (or ethylbenzene or 4-chlorotoluene) was dissolved in 25 mL of anolyte comprising 0.1 M LiBF₄ and 0.06 M LiNO₃ in pure acetonitrile solvent or 4:1 acetonitrile:HFIP mixture. The catholyte was 0.1 M LiBF₄ aqueous solution. The flow rate was 60 mL/s. All error bars represent standard deviation based on three independent samples. e Breakdown of costs at current densities of 20 and 200 mA/cm², as calculated by TEA. Source data for Fig. 2a-b, d and e are provided as a Source Data file.

Finally, we translated our reaction into a flow cell system based on the obtained insights on the reaction mechanism (Fig. 6c, cell configuration in Supplementary Fig. 25). We postulated that an additive such as hexafluoro-2-propanol (HFIP) can help to stabilize the electron-deficient radicals and funnel selectivity towards the benzaldehyde product¹⁴. Fig. 6d shows that without HFIP, the Faradaic efficiency decreases dramatically as the current density increases. The addition of HFIP enabled high Faradaic efficiencies of 60–82% for the electrochemical oxidation of toluene derivatives even at higher current densities of up to 200 mA/cm². This is significant as prior electrochemical oxidations of benzylic C-H tended to be reported at low current densities <20 mA/cm² (Supplementary Table 1)^20,28,30. Technoeconomic analysis shows that higher current densities (i.e., 200 mA/cm²) lead to profitable plant-gate levelized costs due to reduced capital costs, as the same production rates can be enabled by smaller electrolyzer areas (Fig. 6f, Supplementary Fig. 26). Our electrochemical system also fulfils the minimum Faradaic efficiencies required to be profitable for different electricity costs up to 10 ¢/kWh (Supplementary Fig. 27).

Discussion

We developed an electrocatalyst:mediator assembly, in which the electrocatalyst promotes the mediator to highly reactive state capable of activating the inert hydrocarbons. This is demonstrated in the selective oxidation of toluene to benzaldehyde; CoO_x/IrO₂ electrocatalyst electrochemically activates NO₃^- mediator to a NO₃^● radical that can abstract a hydrogen atom from benzylic C-H in toluene. The electrocatalyst:mediator assembly a high Faradaic efficiency of 86( ±1)% towards benzaldehyde at 25 mA/cm², a factor of >3 times higher than direct oxidation on the bare electrocatalyst, as the reaction pathway with NO₃^● radical has an energy barrier that is 0.48 eV lower than the reaction pathway without. The synergistic effect between the CoO_x cluster and IrO₂ surface reduces the NO₃^● radical adsorption energy on the Co_interfacial site, which allows NO₃^● formed on the Co_interfacial site to desorb more readily after activation on the CoO_x/IrO₂ electrocatalyst. When translated into a flow system, a Faradaic efficiency of 60( ±4)% was achieved at the higher current density of 200 mA/cm², which can significantly reduce capital costs by achieving the same production rates with smaller electrolyzer areas.

Methods

Materials

All chemicals were used as purchased without further purification. Acetonitrile (suitable for HPLC, gradient grade, ≥99.9%, MeCN), ethylbenzene (ReagentPlus®, 99%), 4-chlorotoluene (98%), Iridium (III) chloride hydrate (reagent grade, 98%, IrCl₃·xH₂O), Cobalt(II) chloride hexahydrate (ACS reagent, 98%, CoCl₂·6H₂O), Gold (III) chloride trihydrate (≥99.9% trace metals basis), Lithium nitrate (ReagentPlus®, LiNO₃), oxalic acid (ReagentPlus®, ≥99%) and Lithium tetrafluoroborate (98%, LiBF₄), N-Hydroxyphthalimide (97%, for peptide synthesis, NHPI), 2,2,6,6-tetramethyl−1-piperidinyloxy (98%, TEMPO), Lithium bromide (ReagentPlus®, ≥99%, LiBr), Lithium chloride (ACS reagent, ≥99%, LiCl), Lithium perchlorate (99.99% trace metals basis, LiClO₄), p-Xylene (ReagentPlus®, 99%), 2,4-Dichlorotoluene (99%), 4-Chlorotoluene (98%), p-Toluic acid (98%), Ethylbenzene (ReagentPlus®, 99%), Diphenylmethane (99%), Propylbenzene (98%) 1,2,3,4-Tetrahydronaphthalene (ReagentPlus®, 99%) Tetrabutylammonium perchlorate (for electrochemical analysis, ≥99.0%, TBAP) were purchased from Sigma Aldrich. Dichloromethane (anhydrous (max. 0.005% H₂O) ≥99.8% stabilised, AnalaR NORMAPUR® analytical reagent, CH₂Cl₂), isopropanol (≥99.5%, CMOS for microelectronic, J.T.Baker®) were purchased from VWR Chemicals. Ti screen (0.002 inches thickness, Ti mesh) and Nafion® 117 membrane were purchased from FuelCellStore. Hydrochloric acid (Puriss. p.a., ACS Reagent, Reag. ISO, Reag. Ph. Eur., fuming, ≥37%, APHA: ≤10) and Nitric acid (Puriss. p.a., Reag. ISO, Reag. Ph. Eur., for determinations with dithizone, ≥65%) were purchased from Honeywell. Toluene (ACS Reagent) was purchased from TEDIA. Silver nitrate (99.9%-Ag, ACS, AgNO₃) was purchased from Strem Chemicals. The standard calibration gas mixtures for calibrating gas product at gas chromatography system and nitrogen gas (99.99%) were purchased from Air Liquide Singapore Pte. Ltd.

Synthesis of modified IrO₂ electrode

The CoO_x/IrO₂ electrocatalysts were synthesized through dip-coating followed by thermal decomposition²². Specifically, Ti mesh (cut into 10×10 mm²) etched in boiling 0.5 M oxalic acid solution for one hour. The etched Ti mesh was then soaked in a solution of 24 mg IrCl₃·xH₂O and 6 mg CoCl₂·6H₂O (or AuCl₃·3H₂O, total metal chloride salt amounting to 3 mg/mL ink) with 10% concentrated HCl in 10 mL isopropanol solution. The resulting Ti-mesh was then dried at 100 °C for 10 min and calcined at 500 °C for 10 min. The soaking and calcination cycle was repeated for 15 times, such that the catalyst loading is about 1.5 mg/cm² and the resultant materials were denoted as CoO_x/IrO₂ (or Au/IrO₂ when modified with gold).

Electrocatalyst characterization

Scan electron microscopy (SEM) images were obtained on a Hitachi SU8220 FESEM, at an accelerating voltage of 10 kV. Energy-dispersive X-ray (EDX) spectroscopy was performed with an on-board AEMTEK® Octane Elite EDS System, with an accelerating voltage of 20 kV. Transmission electron microscopy (TEM) images were obtained on TEM JEOL 2100 F. X-ray photoelectron spectroscopy (XPS) was conducted on a PHI Quantum 2000 XPS system. X-ray diffraction (XRD) analysis was operated on a Bruker® D8 ADVANCE system with an increment of 0.1°. X-Ray absorption spectroscopy (XAS) was conducted on XAFCA beamline at the Singapore Synchrotron Light Source (SSLS).

For electron paramagnetic resonance (EPR) measurements, electrochemical toluene oxidation was conducted on a reaction mix containing 0.01 M of 2,2,6,6-Tetramethylpiperidine 1-oxyl (TEMPO). Samples were periodically collected with a capillary tube and, within 30 s of sample collection, analyzed using a JEOL JES-X320 system with center field of 327 mT and sweep depth of 10 mT.

Evaluation of electrochemical performance

Electrochemical toluene oxidation experiments were carried out in a H-type cell (Tianjin Aida) equipped with anode (CoO_x/IrO₂ electrocatalyst on Ti mesh support) Ag/Ag⁺ reference electrode (Tianjin Aida), Pt plate cathode, and Nafion® 117 membrane (size 2.5 cm by 2.5 cm, thickness 0.18 mm). The Nafion® 117 membrane was activated in MilliQ water (18.2 Ω) at 70 °C for overnight before use. Unless otherwise stated, the anolyte is 19.5 mL of 0.1 M LiBF₄ and 0.06 M LiNO₃ acetonitrile solution and 0.5 mL toluene, and the catholyte is 20 mL of 0.1 M LiBF₄ MilliQ water (18.2 Ω) solution. All experiments were conducted using a Metrohm Autolab potentiostat PGSTAT204 without iR correction. Each electrochemical toluene oxidation experiment was conducted at 25 mA/cm² for 4800 s (120 C) unless otherwise stated. The electrolyte for each test was prepared individually directly before the testing.

The calibration of Ag/Ag⁺ reference electrode was carried out in a 3-electrode electrochemical cell³¹. Two glassy carbon electrodes with 0.5 mm diameter were employed as working and counter electrode; the to-be-calibrated Ag/Ag⁺ reference electrode (0.1 M TBAP and 0.01 M AgNO₃ acetonitrile solution) was employed as a reference electrode. The potentials of Fc/Fc⁺ (2.5 mM) redox pair were measured in 0.1 M TBAP acetonitrile solution and 0.1 M TBAP / 0.5 M [BMIM]PF₆ acetonitrile solution with a scan rate of 10 mV/s, respectively.

Cyclic Voltammetry (CV) and Linear Scan Voltammetry (LSV) were obtained in a one-pot electrochemical cell equipped with a 3-electrode system, including different materials as working electrode, Pt plate as counter electrode and Ag/Ag⁺ as reference electrode. Acetonitrile was used as solvent with 0.1 M Lithium tetrafluoroborate and 0.06 M Lithium nitrate as electrolyte. 0.5 mL toluene was added into the scan if required.

The double layer capacitance (${C}_{{dl}}$) values of CoO_x/IrO₂ anode from different Co/Ir ratios were determined by collecting cyclic voltammetry at different scan rates ($v$) of 10, 20, 40, 60, 80 and 100 mV/s. The potential range was -0.40 – -0.30 V vs. Ag/Ag⁺. The current density ($\Delta J$) at -0.35 V vs. Ag/Ag⁺ was then plotted against the corresponding $v$. The double-layer capacitance (${C}_{{dl}}$) was calculated from $\Delta J-v$ plot according to the formula:

$$\Delta J=v{C}_{{dl}}$$

The slopes for the $\Delta J-v$ plots are ${C}_{{dl}}$ values of CoO_x/IrO₂ anode from different Co/Ir ratios.

Flow cell tests were carried out in a standard 2-compartment flow cell system (purchased from Tianjin Aida) equipped with anode (CoO_x/IrO₂ electrocatalyst), Ag/Ag⁺ reference electrode, Ni mesh cathode, and Nafion® 117 membrane. The anolyte was 0.1 M LiBF₄ 0.06 M LiNO₃ and 340 µL toluene in 25 mL of pure acetonitrile solvent or 4:1 acetonitrile:HFIP mixture. The catholyte was 0.1 M aqueous LiBF₄ solution.

The generated benzaldehyde, benzyl alcohol and benzoic acid, as well as remaining toluene in anolyte was quantified with quantitative NMR (qNMR) spectroscopy using CH₂Cl₂ as internal standard at a JOEL JNM-ECA500II FT NMR System. Calibration curves were generated from qNMR spectra of standards containing different concentrations of the target chemicals, by integrating the specified peak proportional to the peak area of CH₂Cl₂. The resultant calibration curves showed R² > 0.991.

The calculation of Faradaic efficiency was then calculated through the following equation:

$${FE}(\%)=\frac{{nZF}}{Q} \times 100\%$$

where $n$ is the amount of product; $Z$ is the number of electrons transferred to obtain one molecule of product; $F$ is the Faraday constant (96485 C/mol) and $Q$ stands for the total passed charge in the reaction, which is 120 C unless otherwise stated.

Faradaic efficiency of hydrogen evolution reaction (HER) at the cathode side was evaluated through gas chromatography, using an on-line 990 micro GC system equipped with a thermal conductivity detector and a flame ionization detector. The measurement was using 10 sccm N₂ as carrier gas, and a sample was taken every 10 min. The calculation of FE was processed through the following quotation:

$${FE}(\%)={ZF} \times \frac{{vc}}{i \times {V}_{m}} \times 100\%$$

where $Z$ is the number of electrons transferred to require one molecule of product, in this hydrogen evolution reaction is 2; and $F$ stands for the Faraday constant (96485 C/mol); $v$ is the gas flow rate, in our measurement is 10 sccm, which is measured with a mass flow controller (Alicat Scientific); $c$ is the concentration of H₂ measured by GC; $i$ is the current applied and ${V}_{m}$ is the unit molar volume of gas.