Oxygen-rich g-C₃N₄ quantum dot engineered graphene cathode enabling highly selective 2e⁻ oxygen reduction reaction in a value-added reaction assisted Zn–air battery

Abstract

The electrochemical two-electron oxygen reduction reaction (2e⁻ ORR) offers a sustainable and energy-efficient alternative to the conventional anthraquinone process for hydrogen peroxide (H₂O₂) production. While metal-free carbon-based electrocatalysts have shown promise, optimizing their active sites for the selective 2e⁻ ORR remains a critical challenge. Given their nitrogen and oxygen-rich species, graphitic carbon nitride quantum dots (CNQDs) hold promise for promoting selective 2e⁻ ORR. Herein, we report the rational design of oxygen-rich CNQDs decorated on graphene frameworks (CNQDs/Gr), enabling the precise tailoring of active sites to promote the 2e⁻ ORR pathway. The rich C–O–C moieties within CNQDs effectively address the oxygen-deficient sites in Gr, serving as key active sites for selective 2e⁻ ORR for H₂O₂ generation. Furthermore, the interaction between the C atoms in Gr and neighboring N atoms in CNQDs facilitates π–π stacking, which stabilizes the *OOH intermediate and thus promotes selective H₂O₂ formation. These synergistic effects result in CNQDs/Gr with outstanding H₂O₂ selectivity of 96.5% at 0.6 V vs. RHE in 0.1 M KOH, far exceeding that of pristine Gr (∼60%). Furthermore, the CNQDs/Gr demonstrates outstanding performance in a value-added reaction assisted Zn–air battery (VAR-ZAB), achieving an impressive capacity of 768.58 mA h g⁻¹ with a high H₂O₂ production rate of 136 mmol g⁻¹ h⁻¹. This work presents a novel design strategy for metal-free carbon electrocatalysts, providing new insights into green H₂O₂ electrosynthesis.

---

Introduction

Hydrogen peroxide (H₂O₂) is one of the essential chemicals in the modern chemical industry. Its utility varies from being mainly used as a hygiene product in medical healthcare to a pollution-control agent for wastewater management. Meanwhile, the large-scale production of H₂O₂ still uses the anthraquinone process, which requires a large amount of energy, an intricate separation process, and strict safety controls.¹ To address these limitations, decentralized production methods are being actively explored. Among them, the electrochemical two-electron oxygen reduction reaction (2e⁻ ORR) presents a promising alternative, offering higher energy efficiency and cost-effectiveness.² The key challenge, however, lies in developing a selective and highly active electrocatalyst capable of producing H₂O₂ with high efficiency. Noble metal catalysts such as Pd have been considered benchmarks in alkaline media. Yet, they suffer from stability issues due to dissolution over time, making them unsuitable for long-term applications.^3–5 Therefore, the design of stable electrocatalyst materials for hydrogen peroxide production in alkaline media is critically important.

Extensive efforts have been devoted to engineering metal-free carbon materials to enhance both catalytic selectivity and activity.^6–9 Strategies such as heteroatom doping,^10,11 introducing oxygen-rich functional groups,^12–14 structural modification,⁷ and combining with other metal-free carbon materials¹⁵ have been explored to introduce more active sites and optimize intrinsic catalytic properties. As one of the efficient strategies to enrich the active site for H₂O₂ generation, introducing oxygen-rich functional groups, such as C–O–C or C=O, could modulate the sp² or sp³ hybridized carbon that generally has lower electrochemical adsorption ability, leading to easier formation of *OOH intermediates, which are crucial for selective H₂O₂ production.¹³

Among carbon-based catalysts, carbon nitride (C₃N₄) is particularly attractive due to its nitrogen-rich framework, tunable electronic properties, and high stability in harsh environments.¹⁶ However, its poor electrical conductivity remains a major drawback, limiting its catalytic performance. To overcome this challenge, researchers have explored structural modifications and hybridization strategies to enhance charge transport.^17–19 Some studies have demonstrated the promising application of carbon nitride in 2e⁻ ORR electrocatalysis. For instance, Yang et al. proposed the highly porous graphitic carbon nitride (g-C₃N₄) with tailored nitrogen species (pyridinic C–N=C and graphitic N₃(C)) to achieve high catalytic selectivity for H₂O₂ production.²⁰ They reported that graphitic nitrogen sites played a dominant role in H₂O₂ selectivity, reaching 90% efficiency, while a high surface area improved mass transport. In addition, they found that the graphitic nitrogen content is closely related to the higher catalytic selectivity compared to the pyridinic nitrogen content, which contributes to 90% H₂O₂ selectivity. Similarly, a recent study employed the combination of carbon nitride and carbon nanotubes (CNT) and highlighted the role of π–π stacking interaction in stabilizing *OOH intermediates during the reaction and promoting faster electron transfer.²¹ Moreover, the electrochemical performance exhibits a selectivity of 97%, which could be achieved at 0.50 V. Although several studies have explored carbon nitride and carbon composites for the 2e⁻ ORR, most have focused on bulk or polymeric carbon nitride structures, where the limited density of accessible surface functional groups and less effective electronic coupling with carbon supports could restrict their ability to achieve high selectivity and activity toward H₂O₂ production.

In this study, we report a novel strategy to enhance 2e⁻ ORR by decorating graphitic carbon nitride quantum dots (CNQDs) onto a graphene framework (CNQDs/Gr), as shown in Scheme 1, effectively tailoring the active sites for the 2e⁻ ORR pathway. Enriched with abundant C–O–C moieties, CNQDs serve as crucial active sites for the selective 2e⁻ ORR pathway, functionality that is typically deficient in graphene. The C–O–C sites supplement the oxygenated groups lacking in graphene and play a critical role in promoting H₂O₂ production. Moreover, introducing CNQDs onto the graphene framework facilitates distinct π–π stacking interactions between the pyridinic nitrogen of CNQDs and the edge sites of graphene. This interfacial interaction stabilizes the *OOH intermediate—an essential step in the two-electron reduction of oxygen—thereby favoring selective H₂O₂ production. Beyond its excellent electrocatalytic activity, the CNQDs/Gr composite also demonstrates multifunctionality in a value-added reaction assisted Zn–air battery (VAR-ZAB) system, enabling simultaneous electricity generation and H₂O₂ production. Thus, this study not only highlights a new strategy of carbon nitride-graphene coupling for the efficient electrosynthesis of H₂O₂, but also unravels the role of the tailored active sites in the VAR-ZAB.

Scheme 1 CNQDs/Gr as the electrocatalyst for hydrogen peroxide generation via 2e⁻ ORR in the value-added reaction assisted Zn–air battery (VAR-ZAB).

Experimental

Materials

Urea (NH₂CONH₂), with purity of 99.0–100.5%, nitric acid 65.0% solution, graphite flakes, and Nafion® 117 solution were purchased from Sigma Aldrich. Potassium hydroxide (KOH) for the electrolyte was obtained from Fisher Chemical. Potassium titanium(IV) oxalate (PTO) was purchased from Angene International Limited. All of the materials were used without further purification.

Synthesis of carbon nitride quantum dots (CNQDs)

The CNQDs were synthesized using a top-down method which requires the exfoliation of bulk graphitic carbon nitride (g-CN). First, the bulk graphitic carbon nitride was fabricated using 10 g of urea which was heated for 4 h at 550 °C under an air atmosphere. The resulting g-CN in the form of a porous yellow powder was then transferred into 100 ml of DIW with a weight of 100 mg. The mixture of bulk g-CN and DIW was sonicated for at least 12 h. After sonication, the mixture was turned into a white suspension in which the bigger particles were separated using a centrifuge 3 times at 8000 rpm. The clearer solution was obtained and dialyzed for 1 h to remove impurities to finally produce CNQDs dissolved in DIW.

Synthesis of exfoliated graphene oxide

Highly concentrated H₂SO₄ and H₃PO₄ were mixed in a 9 : 1 ratio. Then, 3.0 g of graphite flakes and 18.0 g of KMnO₄ were combined with the previous acid solution and heated to 50 °C with continuous stirring for 12 h. After the reaction finished, the mixture was cooled and 100 ml of DI water was added followed by reheating to 93 °C. The temperature was maintained at 93 °C and 300 ml of DI water with 15 ml of H₂O₂ (33%) were added. The mixture was once again cooled and centrifuged. The remaining solid powder was washed with 30% HCl four times, followed by DI water four times, and once with ethanol. Then, the final result of GO could be obtained after the clean solid powder was dried overnight at 80 °C. For comparison, pristine graphene was synthesized by annealing the GO for 2 h at 400 °C.

Synthesis of CNQDs/Gr

The suspension of CNQDs in DIW was varied relative to the GO weight for comparison. First, 50 mg of GO was dispersed in 50 ml of DI water followed by the addition of 20 ml of CNQD solution. The mixture was stirred for 12 h. Then, the solid powder was separated with a centrifuge and dried at 80 °C. To get the final result of CNQDs/Gr, the solid powder was annealed for 2 h at 400 °C. The optimization of CNQDs and graphene coupling are achieved by varying the CNQD concentration to 5 ml, 10 ml, 15 ml, 20 ml, and 25 ml (denoted as CNQDs/Gr-X, where X is the volume of CNQDs).

Material characterization

The morphological characterization of the material was performed using transmission electron microscopy (TEM, FEI Tecnai 20 G2 S-Twin transmission electron microscope). The optical characterization of CNQDs was conducted using a JASCO V-360 for the UV-vis absorbance, and a JASCO V676 absorbance spectrophotometer was used to obtain the photoluminescence (PL) excitation–emission spectra. The crystallinity of the material was analysed using an X-ray diffractometer (D2 Phaser, Bruker) with Cu Kα radiation (λ = 1.5418 Å). A RAMaker microscope spectrometer equipped with a 532 nm laser and CCD detector were used to observe the carbon structural defect in the catalyst. The functional groups of the samples were analysed by using a PerkinElmer attenuated total reflection Fourier transform infrared (ATR-FTIR) spectrometer (USA). X-ray photoelectron spectroscopy (XPS, JPS-9030, JEOL) was utilized for the chemical surface characterization of the materials. The L-edge X-ray absorption near edge structure (XANES) was acquired from hard X-ray absorption spectroscopy at the 17C1 beamline station of the National Synchrotron Radiation Research Center (NSRRC) in Hsinchu, Taiwan. In-situ attenuated total reflectance–surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS) was conducted using a Thermo Scientific Nicolet iS50 FTIR spectrometer equipped with a PIKE Technologies VeeMAX III variable-angle ATR accessory and a ZnSe internal reflection element. A custom-made glass spectroelectrochemical cell was mounted on the ATR crystal. A ∼50 nm Au film sputtered on a Si wafer served as the working electrode, onto which the as-prepared catalyst was drop-cast prior to measurement. An Ag/AgCl reference electrode and a Pt wire counter electrode were inserted through the sidearms of the cell. Potential-dependent SEIRAS spectra were acquired under chronoamperometric control by stepping the potential from 0.9 to 0.3 V vs. RHE. All measurements were performed under an O₂ atmosphere using a Metrohm Autolab potentiostat.

Electrochemical performance evaluation

The electrochemical performance evaluation of 2e⁻ oxygen reduction to hydrogen peroxide was investigated in a three-electrode cell using an Autolab potentiostat/galvanostat (PGSTAT302N with FRA32M, Metrohm Autolab). A rotating-ring disk electrode (RRDE), with a diameter of 4 mm and collection efficiency of N = 0.424, was used as the working electrode. The commercial Pt wire was utilized as the counter electrode. The operating potential was obtained from the conversion of Ag/AgCl (3 M KCl) as the reference electrode in O₂ saturated 0.1 M KOH electrolyte and calculated using the equation (eqn (1)) below.

The catalyst loading on the glassy carbon surface was pre-determined to be 0.1 mg cm⁻². The catalyst slurry was prepared using the formulation of 5 mg of catalyst mixed with 50 µl of 20% wt Nafion solution, 200 µl DIW, and 750 µl IPA followed by ultrasonication for 30 minutes to obtain the homogenized catalyst ink. The linear polarization curve from the disk current was acquired with a scan rate of 5 mV s⁻¹ with simultaneous recording for ring current (the applied potential for the Pt ring was set to 1.2 V vs. RHE). The selectivity, electron transfer number, and faradaic efficiency were calculated using the following equations (eqn (2)–(4)).

The parameters used for the above equations are described as follows: I_r is the ring current, I_d is the disk current, N is the collection efficiency, and n is the calculated electron transfer number. The linear polarization curve for the H₂O₂RR was obtained using the scan rate of 5 mV s⁻¹ in N₂ saturated 0.1 M KOH + 1 mM H₂O₂ electrolyte.

The bulk electrolysis method using an H-type cell was also assessed with 20 ml of 6.0 M KOH electrolyte in each of the anode and cathode compartments. The compartments were separated with an alkaline membrane, with Pt as a counter electrode, Ag/AgCl as the reference, and carbon paper which was drop-cast with the catalyst of 1 mg cm⁻² as a working electrode. The electrolyte was replaced every 3 hours. The faradaic efficiency (FE) of H₂O₂ was determined based on the yield of H₂O₂ relative to the total charge transferred, as expressed in eqn (5).

here, C denotes the concentration of H₂O₂ (mol L⁻¹), V is the electrolyte volume (L), F represents the Faraday constant (96 485 C mol⁻¹), and Q corresponds to the total charge passed (C). The total charge Q was obtained by integrating the measured current over the polarization duration, from t = 0 to t, at a specified applied potential as described by eqn (6):

Value-added reaction assisted Zn–air battery assembly and product quantification

The liquid zinc–air battery cell was utilized to test the applicability of the catalyst to generate electricity and H₂O₂ simultaneously. Zn foil with an area of 4 cm² was used as the anode part. 6 M KOH with 0.2 M zinc acetate solution was used as an electrolyte, separated by the alkaline membrane. The cathode was fabricated by dropping the catalyst onto a carbon paper with a gas diffusion layer using 1 mg cm⁻² controlled loading. The cathode part was purged with a controlled air flow of 10 ml min⁻¹. A Biologic Modular 6 Channels Electrochemical Workstation was utilized to assess the performance of the battery. The H₂O₂ production during selected time intervals was measured using a UV-vis titration method. The sample was mixed with DIW, 0.1 M K₂TiO(C₂O₄)₂, and 3.0 M H₂SO₄ before observing the wavelength at 400 nm. The calibration curve and the corresponding UV-vis spectra are presented in Fig. S5b and c.

Results and discussion

Characterization of the carbon nitride quantum dots (CNQDs)

The CNQD particles are prepared by the top-down synthesis which requires the size reduction of bulk graphitic carbon nitride to quantum-dot size.²² The CNQDs' unique properties are further identified by a set of physical, optical, and chemical surface characterizations. The TEM image in Fig. 1a depicts the well-dispersed CNQDs with particle sizes of 2–4.5 nm.²³ The calculation of CNQD interplanar spacing from the higher resolution of the TEM image in Fig. S1a indicates the lattice spacing of CNQDs with the size of 0.31 nm.²⁴ The optical properties of CNQDs were also analyzed using PL spectra. The contour plot presented in Fig. 1b reveals that the emission with the red color signal intensified at the wavelength of 440 nm and the excitation at 308 nm. This result was supported with the peak at 440 nm shown in PL emission spectra (Fig. S1b).²⁵ In the same plot, the absorbance peak from the UV-vis at 237 nm refers to the π–π* transition which specifically belongs to the s-triazine ring in carbon nitride and the peak at 315 nm refers to the n–π* transition of the heptazine heterocycles which are made up of individual triazine rings.²⁶ The inset picture from the same figure also reveals that the CNQD suspension emitted bright blue light after being exposed to the 365 nm light.

Fig. 1 (a) TEM image of the as-synthesized CNQDs. (b) 3D emission-excitation-intensity spectra of CNQDs. High resolution XPS spectra of (c) C 1s and (d) N 1s of CNQDs.

High-resolution XPS was used to gain deeper insight into the chemical surface characteristics of the CNQDs. The survey spectrum in Fig. S2a shows the prominent peaks of C, N, and O as the main elements contained in the CNQDs. The deconvolution peaks of C 1s spectra in Fig. 1c unveil the existence of C–C, C–N, C–O–C, C–N=C, and COOH/C=O at the binding energies of 284.79, 285.90, 288.10, and 289.10 eV, respectively.²⁷ The C–N and C–N=C bonds could be observed as the main attribute of nitrogen configuration in the carbon nitride structure. The N 1s spectrum in Fig. 1d shows that the highest peak deconvolution belongs to the C–N=C at 398.5 eV followed by N–(C)₃ (sp³ C–N) at 399.5 eV and C–NH at 401.1 eV.^28,29 The O 1s spectrum (Fig. S2) illustrates that the as-synthesized CNQDs also contain some functional groups of C=O, C–OH, and C–O at 531.2, 532.3, and 533.7 eV, respectively.²⁸

Characterization of carbon nitride quantum dot-decorated graphene (CNQDs/Gr)

The CNQDs are immobilized on the graphene (Gr) framework, owing to their large surface area, which provides abundant anchoring sites for CNQDs. Additionally, Gr itself is widely recognized as an electrochemically active material for the ORR, due to its excellent electrical conductivity and accessible active sites.^21,26 The resulting CNQDs/Gr was then characterized to investigate the contribution of the intrinsic properties of the CNQDs to the Gr. The TEM image in Fig. 2a reveals CNQDs distinctly scattered across the Gr frameworks, with the particle size between 3 and 5 nm, indicating the successful attachment of CNQDs to the Gr. Meanwhile, the FTIR spectra of CNQDs (Fig. S3a) highlight a characteristic peak at 800 cm⁻¹ appearing in the bulk graphitic carbon nitride (g-CN), which is associated with the triazine ring. This peak also appears in the wavelength spectra of CNQDs/Gr with lower intensity due to the smaller particle size and lower concentration of CNQDs distributed on Gr, indicating a successful coupling of CNQDs onto the Gr framework. The XRD patterns of bulk g-CN presented in Fig. S3b show that the obvious peaks at 13.0° and 27.6° are attributed to the (100) and (002) planes of typical bulk graphitic carbon nitride.³⁰ The prominent peaks at 26° belonging to the (002) plane specific to graphite appear in Gr and CNQDs/Gr samples, implying no significant structural change in the graphene after the attachment of CNQDs.

Fig. 2 (a) TEM image of CNQDs/Gr, (b) C K-edge and (c) N K-edge soft-XAS spectra of CNQDs/Gr and Gr. High resolution XPS spectra of (d) C 1s and (e) N 1s for CNQDs/Gr and Gr. (f) Schematic illustration of the CNQDs and Gr with the electron flow direction which establishes the π–π interaction in CNQDs/Gr.

To investigate the interaction between CNQDs and Gr, the local electronic structure and chemical surface were characterized using X-ray absorption spectroscopy (XAS) and X-ray photoelectron spectroscopy (XPS). The C k-edge spectra (Fig. 2b) show a prominent peak at 285.2 eV from the energy spectra of Gr and CNQDs/Gr, which refers to the sp² C=C hybridized carbon. The strong intensity of the sp² C=C from the Gr sample is established from the signal of the interlayer Gr. After the introduction of CNQDs, CNQDs/Gr spectra possess lower intensity due to the CNQD particle stacking interaction with a larger particle size of Gr.^31,32 Meanwhile, a broader peak at 287.6 eV, which corresponds to the C–O–C, could be observed in CNQDs/Gr as compared to Gr, implying a higher content of oxygen functional groups contributed by CNQDs.³³ The functional groups of C–O–C are very crucial for the formation of *OOH as the rate-determining step of the 2e⁻ ORR.³⁴ Additionally, the small peak observed around ∼289 eV is attributed to π(C=O)* transition originating from C=O (carbonyl/carboxyl) groups.^35,36 While this peak is already detectable in the graphene sample, its intensity is enhanced in the CNQDs/Gr spectrum, suggesting that the oxygen-rich CNQDs introduce additional surface functionalities. The strong peak at 291.5 eV correlates to the orbital transition of C–C (σ*), which is the characteristic peak of graphene, and appears in the Gr and CNQDs/Gr samples.³⁷ On the other hand, the N k-edge spectra in Fig. 2c showed noticeable peaks at 398.4 and 401.1 eV, which belong to the pyridinic (sp² C–N=C) and graphitic (sp³ C–N) features of heptazines, respectively, as the building blocks of the CNQDs.^38,39 Meanwhile, the corresponding peak of graphitic and pyridinic nitrogen for CNQDs/Gr shifted to lower photon energies due to the charge transfer from carbon in the graphene to the adjacent nitrogen of the CNQDs. This phenomenon can induce denser electron density and may induce π–π interaction, which stabilizes the OOH intermediates for further improving the selectivity of 2e⁻ ORR.²¹

To further confirm the phenomenon of π–π stacking interaction between CNQDs and Gr, high-resolution XPS was employed. Fig. 2d shows the deconvoluted C 1s spectra of Gr and CNQDs/Gr samples. The pristine Gr sample shows several characteristic peaks for C=C, C–C/C–N, and C–O–C at the binding energies of 284.79, 285.90, and 287.30 eV, respectively. Meanwhile, upon incorporation of CNQDs resulting in CNQDs/Gr, the carbon spectra show peaks of C–N=C and COOH/C=O at 288.25 and 289.10 eV, respectively, corresponding to the functional groups associated with CNQDs.^26,27 Notably, the C=C bonding of CNQDs/Gr shifts by 0.13 eV toward higher binding energy compared to pristine Gr, suggesting a lower electron density due to the electrons flowing from carbon in Gr to the adjacent nitrogen in CNQDs. Meanwhile, as shown in Fig. 2e, the N 1s spectra provide a comparison of CNQDs and CNQDs/Gr spectra deconvolution peaks. Herein, the peaks at 398.4 and 401.1 eV correspond to pyridinic (C–N=C) and graphitic (C–N) nitrogen, respectively. These characteristic peaks of CNQDs are also observed in the CNQDs/Gr but are shifted to lower binding energies by 0.20 eV (C–N=C) and 0.13 eV (C–N). This shift is attributed to an increase in electron density, where electrons from the carbon of Gr contribute to charge redistribution, thereby also influencing the observed shift in the XPS C 1s spectra. Such a combination of energy shifts between the positive shift in carbon and the negative shift in nitrogen is confirmed to induce the formation of π–π interactions, which helps stabilize the OOH intermediates, thus facilitating the selective 2e process.²¹

Furthermore, the peak assigned to C–O–C increases significantly in CNQDs/Gr compared to Gr. Due to the contribution from the oxygen functional group in CNQDs, the calculated peak deconvolution area of C 1s spectra (Table S1) shows the increase of the percentage of C–O–C of CNQDs/Gr (20.69%) compared to Gr (14.29%) and CNQDs (19.93%). The increasing content of the C–O–C functional group leads to more active sites that are available to generate *OOH intermediates.^14,40,41 Fig. S4a illustrates the peak deconvolution of O 1s spectra from CNQDs/Gr and Gr into three major peaks, which are C=O, C–O–C, and C–OH at the binding energies of 531.25, 532.66, and 533.92 eV, respectively.⁴¹ The C–O–C peak intensity increases in the CNQDs/Gr sample compared to Gr, which supports an increasing trend from the C–O–C bonding in C 1s XPS spectra and C k-edge XAS energy spectra. Furthermore, the Raman spectrum in Fig. S4b gives information on the increasing I_D/I_G ratio of CNQDs/Gr, which refers to the defect degree of a carbon-based material. The defect could originate from the edge sites of the CNQDs, which contain many functional groups, acting as the active site for the reactive *OOH species formation.^42,43Fig. 2f schematically illustrates the π–π stacking interaction between CNQDs and Gr, due to the electron flow from the carbon in Gr into the adjacent nitrogen in CNQDs, as supported by the results of XAS and XPS.

The electrocatalytic performance of CNQDs/Gr towards 2e⁻ ORR

To evaluate the influence of CNQDs/Gr toward the two-electron pathway of the ORR, the electroactivity of pristine Gr, CNQDs/Gr and the physically mixed bulk g-CN with Gr (g-CN/Gr) was investigated, as shown in Fig. 3a. The measurements were carried out in O₂-saturated 0.1 M KOH electrolyte using an RRDE system. Fig. 3a shows the LSV curves in the disk current and the corresponding ring current of Gr, CNQDs/Gr, and g-CN/Gr. The results clearly revealed that CNQDs/Gr possesses the highest ring current among all the samples, followed by g-CN/Gr and Gr, implying that more H₂O₂ product is detected. To validate the result of the 2e⁻ ORR activity, the electron transfer number (n) of Gr, CNQDs/Gr, and g-CN/Gr was calculated and presented in Fig. 3b. At the potential of 0.6 V vs. RHE, the CNQDs/Gr could achieve an electron transfer number of 2.07 surpassing g-CN/Gr (2.32) and Gr (2.61). The corresponding H₂O₂ selectivity was also measured, with the CNQDs/Gr exhibiting the highest selectivity of 96.51% at 0.6 V vs. RHE and maintaining 89.47% at 0.3 V vs. RHE compared to g-CN/Gr (83.73% at 0.6 V vs. RHE) and Gr (70.03% at 0.6 V vs. RHE). The preserved selectivity could indicate that CNQDs/Gr could provide a stable performance across a wide range of potentials. Fig. 3c presents the comparison of the faradaic efficiency for corresponding samples, showing that CNQDs/Gr achieved the highest value of 93.25% at 0.6 V vs. RHE, outperforming g-CN/Gr (72.01% at 0.6 V vs. RHE) and Gr (53.88% at 0.6 V vs. RHE). The enhanced two-electron ORR activity of CNQDs/Gr can be attributed to the high content of C–O–C oxygen functional groups introduced by the CNQDs onto the Gr framework, providing accessible sites for the formation of *OOH intermediates. The activity trend of electron transfer number, H₂O₂ selectivity, and faradaic efficiency shows the consistent sequence from high activity as follows: CNQDs/Gr > g-CN/Gr > Gr. Moreover, the selectivity of the catalyst was further confirmed by evaluating the performance of the H₂O₂ reduction reaction (H₂O₂RR) for each sample.⁴⁴ The LSV curve in Fig. 3d shows that the CNQDs/Gr catalyst exhibits the lowest reduction current compared to g-CN/Gr and Gr. The low activity from the H₂O₂RR indicates more stabilized *OOH intermediates and prevents further reduction of H₂O₂ to H₂O, which could also improve the selectivity of H₂O₂ production.⁴⁵ The above electrochemical results show that CNQDs/Gr effectively stabilizes the *OOH intermediates, attributed to the π–π interaction.

Fig. 3 (a) The LSV curves of Gr, CNQDs/Gr, and g-CN/Gr in O₂-saturated 0.1 M KOH. (b) The calculated electron transfer number (n) and H₂O₂ selectivity of Gr, CNQDs/Gr, and g-CN/Gr. (c) faradaic efficiency of Gr, CNQDs/Gr, and g-CN/Gr. (d) The LSV curves of Gr, CNQDs/Gr, and g-CN/Gr with 10 mM H₂O₂ in N₂-saturated 0.1 M KOH. All of the measurements were conducted using a RRDE at a 1600 rpm rotation speed. (e) Bulk polarization curve of CNQDs/Gr in oxygen and nitrogen saturated 6 M KOH electrolyte in an H-type cell. (f) Long term stability, hydrogen peroxide yield based on catalyst mass, and faradaic efficiency of the CNQDs/Gr. A constant potential of 0.6 V vs. RHE was applied to the H-type cell.

The bulk electrolysis method was carried out to validate the high performance of CNQDs/Gr using an H-type cell. Fig. 3d presents the bulk polarization curve of CNQDs/Gr, which achieved a current density of 77.43 mA cm⁻² at 0.2 V vs. RHE. Furthermore, Fig. 3e shows that a constant potential of 0.6 V vs. RHE was applied for 24 hours to evaluate long-term stability, during which a steady current output was maintained. The measured H₂O₂ yield reached a maximum value of 1305.52 mmol g_cat⁻¹, accompanied by a faradaic efficiency of 87.03%, which is comparable to the RRDE result of 93.25%. The post-ORR XPS analysis (Fig. S5) confirms the chemical stability of the CNQDs/Gr catalyst, with the characteristic binding positions of sp²-hybridized C and N species remaining largely unchanged after the reaction. This stability indicates that the π–π interfacial interaction between CNQDs and graphene is well preserved, maintaining effective charge redistribution across the hybrid interface. Moreover, a slight decrease in the C–O–C peak intensity was observed in both the C 1s and O 1s spectra, which may result from minor surface reduction during the reaction. Nevertheless, the C–O–C group remains the most abundant oxygen-containing species, demonstrating that the oxygen-rich active sites are structurally robust and retain their dominant contribution to the 2e⁻ ORR process.

In situ ATR-FTIR measurements were conducted to gain deeper mechanistic insight into the formation of the *OOH intermediate on the catalyst surface. The spectra were collected in 0.1 M KOH over the potential range of 0.9–0.3 V vs. RHE. As shown in Fig. 4a and b, both Gr and CNQDs/Gr exhibit a characteristic *OOH vibrational band at ∼1220 cm⁻¹,^46–48 with the CNQDs/Gr displaying a more pronounced intensity that increases as the applied potential decreases. This behavior suggests a higher degree of *OOH formation on CNQDs/Gr compared to Gr. In addition, the H–O–H bending mode in the region of 1600–1700 cm⁻¹ was observed for both catalysts and showed a progressive increase in intensity with decreasing potential, indicative of enhanced hydration around the H₂O₂ production site.⁴⁹ Interestingly, this increase was more moderate in CNQDs/Gr than in Gr. A similar trend was observed in the H–O–H stretching mode between 3000 and 3600 cm⁻¹ (Fig. S6), where the CNQDs/Gr again exhibited stronger features than Gr. Taken together, these results indicate that CNQDs/Gr not only promotes the formation of *OOH intermediates but also contributes to their stabilization more effectively than pristine graphene (Gr).

Fig. 4 (a) The in situ ATR-FTIR of Gr and (b) CNQDs/Gr in O₂-saturated 0.1 M KOH electrolyte.

CNQD concentration-dependent activity of CNQDs/Gr for H₂O₂ formation

To better understand the effect of CNQD loading of CNQDs/Gr composites for H₂O₂ formation, a series of CNQDs/Gr with varying CNQD concentrations were evaluated. A similar electrochemical evaluation was also conducted to elucidate the relationship between the various CNQD concentrations of the CNQDs/Gr and 2e⁻ ORR performance. Fig. 5a displays the LSV curves corresponding to the ring and disk currents of different CNQDs/Gr samples. Notably, CNQDs/Gr-20 exhibits the highest ring current among all CNQDs/Gr samples, indicating greater production of H₂O₂. Fig. 5b presents the H₂O₂ selectivity and electron transfer number (n), with CNQDs/Gr-20 showing the highest selectivity across a broad potential range. The incorporation of CNQDs into Gr significantly enhances H₂O₂ selectivity, surpassing 70% across various CNQD concentrations of CNQDs/Gr. Moreover, the electron transfer number remains below 2.7 for all CNQDs/Gr samples, approaching the ideal value of 2 electrons and confirming the catalyst selectivity for the 2e⁻ ORR. This enhanced activity highlights the key role of CNQDs in CNQDs/Gr towards 2e⁻ ORR activity. Correspondingly, the faradaic efficiency (FE) of Gr also improved upon CNQD integration across the various concentrations (Fig. S7a), which verifies the activity of CNQDs in facilitating efficient H₂O₂ generation. Furthermore, as shown in Fig. S7b, the highest performing CNQDs/Gr-20 maintains above 80% faradaic efficiency over a wide range of applied potentials, implying its stable catalytic performance. Fig. 5c illustrates the performance summary of CNQDs/Gr with various concentrations of CNQDs at 0.6 V vs. RHE, the potential at which the highest activity is observed, with CNQDs/Gr-20 exhibiting the highest activity, indicating an optimal CNQD loading for H₂O₂ selectivity and faradaic efficiency.

Fig. 5 (a) The LSV curves and the corresponding ring current of CNQDs/Gr with various GQD concentrations in O₂-saturated 0.1 M KOH. (b) The calculated H₂O₂ selectivity and electron transfer number of CNQDs/Gr with various GQD concentrations. (c) The relationship between the H₂O₂ selectivity and faradaic efficiency with the various GQD concentrations of CNQDs/Gr at 0.6 V vs. RHE. (d) The C K-edge soft-XAS spectra of the CNQDs/Gr with various GQD concentrations.

To gain perspective into the electronic state alteration of CNQDs/Gr through the GQD concentration within the graphene framework, soft-XAS was conducted. Fig. 4d presents the C K-edge spectra, wherein the prominent sp² carbon peak at 285.8 eV is notably intensified along with the increased CNQD loading. This enhanced peak intensity upon incorporating CNQDs into the graphene framework suggests an enhanced π–π stacking interaction between CNQDs and graphene, consistent with the observed electrochemical performance.⁵⁰ However, taking a look at the electrochemical performance, the CNQDs/Gr-25 exhibited reduced electrochemical performance, which deviates from this trend. Raman spectroscopy (Fig. S9) reveals that the I_D/I_G ratio of CNQDs/Gr-25 decreases compared to CNQDs/Gr-20, indicating a reduction in structural defects. This decrease in defects may lead to fewer active sites that are essential for catalytic activity.⁵¹ Overall, the enhancement of the 2e⁻ ORR performance was attributed to several points: (1) the highly active C–O–C functional groups in CNQDs facilitate the formation of reactive OOH species during the reaction, and (2) π–π stacking interaction exhibited from CNQDs and graphene could act as the active sites for stabilizing the *OOH intermediates, which is followed by the optimum CNQD concentration. The collaborative effect from these two phenomena enhances both the selectivity and activity of the CNQD catalyst.

Value-added reaction assisted Zn–air battery (VAR-ZAB) application

The performance of the CNQDs/Gr for H₂O₂ formation was evaluated in a value-added reaction assisted Zn–air battery (VAR-ZAB), and pristine Gr was also used as a comparison. In this system, the anode part of the battery undergoes the Zn oxidation reaction, while the cathode part of the battery performs the 2e⁻ ORR, which produce H₂O₂ simultaneously (Fig. 6a). After a 10 minutes rest period post-charging, the open-circuit voltage (OCV) was recorded. The CNQDs/Gr-based VAR-ZAB (Fig. S10a) achieved an OCV of 1.69 V, significantly higher than the 1.44 V observed for the Gr-based one. The power density of the VAR-ZAB was measured by analyzing the polarization curve of the cell within a voltage range of 1.4 to 0.1 V. The VAR-ZAB with the CNQDs/Gr catalyst demonstrated a maximum power density of 47.48 mW cm⁻², outperforming the Gr-based system, which reached only 43.03 mW cm⁻², as shown in Fig. 6b. The battery was discharged at a constant potential of 0.6 V, identified as the optimum potential for achieving the highest H₂O₂ yield with the CNQDs/Gr catalyst for the VAR-ZAB. As shown in Fig. 6c, specific capacity measurements which were conducted at a constant current density of 10 mA cm⁻² revealed that the VAR-ZAB with a CNQDs/Gr catalyst delivers a higher specific capacity of 768.58 mA h g⁻¹, outperforming the one with a Gr catalyst (685.22 mA h g⁻¹). Additionally, the time-based specific discharge in Fig. S10b shows that the CNQDs/Gr lasted for 132.91 min to reach 0 V, which is higher compared to Gr that lasted for 111.23 min. On the other hand, to track H₂O₂ production during discharge for the VAR-ZAB, the electrolyte was sampled every 15 minutes and analyzed using UV-vis titration. The result was plotted using a linear profile in Fig. 6d to calculate the H₂O₂ production rate, normalized based on the catalyst loading. Remarkably, the VAR-ZAB with CNQDs/Gr shows an excellent production rate of 59.08 mmol g_cat⁻¹ min⁻¹ compared to the one with Gr (24.19 mmol g_cat⁻¹ min⁻¹). The production rate presented in Fig. 6e also highlights the consistent efficiency of the VAR-ZAB with the CNQDs/Gr catalyst based on the catalyst loading (yield@M), which reaches up to 136 mmol g⁻¹ h⁻¹ and the geometric area of the catalyst (yield@A), which reaches up to 3992 mmol cm⁻² h⁻¹. The comparison of faradaic efficiency and H₂O₂ production yield in Fig. 6f validates the superior performance of CNQDs/Gr relative to recently published catalysts for H₂O₂ electro-generation, with detailed values for each catalyst given in Table S1.

Fig. 6 (a) Schematic diagram of the liquid 2e⁻ ORR zinc–air battery. (b) Power density and discharge polarization curves between Gr and CNQDs/Gr. (c) Specific capacity of Gr and CNQDs/Gr. (d) Linear plot comparison of the H₂O₂ production yield of Gr and CNQDs/Gr. (e) H₂O₂ production rate based on catalyst loading and electrode geometric area using CNQDs/Gr as catalyst. (f) Comparison of H₂O₂ production yield and faradaic efficiency of CNQDs/Gr with other reported catalysts.

Conclusion

In summary, this work presents a pioneering way to enhance the selective 2e⁻ ORR by engineering CNQDs onto a graphene framework. This design effectively tailors active sites of the oxygen-containing functional groups, which are typically deficient in pristine graphene, for efficient H₂O₂ generation via 2e⁻ ORR. Due to enrichment with CNQD-originated C–O–C moieties, as confirmed by soft-XAS, which compensate for the oxygen-deficient sites in graphene, new active sites are created which promote the selective 2e⁻ ORR pathway. Furthermore, the π–π stacking interactions between the pyridinic nitrogen in CNQDs and graphene stabilize the critical *OOH intermediate, a key intermediate in the two-electron reduction mechanism, thereby boosting the selectivity of H₂O₂ formation. Such a synergistic effect can be finely tuned by adjusting the CNQD loading. The resulting CNQDs/Gr exhibits outstanding electrocatalytic performance, achieving H₂O₂ selectivity of over 95% at 0.6 V vs. RHE, significantly higher than that of pristine graphene (∼60%) while maintaining over 80% H₂O₂ selectivity across a broad potential range. In practical application within a VAR-ZAB system, the CNQDs/Gr as the air catalyst delivers a high specific capacity (768.58 mA h g⁻¹) and a H₂O₂ production rate up to 136 mmol g⁻¹ h⁻¹. Overall, this work offers valuable insights into tailoring active sites in metal-free carbon catalysts, providing a promising platform for sustainable and efficient H₂O₂ electrosynthesis and its applications.

Author contributions

The manuscript was written through the contributions of all authors. All authors have approved the final version of the manuscript. These authors contributed equally.

Conflicts of interest

The authors declare no competing financial interest.

Data availability

Data will be made available on request from the authors. The data that support the findings of this study are available from the corresponding author, M. H. Yeh, upon reasonable request.

The data supporting this article have been included as part of the Supplementary Information (SI) which contains supporting characterizations, electrochemical performances, and further experimental details. See DOI: https://doi.org/10.1039/d5ta09143h.

Acknowledgements

This work was financially supported by the National Science and Technology Council (NSTC) in Taiwan (NSTC 112-2221-E-011-016-MY3, NSTC 113-2628-E-011-004-MY3, NSTC 113-2222-E-027-009, NSTC 113-2221-E-606-010-MY3, NSTC 114-2221-E-027-010-MY3, and NSTC 114-2218-E-011-002), the National Taiwan University of Science and Technology & Institut Teknologi Sepuluh Nopember Joint Research Program (ITS-NTUST-2025-03), and the industry-academia cooperation project (NTUST-Might Electronic-No. 10959). This work was also financially supported by the “Sustainable Electrochemical Energy Development Center (SEED)” from The Featured Areas Research Center Program within the framework of the Higher Education Sprout Project by the Ministry of Education (MOE) in Taiwan. The research facilities for soft-XAS were provided by the beamline of BL20A1 at the National Synchrotron Radiation Research Center (NSRRC) in Taiwan. Thanks to Ms. C.-Y. Chien of the Ministry of Science and Technology (National Taiwan University) for the assistance with TEM experiments.

References

1. T. Nishimi, T. Kamachi, K. Kato, T. Kato and K. Yoshizawa, Eur. J. Org Chem., 2011, 2011, 4113–4120.
2. J. An, Y. Feng, Q. Zhao, X. Wang, J. Liu and N. Li, Environ. Sci. Ecotechnology, 2022, 11, 100170.
3. E. Pizzutilo, S. J. Freakley, S. Cherevko, S. Venkatesan, G. J. Hutchings, C. H. Liebscher, G. Dehm and K. J. J. Mayrhofer, ACS Catal., 2017, 7, 5699–5705.
4. G. V. Fortunato, E. Pizzutilo, A. M. Mingers, O. Kasian, S. Cherevko, E. S. F. Cardoso, K. J. J. Mayrhofer, G. Maia and M. Ledendecker, J. Phys. Chem. C, 2018, 122, 15878–15885.
5. M. Selinsek, B. J. Deschner, D. E. Doronkin, T. L. Sheppard, J.-D. Grunwaldt and R. Dittmeyer, ACS Catal., 2018, 8, 2546–2557.
6. Y. Zhou, G. Chen and J. Zhang, J. Mater. Chem. A, 2020, 8, 20849–20869.
7. D. Zhang, E. Mitchell, X. Lu, D. Chu, L. Shang, T. Zhang, R. Amal and Z. Han, Mater. Today, 2023, 63, 339–359.
8. W. Peng, H. Tan, X. Liu, F. Hou and J. Liang, Energy Fuels, 2023, 37, 17863–17874.
9. Y. Bu, Y. Wang, G.-F. Han, Y. Zhao, X. Ge, F. Li, Z. Zhang, Q. Zhong and J.-B. Baek, Adv. Mater., 2021, 33, 2103266.
10. A. Byeon, W. C. Yun, J. M. Kim and J. W. Lee, Chem. Eng. J., 2023, 456, 141042.
11. Z. Tian, Q. Zhang, L. Thomsen, N. Gao, J. Pan, R. Daiyan, J. Yun, J. Brandt, N. López-Salas, F. Lai, Q. Li, T. Liu, R. Amal, X. Lu and M. Antonietti, Angew. Chem., Int. Ed., 2022, 61, e202206915.
12. Y. Tian, D. Deng, L. Xu, M. Li, H. Chen, Z. Wu and S. Zhang, Nano-Micro Lett., 2023, 15, 122.
13. G. Zhao, T. Chen, A. Tang and H. Yang, Chem.–Eur. J., 2024, 30, e202304065.
14. Q. Zhang, X. Tan, N. M. Bedford, Z. Han, L. Thomsen, S. Smith, R. Amal and X. Lu, Nat. Commun., 2020, 11, 4181.
15. H. He, S. Liu, Y. Liu, L. Zhou, H. Wen, R. Shen, H. Zhang, X. Guo, J. Jiang and B. Li, Green Chem., 2023, 25, 9501–9542.
16. W. Niu and Y. Yang, ACS Energy Lett., 2018, 3, 2796–2815.
17. S. M. Lyth, Y. Nabae, S. Moriya, S. Kuroki, M.-a. Kakimoto, J.-i. Ozaki and S. Miyata, J. Phys. Chem. C, 2009, 113, 20148–20151.
18. K. Qiu and Z. X. Guo, J. Mater. Chem. A, 2014, 2, 3209–3215.
19. K. Kwon, Y. J. Sa, J. Y. Cheon and S. H. Joo, Langmuir, 2012, 28, 991–996.
20. Z. Yang, C. Zhang, Y. Mei, Y. Zhang, Y. Feng, M. Shao and J. Hu, Adv. Mater. Interfaces, 2022, 9, 2201325.
21. J. Qu, G. Long, L. Luo, Y. Yang, W. Fan and F. Zhang, Small, 2024, 20, 2400695.
22. H.-x. Zhong, Q. Zhang, J. Wang, X.-b. Zhang, X.-l. Wei, Z.-j. Wu, K. Li, F.-l. Meng, D. Bao and J.-m. Yan, ACS Catal., 2018, 8, 3965–3970.
23. Z. Song, T. Lin, L. Lin, S. Lin, F. Fu, X. Wang and L. Guo, Angew. Chem., Int. Ed., 2016, 55, 2773–2777.
24. M. Majdoub, D. Sengottuvelu, S. Nouranian and A. Al-Ostaz, ChemSusChem, 2024, 17, e202301462.
25. C. Zheng, X. Qiu, J. Han, Y. Wu and S. Liu, ACS Appl. Mater. Interfaces, 2019, 11, 42243–42249.
26. S. Hou, C. Yu, X. Song, Y. Ding, J. Chang, Y. Liu, L. Chen, Q. Wei, X. Zhang and J. Qiu, Adv. Funct. Mater., 2023, 33, 2212112.
27. G. Bai, Z. Song, H. Geng, D. Gao, K. Liu, S. Wu, W. Rao, L. Guo and J. Wang, Adv. Mater., 2017, 29, 1606843.
28. H. Liu, X. Lv, J. Qian, H. Li, Y. Qian, X. Wang, X. Meng, W. Lin and H. Wang, ACS Nano, 2020, 14, 13304–13315.
29. Z. Zhou, Y. Shen, Y. Li, A. Liu, S. Liu and Y. Zhang, ACS Nano, 2015, 9, 12480–12487.
30. F. Fina, S. K. Callear, G. M. Carins and J. T. S. Irvine, Chem. Mater., 2015, 27, 2612–2618.
31. W. Zhou, S. J. Pennycook and J.-C. Idrobo, J. Phys.: Condens. Matter, 2012, 24, 314213.
32. W. Hua, B. Gao, S. Li, H. Ågren and Y. Luo, Phys. Rev. B: Condens. Matter Mater. Phys., 2010, 82, 155433.
33. S. Wang, S. Zhou, Z. Ma, N. Gao, R. Daiyan, J. Leverett, Y. Shan, X. Zhu, Y. Zhao, Q. Liu, R. Amal, X. Lu, T. Liu, M. Antonietti, Y. Chen, Q. Zhang and Z. Tian, Angew. Chem., Int. Ed., 2025, e202501896.
34. Y. J. Sa, J. H. Kim and S. H. Joo, Angew. Chem., Int. Ed., 2019, 58, 1100–1105.
35. L. Curti, O. W. Moore, P. Babakhani, K.-Q. Xiao, C. Woulds, A. W. Bray, B. J. Fisher, M. Kazemian, B. Kaulich and C. L. Peacock, Communications Earth & Environment, 2021, 2, 229.
36. A. Kolodziejczyk, J. Wheeler, N. T. Tran, C. Jaye and D. Knorr Jr, Langmuir, 2023, 39, 18289–18301.
37. C. He, J. Lei, X. Li, Z. Shen, L. Wang and J. Zhang, Angew. Chem., Int. Ed., 2024, 63, e202406143.
38. N. Meng, J. Ren, Y. Liu, Y. Huang, T. Petit and B. Zhang, Energy Environ. Sci., 2018, 11, 566–571.
39. S. Sainio, N. Wester, A. Aarva, C. J. Titus, D. Nordlund, E. I. Kauppinen, E. Leppänen, T. Palomäki, J. E. Koehne, O. Pitkänen, K. Kordas, M. Kim, H. Lipsanen, M. Mozetič, M. A. Caro, M. Meyyappan, J. Koskinen and T. Laurila, J. Phys. Chem. C, 2021, 125, 973–988.
40. T. Lu, M. Sun, F. Wang, S. Chen, Y. Li, J. Chen, X. Liao, X. Sun, Y. Liu, F. Wang, B. Huang and H. Wang, ACS Nano, 2024, 18, 15035–15045.
41. K. Lee, J. Lim, M. J. Lee, K. Ryu, H. Lee, J. Y. Kim, H. Ju, H.-S. Cho, B.-H. Kim, M. C. Hatzell, J. Kang and S. W. Lee, Energy Environ. Sci., 2022, 15, 2858–2866.
42. K.-H. Wu, D. Wang, X. Lu, X. Zhang, Z. Xie, Y. Liu, B.-J. Su, J.-M. Chen, D.-S. Su, W. Qi and S. Guo, Chem, 2020, 6, 1443–1458.
43. Y. Wang, T. Zhang, D. Li, P. Li, Q. Hu, Q. Zhuang, L. Duan and J. Liu, J. Mater. Chem. A, 2024, 12, 23398–23405.
44. Y. Yuan, H. Li, Z. Jiang, Z. Lin, Y. Tang, H. Wang and Y. Liang, Chem. Sci., 2022, 13, 11260–11265.
45. E. Jung, H. Shin, B.-H. Lee, V. Efremov, S. Lee, H. S. Lee, J. Kim, W. Hooch Antink, S. Park, K.-S. Lee, S.-P. Cho, J. S. Yoo, Y.-E. Sung and T. Hyeon, Nat. Mater., 2020, 19, 436–442.
46. Z. Sang, Y. Qiao, R. Chen, L. Yin, F. Hou and J. Liang, Nat. Commun., 2025, 16, 4050.
47. J. Zhang, H. Zeng, B. He, Y. Liu, J. Xu, T. Niu, C. Pan, Y. Zhang, Y. Lou, Y. Wang, Y. Dong and Y. Zhu, Appl. Catal., B, 2024, 358, 124448.
48. Z. Pei, Y. Li, G. Fan, Y. Guo, D. Luan, X. Gu and X. W. Lou, Small, 2024, 20, 2403808.
49. J. Liu, Z. Zhang, X. Wang, C. Han and M. Yang, J. Am. Chem. Soc., 2025, 147, 36618–36625.
50. M. I. De Barros Bouchet, J. M. Martin, J. Avila, M. Kano, K. Yoshida, T. Tsuruda, S. Bai, Y. Higuchi, N. Ozawa, M. Kubo and M. C. Asensio, Sci. Rep., 2017, 7, 46394.
51. A. Olean-Oliveira, N. Hasnain, R. Martínez-Hincapié, U. Hagemann, A. Jain, D. Segets, I. Spanos and V. Čolić, ACS Catal., 2024, 14, 17675–17689.

---