Crystalline nitrogen-doped-carbon anchored well-dispersed Fe₃O₄ nanoparticles for real-time scalable neutral H₂O₂ electrosynthesis

Abstract

Salt-free neutral H₂O₂ electrosynthesis via a 2-electron oxygen reduction reaction (2e⁻-ORR) remains challenging owing to the absence of efficient electrocatalysts and well-matched practical processes. Herein, we report an important progress and understanding of neutral H₂O₂ electrosynthesis of 2e⁻-ORR at a scalable rate using crystalline nitrogen-doped-carbon anchored Fe₃O₄ nanoparticles (NPs, Fe₃O₄@TNC) as efficient electrocatalysts, which were derived from the pyrolysis of a mixture of g-C₃N₄ and Fe@Tpy, achieving a salt-free, real-time and continuous H₂O₂ production process. Based on rotating ring-disk electrodes, Fe₃O₄@TNC achieved nearly 100% selectivity from 0 to 0.75 V vs. RHE and a limiting diffusion current density up to 5.2 mA cm⁻² at 0 V vs. RHE. It was revealed that the exposed (220) facet of Fe₃O₄ NPs obtained a thermodynamically optimal binding of *OOH and rapid *OOH-mediated kinetic pathway. The integration of Fe₃O₄@TNC into scalable cells exhibited superior performance and techno-economic potential for neutral H₂O₂ electrosynthesis as industrially relevant current densities were achieved with remarkable real-time continuous production while maintaining relatively large faradaic efficiency. This work provides in-depth mechanistic insights into neutral H₂O₂ electrosynthesis and offers an advanced and economical process for integrating efficient electrocatalysts and scalable electrolyzer for industrially relevant neutral H₂O₂ production.

---

Broader context

Neutral H₂O₂ electrosynthesis via oxygen reduction reaction is an attractive route for sustainable and on-site application. However, it remains challenging to design efficient electrocatalysts at industrially relevant and affordable production rates. Herein, crystalline nitrogen-doped-carbon-anchored well-dispersed Fe₃O₄ nanoparticles (NPs, Fe₃O₄@TNC) were constructed by pyrolyzing a mixture of g-C₃N₄ and Fe²⁺-terpyridine. Fe₃O₄@TNC showed superior performance for neutral H₂O₂ production, achieving nearly 100% selectivity from 0 to 0.75 V vs. RHE and a limiting diffusion current density up to 5.2 mA cm⁻² at 0 V vs. RHE. It was revealed that the mainly exposed (220) facet of Fe₃O₄ NPs facilitated thermodynamically optimal binding of *OOH and rapid *OOH-mediated kinetic pathway. Moreover, integrating Fe₃O₄@TNC into scalable cells led to a record production rate of 8.89 mol h⁻¹ g_cat.⁻¹ and 19.23 mM for real-time continuous salt-free neutral H₂O₂ production, showing a techno-economical H₂O₂ production rate.

Introduction

Hydrogen peroxide (H₂O₂) is an important green oxidant in large-scale industrial processes and smaller on-site activities,¹ such as industrial bleaching,² chemical synthesis,³ disinfection,⁴ and fuel cell technologies.⁵ Compared to the current process for industrial H₂O₂ production,⁶ H₂O₂ electrosynthesis via a 2-electron oxygen reduction reaction (2e⁻-ORR) is a promising alternative approach that is applicable on different scales for H₂O₂ production in a green and sustainable manner.^1,7–9 The current main challenge is achieving high efficiency at industrially relevant current densities for a positive techno-economic value of H₂O₂ electrosynthesis.^9,10 Although a desired system of neutral H₂O₂ electrosynthesis was presented,^10–15 it is not yet satisfactory for large-scale productivity at an affordable cost owing to the competitive reduction reaction of the accumulated H₂O₂, extra separation purity from the salt-containing electrolyte, and further concentration costs. Therefore, it is highly promising to develop a salt-free and scalable neutral H₂O₂ electrosynthesis system, particularly including efficient electrocatalyst and a well-matched process.

To date, although many efficient electrocatalysts have been developed for neutral 2e⁻-ORR electrocatalysis,^13,15–18 they rarely show versatile production potential for neutral H₂O₂ electrosynthesis owing to their relatively large overpotential, low current density, and lack of a process for continuous neutral H₂O₂ electrosynthesis.^19–21 In particular, this kind of neutral electrolytes or salt-free aqueous solution usually has high ohmic loss and low ionization, resulting in sluggish ORR kinetics and low binding of OOH*.²² For example, although porous carbon-supported Pd nanoparticles (NPs) achieved a high selectivity of neutral 2e⁻-ORR of about 95%, its ORR current density is still lower than the alkaline activity of this kind of catalysts. Similarly, transition metal-based or metal-free carbon electrocatalysts also present a similar trend of selectivity for neutral 2e⁻-ORR, and their overpotential and current density are also far less than those under alkaline conditions.^23,24 Besides, even though these efficient electrocatalysts in a flow cell contribute a considerable H₂O₂ yield, their neutral sluggish kinetics for 2e⁻-ORR also limit the continuous accumulation of H₂O₂ due to a thermodynamic decomposition potential of H₂O₂ of about 1.766 V.¹⁹ Therefore, it is highly challenging to design efficient electrocatalysts and rational neutral H₂O₂ electrosynthesis systems to realize superior and scalable neutral H₂O₂ production.

Compared with various types electrocatalysts, carbon-based metal electrocatalysts, benefiting from their low-cost, easily adjustable structure of active sites, optimizable number of active sites, and high conductivity, show more positive onset potential and relatively large current density of neutral 2e⁻-ORR.^25–27 However, owing to water as the proton supply, the kinetics of *OOH formation over carbon-based single-atom metal electrocatalysts is slow, requiring a more negative potential for H₂O₂ electrosynthesis.²⁶ Fortunately, carbon-supported metal NPs can achieve higher current density of 2e⁻-ORR at a more positive potential due to the available active sites accelerating the proton coupling electron transfer process to hydrogenate the adsorbed O₂, thus forming *OOH.^28,29 On the other hand, pure metal NPs as active site for ORR thermodynamically favor cleavage of the O–O bond because of the favorable adsorption of O₂ in the side-on manner.³⁰ In contrast, carbon-anchored metal compound NPs, owing to the well-dispersed metal atom, show great potentials for theoretically higher efficient neutral 2e⁻-ORR at more positive potential ranges.²⁶ However, neutral H₂O₂ electrosynthesis still has not achieved a practical rate and industrially relevant current densities owing to the rapid consumption of proton in water, which causes alkalization of aqueous solution.¹⁹ Therefore, it is significantly meaningful to integrate efficient carbon-anchored metal compound NPs electrocatalysts and scalable electrolyzers for developing an advanced neutral H₂O₂ electrosynthesis process.

Herein, we report crystalline nitrogen-doped-carbon anchored well-dispersed Fe₃O₄ nanoparticles (NPs, Fe₃O₄@TNC) as efficient electrocatalysts toward real-time continuous and scalable neutral H₂O₂ production (Scheme 1). Experimental observations and theoretical calculations indicated that Fe₃O₄@TNC with a main exposure (220) facet achieved a high onset potential of about 0.75 V vs. RHE, an exclusive neutral H₂O₂ selectivity at a record current density, attributed to the main exposed Fe₃O₄ (220) facet with an optimal binding of *OOH and rapid *OOH-mediated kinetic pathway. The integration of Fe₃O₄@TNC into scalable cells exhibited scalable neutral H₂O₂ electrosynthesis with a positive techno-economic potential.

Scheme 1 Schematic of the synthesis procedure of Fe₃O₄@TNC and the integration of Fe₃O₄@TNC into solid-state electrolysis (SE) cell for scalable neutral H₂O₂ electrosynthesis via O₂ reduction.

Results and discussion

Synthesis and structural characterization of catalysts

The formation mechanism of Fe₃O₄@TNC electrocatalyst was first revealed by X-ray photoelectron spectroscopy (XPS), thermogravimetry (TG), and Fourier transform infrared (FTIR) spectroscopy to monitor the pyrolysis process of precursors. The XPS spectra in Fig. S1 (ESI†) suggest that both g-C₃N₄ and Tpy as precursors mainly contain two kinds of carbon and nitrogen elements. Moreover, their high-resolution XPS C 1s and N 1s spectra were deconvoluted into a series of peaks that correspond to the typical position of g-C₃N₄ and Tpy.^31,32 The HR-XPS O 1s spectra for g-C₃N₄ and Tpy can also be deconvoluted into H₂O adsorption of two peaks (Fig. S1, ESI†). Fe@Tpy as the precursor includes C, N, Fe, Cl, and O five kinds of elements (Fig. S1, ESI†), which are ascribed to FeCl₂ and methanol used for the assembly of Fe@Tpy.³² In contrast to Tpy, a new peak appears in the HR-XPS C 1s and O 1s spectrum, assigned to the C–O bond possibly due to the existence of methanol. TG-FTIR was further used to investigate the pyrolysis process of the mixture of both Fe@Tpy and g-C₃N₄ (Fig. 1(a) and Fig. S2, S3, ESI†). As the pyrolysis temperature increases, the peak at 1293 cm⁻¹ appears and gradually increases for pure Fe@Tpy and the mixture, assigned to the vibration of C–O bond,³³ indicating that methanol was partially encapsulated in Fe@Tpy.

Fig. 1 (a) 3D colormap surface of FTIR for the escaping gases for carbonizing the mixture of g-C₃N₄ and Fe@Tpy. (b) TEM image and (c) HRTEM image of Fe₃O₄@TNC. (d) HAADF-STEM and EDS mapping (C, N, O, and Fe) of Fe₃O₄@TNC. (e) X-ray diffraction pattern of NC, TNC, Fe@Tpy, and Fe₃O₄@TNC. (f) Raman spectra of TNC, Fe₃O₄@TNC, and Fe₃O₄@TNC-AE. (g) Fe K-edge XANES spectra of Fe₃O₄@TNC, Fe foil, FeO, Fe₂O₃, Fe₃O₄, and FePc. (h) Linear fit of Fe valence versus Fe K-edge energy position, (i) FT k²-weighted χ(k)-functions of the EXAFS spectra of the Fe K-edge for Fe₃O₄@TNC and the reference samples.

Moreover, with the decomposition of g-C₃N₄ at the beginning of 620 °C (Fig. 1(a) and Fig. S2, ESI†), it could be observed that the peak of C–O for the mixture also decreased significantly at about 700 °C (Fig. 1(a)). The peak of C–O keeps increasing from about 500 to 800 °C during the pyrolysis of Fe@Tpy (Fig. S2 and S3a, ESI†). These results suggest that methanol should be an oxygen source for Fe₃O₄@TNC. Compared with the XPS of C, N, and O 1s for TNC derived from carbonizing the mixture of g-C₃N₄ and Tpy, Fe₃O₄@TNC presents new peaks for C–O (287.0 eV), Fe–N (399.5 eV), and Fe–O (530.1 eV)³⁴ (Fig. S4 and Tables S1, S2, ESI†). Moreover, in contrast to the Fe 2p HR-XPS of Fe@Tpy (Fig. S1j, ESI†), Fe₃O₄@TNC presents a low-intensity peak for Fe of zero value (Fig. S4g, ESI†), suggesting the formation of a little bit of metallic Fe. Besides, the intensity of the Fe²⁺ and Fe³⁺ peaks dominate in the Fe 2p HR-XPS of Fe₃O₄@TNC (Fig. S4g, ESI†), confirming the main existence of Fe metal in the zero oxidation state. In contrast to the porous layer structure of NC, which originated from the pyrolysis of pure g-C₃N₄, TNC displays a typical lamellar stack, and Fe₃O₄@TNC shows an interconnected and stacked structure consisting of stacked linear and layer structure (Fig. S5, ESI†) derived from the carbon framework of the linear polymer of Fe@Tpy and g-C₃N₄, indicating that iron species were confined by the nitrogen-doped carbon matrix.

Transmission electron microscopy (TEM) images further confirm the presence of nanoparticles of about 25 nm size that are well-distributed on the carbon matrix (Fig. 1(b) and Fig. S6, ESI†). High-resolution TEM (HRTEM) images further show distinct lattice fringes with interlayer spacings of 0.294, 0.208, and 0.343 nm (Fig. 1(c)), which are assigned to the (220) and (400) crystal planes of cubic Fe₃O₄ and the (002) crystal plane of graphitized nitrogen-doped carbon matrix,³⁵ respectively. Besides, it could be observed in Fig. 1(c) and Fig. S6a (ESI†) that some Fe₃O₄ NPs were encapsulated by the nitrogen-doped carbon matrix, demonstrating nitrogen-doped carbon support with a good crystallinity. The selected-area electron-diffraction (SAED) pattern displays a ring-like pattern and the exposed planes of Fe₃O₄ NPs, further confirming the graphitized nature of the nitrogen-doped carbon matrix with a good crystallinity and the facet species of Fe₃O₄ NPs over Fe₃O₄@TNC (Fig. S7, ESI†). The high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image further confirms the mainly uniform monodispersed bright NPs (Fig. 1(d)). The energy-dispersive X-ray spectroscopy (EDS) elemental mapping verifies the elemental composition of Fe₃O₄@TNC with uniform distribution of C, N, O, and Fe elements. Particularly, the bright region of oxygen and iron elements is well overlapped (Fig. 1(d)), further confirming the well-dispersed Fe₃O₄ NPs anchored on the nitrogen-doped carbon (donated as Fe₃O₄@TNC). The crystalline structure of Fe₃O₄ NPs and nitrogen-doped carbon in Fe₃O₄@TNC was further confirmed by the X-ray diffraction (XRD) pattern (Fig. 1(e)). The diffraction peak at about 26.5° is attributed to the (002) plane of graphitic carbon, suggesting a nitrogen-doped carbon support with a good crystallinity, while other diffraction peaks of Fe₃O₄@TNC, including (220), (311), (440), and (400), correspond well to cubic phase Fe₃O₄ (card no. 01-076-1849).^34,36 Particularly, the (220) peak of Fe₃O₄ over Fe₃O₄@TNC with strong intensity relative to other peaks indicates the main facet over Fe₃O₄@TNC. Besides, Fe (110) in Fe₃O₄@TNC is also observed in Fig. 1(e) (card no. 06-0696),³⁴ which agrees with the Fe 2p XPS spectra (Fig. S4g, ESI†). Therefore, it could be demonstrated that the obtained Fe-based components in Fe₃O₄@TNC include Fe₃O₄ and metallic Fe components, with the Fe₃O₄ component dominating particularly.^34,37

To further confirm the exposed Fe-based components over Fe₃O₄@TNC, Fe₃O₄@TNC was etched by acid washing (denoted as Fe₃O₄@TNC-AE), the XRD pattern of Fe₃O₄@TNC-AE shows a remarkable enhancement of the intensity of Fe₃O₄ (311), while the intensity of Fe (110) and Fe₃O₄ (440) and the (400) peak is stronger than that of Fe₃O₄@TNC (Fig. S8, ESI†), indicating that these Fe-components were encapsulated by crystalline nitrogen-doped-carbon supports. More importantly, in contrast to Fe₃O₄@TNC, the intensity of the Fe₃O₄ (220) peak decreased obviously after acid etching (Fig. 1(e)), further demonstrating the (220) peak of Fe₃O₄, which mainly stays in the exposed state rather than the encapsulated situation relative to the other peaks. The I_D/I_G ratio in the Raman spectrum also increases from 1.10 to 1.22 after acid etching and is far lower than that of TNC (Fig. 1(f)), confirming that the carbon lattice fringes are etched.³⁸ This result suggests that the formation of Fe₃O₄ or Fe NPs over Fe₃O₄@TNC favors the formation of the crystalline nitrogen-doped-carbon. Besides, N 1s, O 1s, and Fe 2p HR-XPS of Fe₃O₄@TNC-AE also show peaks for Fe–N, Fe–O, and Fe of zero value, respectively (Fig. S8, ESI†). In particular, the content of O and Fe in Fe₃O₄@TNC-AE is lower than that in Fe₃O₄@TNC, suggesting that the crystalline carbon matrix was destroyed and simultaneously the encapsulated Fe-component was further exposed. Therefore, the above results suggest that Fe₃O₄ (220) dominates in the exposed plane of Fe₃O₄@TNC, where the Fe (110) and Fe₃O₄ (440), (400), (311) are mainly encapsulated by the carbon matrix, which is because this strategy for acid etching enables to etch the carbon matrix of the encapsulated Fe-based components.³⁹ After the acid etching of Fe₃O₄@TNC, the specific surface areas decreased sharply from 537.68 to 251.80 m² g⁻¹, and the distribution of the pore size also decreased from 22.18 to 6.12 nm (Fig. S9 and Table S3, ESI†). The content of Fe decreased significantly (Fig. S10 and Table S1, ESI†), further verifying that the carbon matrix and exposed Fe₃O₄ NPs in Fe₃O₄@TNC were etched, leading to the formation of a smaller porous structure. Moreover, it could be concluded that the encapsulated Fe-based components in Fe₃O₄@TNC enables to support the porous layer structure of the crytalline nitrogen-doped-carbon matrix, consequently realizing the mainly exposed (220) facet of Fe₃O₄ NPs with a well-dispersed.

To further reveal the chemical state and coordination environment for Fe-based components in Fe₃O₄@TNC, the Fe K-edge X-ray absorption near-edge structure (XANES) spectrum in Fig. 1(g) suggests that Fe₃O₄@TNC is different from those of Fe foil and iron phthalocyanine (FePc) but is very close to that of FeO, Fe₂O₃, and Fe₃O₄. The valence state (+2.61) of Fe in Fe₃O₄@TNC is higher than that of FePc (+2.44) and somewhat lower than that of Fe₃O₄ (+2.67) (Fig. 1(h)), suggesting that Fe₃O₄ dominates in Fe-based components of Fe₃O₄@TNC and electron transfer occurs from the coordination network to Fe sites in Fe₃O₄@TNC.⁴⁰ Moreover, k²-weighted Fourier transform (FT) EXFAS spectra demonstrate that Fe₃O₄@TNC displays two kinds of peaks at 1.52 and 2.72 Å (Fig. 1(i)), corresponding to the Fe–N and Fe–O scattering path and Fe–Fe scattering path, respectively, which is becuase the peak at 1.52 Å is close to the Fe–N path of FePc (1.58 Å) and the Fe–O path of Fe₂O₃ (1.54 Å) and Fe₃O₄ (1.61 Å) but far away from FeO (1.72 Å), and the peak of the Fe–Fe path at 2.72 Å is positioned between FeO (2.76 Å) and Fe₂O₃ (2.67 Å). The wavelet transforms of K-edge EXAFS (WT-EXAFS) in Fe₃O₄@TNC and the reference samples are shown in Fig. S11 (ESI†); the two intensities maximum of Fe₃O₄@TNC are attributed to the contribution of the backscattering of Fe–O–Fe, Fe–O, Fe–N, and Fe–Fe bonds, which agrees with the XPS spectra (Fig. S4 and Table S2, ESI†). Particularly, the backscattering center of Fe–O–Fe and Fe–O bonds enables them to nearly align to the two intensity maxima of the reference Fe₃O₄. This further suggests that Fe-based components in Fe₃O₄@TNC are mainly Fe₃O₄.⁴¹ Besides, the intensity of Fe₃O₄@TNC-AE in the K-edge energy position around 7116 eV of XANES spectra is higher than Fe₃O₄@TNC and the reference Fe₃O₄ and FePc (Fig. S12a, ESI†). FT-EXAFS of Fe₃O₄@TNC-AE exhibits a main intensity maximum, possibly ascribed to the backscattering of Fe–Fe bonds (Fig. S12b, ESI†). The contribution of the backscattering of Fe–O–Fe, Fe–O, and Fe–N bonds is also presented in Fe₃O₄@TNC-AE (Fig. S12c, ESI†), further confirming that Fe metal was exposed in Fe₃O₄@TNC-AE and also included partial Fe₃O₄, which agrees with the XRD pattern (Fig. S8, ESI†). Based on the above results, it could be concluded that the Fe-based components of Fe₃O₄@TNC include Fe₃O₄ NPs with the main exposed Fe₃O₄ (220) facet and the main encapsulated Fe₃O₄ (440), (400), and (311) facets, Fe NPs with the main encapsulated (110) facet, and these Fe-based components are anchored in crystalline nitrogen-doped-carbon support.

Electrocatalytic performance evaluations of the catalysts

The electrochemical ORR performance was first evaluated using a rotating ring-disk electrode (RRDE) at room temperature in an O₂-saturated neutral electrolyte of 0.1 M LiClO₄. All potentials were iR-compensated (80%) and converted to the reversible hydrogen electrode (RHE) scale. As shown in Fig. 2(a), Fe₃O₄@TNC achieved the largest disk current density (j_disk) of 4.69 mA cm⁻² and ring current of 0.48 mA (I_ring) at 0.3 V and a positive onset potential of 0.75 V compared with NC, Fe@Tpy, TNC, and Fe₃O₄@TNC-AE. The faradaic efficiency (FE) of Fe₃O₄@TNC for H₂O₂ selectivity was nearly 100% at the potential region from 0.60 to 0 V and also about 96% at 0.70 V (Fig. 2(b)). The electron transferred number (n) was calculated to be 2.05 at 0.60 V and 2.46 at 0.70 V (Fig. 2(c)). The rotating speeds of the RRDE have little influence on the H₂O₂ selectivity of Fe₃O₄@TNC (Fig. S13, ESI†). These results revealed Fe₃O₄@TNC shows superior 2e⁻-ORR performance in both activity and selectivity toward neutral H₂O₂ electrosynthesis. In comparison, Fe₃O₄@TNC-AE exhibits a wide fluctuation of electron transfer ranging between 2.5 and 3.69 (Fig. 2(c)), suggesting that the exposed Fe₃O₄ (440), (400), and (311) facets and Fe (110) over Fe₃O₄@TNC-AE are not conducive to 2e⁻-ORR. The structural difference between Fe₃O₄@TNC and Fe₃O₄@TNC-AE can reveal that the exposed Fe₃O₄ (220) in Fe₃O₄@TNC should be the main active site for H₂O₂ electrosynthesis via 2e⁻-ORR.

Fig. 2 (a) Polarization curves recorded at 1600 rpm with a scan rate of 5 mV s⁻¹ in an O₂-saturated electrolyte containing 0.1 M LiClO₄ and the ring electrode at a constant potential of 1.2 V, including disk current density (j_disk) and ring current (I_ring). (b) H₂O₂ faradaic efficiency. (c) Electron transfer number (n). (d) Tafel plots. (e) TOF values of Fe₃O₄@TNC and Fe₃O₄@TNC-AE. (f) Capacitance current densities measured at 1.0 V using 40 μF cm⁻² as the specific capacitance standard. (g) Mass activity for H₂O₂ production at 0.3 V. (h) Stability test of Fe₃O₄@TNC at a potential of 0.5 V. (i) Comparison of H₂O₂ selectivity of Fe₃O₄@TNC and previously reported catalysts (Table S5, ESI†) at 2 mA cm⁻² of j_disk.

To better understand the role of each component in Fe₃O₄@TNC in the process of neutral H₂O₂ electrosynthesis, the study explored the effect of different factors such as the type of Fe-based components, crystalline degree of graphitic carbon matrix, and nitrogen-doped contents and types in Fe₃O₄@TNC. These factors were adjusted by varying the pyrolysis temperature during the fabrication process. Under a low carbonization temperature, the as-prepared sample (Fe₃O₄@TNC-700) shows a strong intensity for the typical (002) peak of g-C₃N₄, but no Fe-based component-relevant diffraction peak appears⁴² (Fig. S14, ESI†), while showing a high content of N residual to form pyridinic-N, pyrrolic-N, graphitic-N and Fe–N bond (Fig. S8 and Table S1, ESI†). In addition, a low intensity of Fe⁰ and Fe–O peaks also appeared³⁴ (Fig. S15, ESI†). These structural changes result in the as-prepared sample with a low graphitic carbon degree (I_D/I_G, 1.56), low specific surface area (201.42 m² g⁻¹), and larger pore size of about 19.66 nm (Fig. S16, S17 and Table S3, ESI†), and also lead to an ultralow activity and selectivity for H₂O₂ electrosynthesis (Fig. S18, ESI†), further confirming that both the formation of Fe₃O₄ NPs and graphitic degree of the carbon matrix are critical for neutral H₂O₂ electrosynthesis performance. After the pyrolysis temperature increases, the as-prepared sample (Fe₃O₄@TNC-900) displays three peaks with slightly strong intensity, assigned to graphitic carbon (002), cubic phase Fe₃O₄ (311), and Fe (110),³⁶ (Fig. S14, ESI†). However, the cubic phase Fe₃O₄ (220) in Fe₃O₄@TNC-900 almost disappears, further demonstrating that a higher pyrolysis temperature is unfavorable for the formation of Fe₃O₄ (220). In addition, the Fe⁰ content sharply increases due to the reduction of Fe²⁺ under high temperatures (Fig. S15, ESI†), the crystalline degree of graphitic carbon increases due to the I_D/I_G ratio decreasing from 1.65 to 1.05 (Fig. S16, ESI†), and the specific surface area and pore size decreased to 478.31 m² g⁻¹ and 16.0 nm, respectively (Fig. S17 and Table S3, ESI†). This kind of structural change also causes low selectivity of H₂O₂ electrosynthesis and decrease of the ORR activity (Fig. S18, ESI†), further confirming that the Fe₃O₄ (311) and (400), and Fe (110) planes in Fe₃O₄@TNC are unfavorable for 2e⁻-ORR. Moreover, the different usage of g-C₃N₄ for the preparation of Fe₃O₄@TNC has a certain effect on the onset potential and activity of 2e⁻-ORR (Fig. S19, ESI†), demonstrating that the crystalline degree of nitrogen-doped-carbon also affects the 2e⁻-ORR performance over the exposed (220) facet of Fe₃O₄ NPs. Furthermore, it can be concluded that proper pyrolysis temperature and usage of g-C₃N₄ are beneficial for the formation of the mainly exposed Fe₃O₄ (220) and a proper crystalline nitrogen-doped-carbon.

The relationship between Fe-based components in Fe₃O₄@TNC and its superior performance was revealed by electrochemical evaluation. The Tafel slope of 147 mV dec⁻¹ for Fe₃O₄@TNC is smaller than that for Fe₃O₄@TNC-AE (210 mV dec⁻¹), TNC (167 mV dec⁻¹), Fe@Tpy (292 mV dec⁻¹), and NC (445 mV dec⁻¹) (Fig. 2(d)), demonstrating more favorable reaction kinetics of H₂O₂ electrosynthesis over the main exposed (220) facet of Fe₃O₄ NPs in Fe₃O₄@TNC.⁸ Moreover, the turnover frequency (TOF) values of Fe₃O₄@TNC are higher than those for Fe₃O₄@TNC-AE at potentials of 0.2, 0.4, and 0.6 V; particularly, a TOF value of 18.5 s⁻¹ was achieved for Fe₃O₄@TNC at 0.6 V, which is about 11-fold for Fe₃O₄@TNC-AE (Fig. 2(e)), further confirming the exposed Fe₃O₄ NPs (220) plane in Fe₃O₄@TNC showing a superior intrinsic performance of neutral H₂O₂ production per unit time, but the exposed Fe₃O₄ (311) plane in Fe₃O₄@TNC-AE shows less selectivity toward 2e⁻-ORR. Electrochemically active surface area (ECSA, Fig. 2(f), Fig. S20 and Table S4) with a double layer capacitance of 11.1 mF cm⁻² was observed for Fe₃O₄@TNC, which is also higher than that of TNC (4.6 mF cm⁻²) and Fe₃O₄@TNC-AE (6.7 mF cm⁻²), indicating superior electron transfer between the active sites of Fe₃O₄@TNC and electrolytes and more active sites available for oxygen activation.⁴³ As indicated in Fig. 2(g), Fe₃O₄@TNC exhibited a mass current density of 93.1 A g⁻¹ at 0.3 V under O₂-saturated electrolyte, which is approximately 1.9-fold, 7.6-fold, 133-fold, and 465.5-fold higher than that of Fe₃O₄@TNC-AE, TNC, Fe@Tpy, and NC, respectively, further indicating that the exposed Fe₃O₄ (220) plane in Fe₃O₄@TNC should be the active sites for 2e⁻-ORR. However, the mass current density of Fe₃O₄@TNC is about 12.6 A g⁻¹ at 0.3 V using air as the feeding gas (Fig. 2(g)), which is far lower 93.1 A g⁻¹, suggesting that the limited mass transportation of O₂ has a large influence on the activity of neutral H₂O₂ electrosynthesis. Thus, it can be further concluded that Fe₃O₄@TNC with a large specific surface area (537.68 m² g⁻¹) and proper pore width (22.18 nm) presents a huge potential for an instantaneously large yield rate of H₂O₂ under satisfactory O₂ supply.

The long-term stability of Fe₃O₄@TNC is revealed in Fig. 2(h). After continuous electrocatalysis for 12 hours, there is no obvious decay in the current density at 400 ppm of the rotating rate or static condition. Post-analysis of TEM images confirmed that obvious changes in Fe₃O₄@TNC before and after the long-term electrolysis (Fig. S21, ESI†). The potential of 0.55 V and H₂O₂ selectivity (nearly 100%) at the current density of 2 mA cm⁻² observed for Fe₃O₄@TNC are among the high values for all previously reported benchmarking H₂O₂ electrosynthesis catalysts under neutral electrolyte for ORR (Fig. 2(i) and Table S5, ESI†), further suggesting the main exposed Fe₃O₄ (220) plane over Fe₃O₄@TNC with a huge potential for realizing the desired neutral H₂O₂ electrosynthesis with low overpotential under large current density.

Theoretical calculations and in situ characterization

The above comparable experiments have demonstrated that the exposed Fe₃O₄ (220) anchored on a porous nitrogen-doped carbon matrix is critically important for the superior performance of neutral H₂O₂ electrosynthesis via 2e⁻-ORR. Density functional theory (DFT) calculations were performed to unveil the catalytic mechanisms of H₂O₂ synthesis. According to the structural characterization, both Fe-based components- and carbon matrix-relevant models in Fe₃O₄@TNC were constructed, including nitrogen-doped carbon with vacancy (CNV), Fe₃O₄- and Fe-based relevant configurations, and their heterojunction configurations (Fig. S22, ESI†). First, O₂ adsorption behavior was investigated on the surface Fe atom of the above models (Fig. S23, ESI†). Both CNV and Fe₃O₄(400) displayed a positive value, and the rest of the models all presented a negative value (Fig. 3(a)), suggesting that the adsorption of O₂ on these models is spontaneous, other than CNV and Fe₃O₄(400). Moreover, in comparison with Fe₃O₄-based facets, the structural heterojunction between CNV and various Fe₃O₄-based facets all present a different change trend for the adsorption of O₂. For instance, in contrast to −2.34 eV of Fe₃O₄(220) and −2.14 eV of Fe₃O₄(311), the heterojunctions of both Fe₃O₄(220)–CNV and Fe₃O₄(311)–CNV are at about −1.37 eV and −2.56 eV, respectively (Fig. 3(a)). In addition, compared with Fe(110) (−0.81 eV), Fe(110)–CNV has an O₂ adsorption energy of −2.16 eV. These results suggest that there is a strong interaction between the carbon-based matrix and Fe₃O₄, which affects the adsorption strength of O₂, thereby affecting the adsorption of *OOH.³³Fig. 3(b) shows the free energy profile of H₂O₂ formation at an equilibrium potential of 0.7 V on the above models. DFT predicts that the formation of *OOH on Fe₃O₄(220)–CNV is an endothermic process and has the lowest barrier energy of 0.07 eV compared to CNV, Fe(110), Fe(110)–CNV, Fe₃O₄(400), Fe₃O₄(400)–CNV, Fe₃O₄(220), Fe₃O₄(311), and Fe₃O₄(311)–CNV, suggesting that the exposed Fe₃O₄(220) in Fe₃O₄@TNC is the most favorable for 2e⁻-ORR for H₂O₂ electrosynthesis. The ORR pathway is further investigated by the tendency of O–O bond dissociation (ΔG_*OOH). Fig. 3(c) shows a volcano plot for oxygen reduction via the 2e⁻ pathway as a function of ΔG_*OOH. Compared with 3.83 eV of pure Fe₃O₄(220), the free energy of *OOH (ΔG_*OOH) on the Fe atom of Fe₃O₄(220)–CNV is 4.19 eV, closer to the ideal value of 4.22 eV. However, Fe₃O₄(311)–CNV has ΔG_*OOH of 2.25 eV compared to 3.48 eV for Fe₃O₄(311), further confirming a strong interaction at the inside of the heterojunction between CNV and Fe₃O₄ NPs. Besides, both Fe₃O₄(400)–CNV and Fe(110)–CNV of ΔG_*OOH are 4.55 and 3.63 eV, respectively, which is relatively closer to 4.22 eV. The charge density difference in Fe₃O₄(311)–CNV and Fe₃O₄(220)–CNV for the adsorption of O₂ and the OOH intermediate further reveal the effect of the heterojunction between CNV and Fe₃O₄ NPs (Fig. 3(d)). When O₂ and OOH intermediate adsorb on the Fe atom of Fe₃O₄(220)–CNV, the adsorbed O₂ and OOH achieve 0.69 e⁻ and 0.54 e⁻ electrons, respectively, but the adsorbed O₂ and OOH on the Fe atom of Fe₃O₄(311)–CNV accumulate the electrons of 0.73 e⁻ and 0.56 e⁻, respectively. Therefore, the nitrogen-doped carbon matrix as support could modify the surface electron density of the exposed Fe₃O₄ facet, thereby adjusting the number of electron transfers from the Fe atom in Fe₃O₄ to the adsorbed OOH with a desired binding strength.

Fig. 3 (a) Adsorption energy of O₂, (b) free energy diagram at U = 0.7 V for the reduction of O₂ to H₂O₂, (c) computed activity volcano plots on the possible relevant crystalline facet of Fe₃O₄@TNC. (d) Charge density difference for both O₂ and *OOH adsorbed on Fe₃O₄ (220)–CNV and Fe₃O₄ (311)–CNV. (e) In situ ATR-FTIR spectra collected on Fe₃O₄@TNC catalyst in O₂-saturated 0.1 M LiClO₄ electrolyte at different potentials from OCP, 1.0 to −0.3 V. (f) In situ electrochemical Raman spectra of the ORR process on Fe₃O₄@TNC at various potentials.

Combining the structural characterization of Fe₃O₄@TNC and Fe₃O₄@TNC-AE with their electrocatalytic performance, although both contain various Fe-based components, such as Fe(110), Fe₃O₄(400), Fe₃O₄(220), and Fe₃O₄(311), the exposed Fe-based components in Fe₃O₄@TNC should be the Fe₃O₄(220) facet with an excellent performance of 2e⁻-ORR, while the mainly exposed Fe₃O₄(311) in Fe₃O₄@TNC-AE exhibited a low selectivity for H₂O₂ electrosynthesis (Fig. 2(b) and Fig. S8, ESI†). Therefore, the agreement between both theoretical calculations and experimental observations suggests that the exposed Fe₃O₄ facet in Fe₃O₄@TNC should be the (220) facet and the other Fe-based NPs should be mainly encapsulated in the carbon matrix. In situ attenuated total reflectance Fourier-transform infrared (ATR-FTIR) spectroscopy was used to monitor the interactions between oxygen intermediates and Fe₃O₄@TNC in O₂-saturated 0.1 M LiClO₄ electrolyte (Fig. S24a and b, ESI†). As illustrated in Fig. 3(e), the intensity of the peak at 1054 cm⁻¹ presents a gradually increasing trend as the potential, associated with the adsorbed ClO₄⁻ in the electrolyte.⁸ Moreover, the bands at 1450 cm⁻¹, typically assigned to the adsorbed *O₂ species,⁴⁴ present an obvious intensity from the potential of 0.4 to −0.1 V, indicating the hydrogenation of the adsorbed *O₂ with a fast kinetics. Besides, the adsorption bands at about 890, 1240, and 1396 cm⁻¹ appear on applying a potential of 0.8 V are assigned to the M–O and O–O stretching vibration of the adsorbed *OOH and the adsorbed *HOOH species,^13,38 respectively. It is observed that the peak intensity of both adsorbed *OOH and *HOOH species is close, further demonstrating this hydrogenation step from *OOH to *HOOH with a rapid kinetics process.

Furthermore, the bands of the adsorbed *OOH and *HOOH species are enhanced from the potential of 0.8 to 0.4 V and slightly rise by further decreasing the potential during neutral H₂O₂ electrosynthesis, attributed to the limited O₂ supply, which follows the trend for the potential at about 0.4 V of the limited current in polarization curve of Fe₃O₄@TNC (Fig. 2(a)). Overall, these results demonstrate the rapid kinetic process of O₂ hydrogenation via *OOH-mediated two-electron ORR pathway on the Fe₃O₄@TNC catalyst, which agrees with the thermodynamically spontaneous process from the DFT results (Fig. 3(b)). The strong electromagnetic field of Fe₃O₄ NPs in Fe₃O₄@TNC could effectively enhance the Raman signal of the species on the exposed Fe₃O₄ NP surface, which allows us to detect trace intermediates during the ORR process. In situ electrochemical Raman measurements were performed in 0.1 M LiClO₄ electrolyte using a customized cell (Fig. S24c and d, ESI†). Both D-band and G-band Raman scattering of Fe₃O₄@TNC were observed at about 1350 and 1640 cm⁻¹ under open circuit potential (OCP) or N₂-saturated electrolyte condition, respectively, which agreed with the Raman spectrum of Fe₃O₄@TNC (Fig. 1(f)). During ORR electrocatalysis, as observed in Fig. 3(f), the intensity of the broad Raman band at about 1093 cm⁻¹ is slightly enhanced as the applied potential increases from the onset potential of 0.7 V for ORR, corresponding to the absorbed O–O stretching vibration of superoxide species (*O₂⁻).⁴⁵ Moreover, a new band at about 1528 cm⁻¹ appears at a higher potential than 0.7 V, assigned to the absorbed *OOH species,²¹ and the intensity of *OOH slightly increases as the potential rises. Besides, the intensity of *O₂⁻ and *OOH becomes obvious under a sufficient potential due to the fast consumption rate, further suggesting the neutral 2e⁻-ORR on Fe₃O₄@TNC via the *OOH-mediated pathway with a fast kinetic process.

Scalable neutral production of H₂O₂ electrosynthesis

To evaluate the potential for scalable neutral H₂O₂ electrosynthesis, Fe₃O₄@TNC was first investigated at a large potential range from 0.75 to −0.7 V in O₂-saturated neutral electrolyte of 0.1 M LiClO₄ using RRDE (Fig. S25, ESI†). Fe₃O₄@TNC with the mainly exposed Fe₃O₄ (220) plane displays a large rising trend for current density in the potential range from 0.75 to −0.7 V, with nearly 100% selectivity for H₂O₂ production, further confirming its great application potential for industrially relevant yield rate of neutral H₂O₂ production under satisfactory O₂ supply and large current density. In addition, Fe₃O₄@TNC was deposited on a hydrophobic gas-diffusion electrode as a cathode for enhancing O₂ supply and Pt gauze as anode. As illustrated in Fig. 4(a), the two-electrode system was assembled in a solid-state electrolyte (SE) cell with a sandwiched double proton exchange membranes (PEM) configuration containing a porous polymer ion conductor in the SE layer to eliminate the iR-drop between the cathode and anode.^12,14,46 The optimized conditions, including the mass loading of the catalyst and water flow rate for the selectivity of H₂O₂ production, are summarized in Fig. S26 (ESI†) under enough O₂ supply with a flow rate of 100 sccm. Increasing the water flow rate within 5–40 mL min⁻¹ is favorable for bringing out the generated H₂O₂ and minimizing its further electroreduction (Fig. S27, ESI†), particularly, the current density of SE cells up to 60 mA cm⁻². A higher mass loading of Fe₃O₄@TNC requires a larger flow rate of DI water, which obtains a relatively high H₂O₂ FE of at least 40 mL min⁻¹ due to the high resistance of O₂ gas and generated H₂O₂ mass transport.¹⁵ Therefore, the optimal mass loading of Fe₃O₄@TNC and DI water flow were respectively determined to be 0.1 mg cm⁻² and 30 mL min⁻¹, which ensure not only the proton resources supply but also efficiently brings out the produced H₂O₂. The I–V curve of the 4-cm² three-chamber double-layer-PEM SE cell is plotted in Fig. 4(b); as the overall current density increased from 2.5 to 60 mA cm⁻², the cell voltage of the SE reactor gradually increased from 1.9 to 2.3 V, and the trend for FE of H₂O₂ decreased gradually from 98% to 83% at the cathode side (Fig. 4(b)). Besides, it can be observed in Fig. 4(b) that a small amount of partially generated H₂O₂ could get through the PEM into the middle layer, corresponding to about 1–4% H₂O₂ FE. The maximum total H₂O₂ FE in the SE cell is nearly 100% in the current density range from 2.5 to 20 mA cm⁻² (Fig. 4(c)). Subsequently, H₂O₂ FE decreased to 87% under a high current density of 60 mA cm⁻², suggesting that a small amount of the partially generated H₂O₂ is further reduced to H₂O at the cathode side. Even then, the yield rate of H₂O₂ increased sharply as the applied current density increased. Particularly, 8.57 and 8.89 mol h⁻¹ g_cat.⁻¹ were obtained at a current density of 50 and 60 mA cm⁻², respectively.

Fig. 4 (a) Schematic of H₂O₂ electrosynthesis via ORR in the SE cell with double-PEM configuration. (b) The I–V curve and corresponding FEs for generating H₂O₂ in the middle chamber and cathode, (c) the corresponding total yield rate of H₂O₂ and FEs. (d) Chronopotentiometry stability test by directly flowing 1 M Na₂SO₄ solution in the middle chamber at 50 mA cm⁻² current density. (e) Chronopotentiometry test using 300 mA cm⁻² for H₂O₂ production by varying H₂O flow rate from 1 to 12 mL min⁻¹. (f) Techno-economic evaluation of neutral H₂O₂ electrosynthesis with two different conditions in the SE cell.

The stability of H₂O₂ electrosynthesis was further investigated by holding a cell current density of 50 mA cm⁻² (200 mA total current). As shown in Fig. 4(d), the continuous generation of H₂O₂ solution on the cathode side can stably run for over 60 h without decline with about 91% of H₂O₂ FE, suggesting Fe₃O₄@TNC with an excellent stability for H₂O₂ electrosynthesis under scalable H₂O₂ production. The structural and morphological stability of Fe₃O₄@TNC after 60 h of testing is reflected in the XRD and TEM images (Fig. S28, S29 and Table S6, ESI†), with minimal change in the (002) crystal plane. To further investigate the scalable H₂O₂ production of SE cell using Fe₃O₄@TNC as a catalyst, air atmosphere was used to replace pure O₂ gas and the optimized operating condition of SE cell was continuously used. ORR activity is observed under low voltage of SE cell due to the close current density (Fig. S30a, ESI†), and H₂O₂ FE is nearly 100% within the current density of 2.5–20 mA cm⁻². Subsequently, the H₂O₂ FE sharply decreases from 30 to 50 mA cm⁻² as the current density of SE cell increases (Fig. S30b, ESI†). This result is mainly due to the limited O₂ supply responding to the diffusion-limited current density (Fig. S31, ESI†). Moreover, continuous H₂O₂ generation at the cathode side occurred stably for 40 h, maintaining about 90% selectivity of H₂O₂, and the voltage of the SE cell increases slightly (Fig. S30c, ESI†), further confirming engaging Fe₃O₄@TNC into the SE cell showing a huge application potential for real-time continuous H₂O₂ production with air and pure water. Furthermore, to explore the industrially relevant and economical potential for neutral H₂O₂ electrosynthesis over Fe₃O₄@TNC in the SE cell, different flow rates of DI water were adopted to regulate the real-time H₂O₂ concentration produced from the cathode under a current density of 300 mA cm⁻². As displayed in Fig. 4(e), as the flow rate of DI water decreases from 12 to 1 mL min⁻¹, the concentration of real-time generated H₂O₂ at the cathode side presents a rising trend from 5.01 to 19.23 mM, showing a large region of H₂O₂ concentration, indicating a real-time H₂O₂ supply potential through the SE cell with the Fe₃O₄@TNC catalyst. Moreover, techno-economic evaluation, derived from SE electrolyzer results under current densities of 50 and 300 mA cm⁻², estimated the plant-gate levelized costs for H₂O₂ production using Fe₃O₄@TNC at a market price of about 73% (US$586) and 40% (US$316) (Fig. 4(f) and Fig. S32, S33, ESI†), respectively, compared with the market price of about US$800 tonne⁻¹ for electron-grade 30% H₂O₂ (ref. 9,10). Moreover, the industrially relevant current density of 300 mA cm⁻² for neutral H₂O₂ electrosynthesis achieves larger room for benefits with respect to 50 mA cm⁻² (Fig. S34, ESI†), demonstrating that this process of H₂O₂ electrosynthesis via the integration of efficient electrocatalysts into the SE cell presents a huge potential for economical and scalable H₂O₂ production in pure water.

Conclusion

In summary, a Fe₃O₄@TNC catalyst was developed by pyrolyzing the mixture of both g-C₃N₄ and Fe@Tpy, which demonstrated a scalable and economical rate of H₂O₂ production in the SE cell. This strategy employed crystalline nitrogen-doped-carbon supported exposed and encapsulated Fe₃O₄ NPs as the electrocatalysts. We first found that the exposed Fe₃O₄ NPs with the (220) facet on TNC is the most active site, which facilitates the thermodynamically lowest barrier energy process for 2e⁻-ORR via *OOH mediated with rapid kinetics for neutral H₂O₂ electrosynthesis due to the interaction between the exposed (220) facet of Fe₃O₄ NPs and crystalline nitrogen-doped-carbon. Under the RRDE test conditions, the Fe₃O₄@TNC catalyst exhibits a high onset potential of 0.75 V, a nearly 100% selectivity toward H₂O₂ production at a broad potential range from 0.75 to −0.6 V, and a current density of 2 mA cm⁻² at 0.5 V vs. RHE in the neutral media, which is superior to that of the previously reported electrocatalysts. The scalable current densities up to 60 mA cm⁻² (240 mA) with 87% FE were obtained for continuous H₂O₂ production at a record rate of 8.89 mol h⁻¹ g_cat.⁻¹ in the SE cell. Particularly, a real-time continuous 19.23 mM of H₂O₂ production was achieved at a current density of 300 mA cm⁻², showing a huge techno-economic potential for industrially relevant H₂O₂ production in pure water.

Experimental section

Synthesis of Tpy

A mixture of terephthalaldehyde (5.0 g, 37.3 mmol), 2-acetylpyridine (21.67 g, 179.1 mmol), anhydrous ethanol (300 mL), and powdered NaOH (8.95 g, 223.8 mmol) was added to a 1000 mL round-bottom flask. The mixture was stirred at room temperature for 12 h, 200 mL of NH₃ H₂O (25–28%) was added and heated at 85 °C for 48 h to get a brown suspension solution. After being cooled to room temperature, the brown precipitate was filtered and then recrystallized twice under reflux with a mixed solvent of CHCl₃/CH₃OH (v/v, 2/1) to obtain the target compound (1.4 g, 1.96 mmol) as a grey solid (donated as Tpy) in 35% yield.

Synthesis of Fe@Tpy

Typically, the obtained Tpy (1.85 mmol) was dissolved in 100 mL CH₂Cl₂ at 85 °C (1.85 × 10⁻² mol L⁻¹), and FeCl₂·4H₂O was used as the metal source to form a metal ion methanol solution (1.85 × 10⁻² mol L⁻¹). The metal ions with 1 : 1 ratio of mass of Tpy were easily dropped into the Tpy solution, reacted at 85 °C for 16 h, and the target product was separated at the end of the reaction, labeled as Fe@Tpy.

Synthesis of TNC, Fe₃O₄@TNC, and Fe₃O₄@TNC-AE

The synthesis of g-C₃N₄ was through the thermal polymerization of melamine in a tube furnace at 550 °C for 3 h, with a heating rate of 5 °C min⁻¹ under an Ar atmosphere. The mixture of Fe@Tpy and g-C₃N₄ was carbonized under different temperatures and Ar atmosphere (700, 800, and 900 °C) using a heating rate of 5 °C min⁻¹ for 2 h. The final products were obtained after cooling and denoted as Fe₃O₄@TNC-700, Fe₃O₄@TNC-800, and Fe₃O₄@TNC-900. Additionally, pyrolyzing the mixture for the different ratios of g-C₃N₄ to Fe@Tpy, a series of powders was prepared after the pyrolysis of these mixtures at 800 °C for 2 h using a heating rate of 5 °C min⁻¹, which were labeled as Fe₃O₄@TNC-20, Fe₃O₄@TNC-40, Fe₃O₄@TNC-60, and Fe₃O₄@TNC-80. Generally, TNC is prepared by directly pyrolyzing the mixture of Tpy and g-C₃N₄ under the same preparation conditions of Fe₃O₄@TNC-800, particularly, the usage ratio between Tpy and g-C₃N₄, and the pyrolysis temperature. Typically, Fe₃O₄@TNC-AE was achieved by the etching of Fe₃O₄@TNC-800. First, Fe₃O₄@TNC was dispersed into a 2 M HNO₃ solution and sonicated for 1 h. The resultant ink solution was further stirred at room temperature for 24 hours and then washed in ultrapure water. The final product was obtained after drying, known as Fe₃O₄@TNC-AE.

Computational details

Spin-polarized DFT calculations were performed using the Vienna ab initio simulation package.⁴⁷ The generalized gradient approximation toward the Perdew–Burke–Ernzerhof level was adopted in the projector-augmented wave potentials.⁴⁸ The Monkhorst–Pack type of k-point sample was shifted to the gamma center to adapt for all the structures in this work. The Gaussian smearing of 0.05 eV was adopted to satisfy the self-consistence and convergence of the electron density. All the atomic positions were allowed to relax until the forces were less than 0.005 eV Å⁻¹, and the electron convergence energy was set to 10⁻⁵ eV. The vacuum was set to 15 Å for all the interfacial structures to avoid interactions between its periodic images. The plane wave cutoff was set to 520 eV. The DFT-D3 scheme was adopted to correct the van der Waals interaction.⁴⁹ Here, we used the computational hydrogen electrode (CHE) model proposed by Nørskov et al. to calculate the free energy levels of all intermediates:^50,51 ΔG_ads = ΔE_ads + ΔZPE − TΔS + eU, where ΔE_ads is the binding energy of adsorption species *O₂ and *OOH. ΔZPE, ΔS, U, and e⁻ are the ZPE changes, entropy changes applied potential at the electrode, and charge transferred, respectively. The contributions of each component for ΔG_ads were obtained from literature.⁵¹ As the ground state of the O₂ molecule is poorly described in DFT calculations, we used gas-phase H₂O and H₂ as reference states as they are readily treated in the DFT calculations. In our simulations, solvation effects were not considered.

Neutral H₂O₂ production in a solid-state electrolyte cell with double-PEM configuration

A solid electrolyte (SE) cell with a sandwiched double-PEM design was used for continuous H₂O₂ electrosynthesis. The cell structures and production setup are shown in Fig. 4(a). The cathode was supplied with a mixture of O₂ gas (100 sccm) and DI water (30 mL min⁻¹), with the gas flow regulated by a mass flow controller (MFC) and the water flow controlled by a peristaltic pump. The cathode and anode were positioned on opposite sides of 0.5 cm thick PEEK plates, each containing 2 cm × 2 cm channels. The catalyst layers faced the flowing liquid electrolyte, providing a geometric surface area of 4 cm⁻². A measuring cylinder calibrated the flow rate of H₂O₂ produced at the outlet. The rapid flow of water in the gas/liquid mixture via the cathode chamber is good for removing the produced H₂O₂ molecules and reducing additional H₂O₂ electroreduction. The SE in the middle chamber consisted of Dowex 50W X8 hydrogen form (Sigma-Aldrich), a styrene–divinylbenzene sulfonated copolymer cation conductor. A syringe pump was used to introduce a solution containing H₂SO₄ and/or Na₂SO₄ into the SE layer. At the anode side, a 0.1 M H₂SO₄ solution was circulated at 3 mL min⁻¹, while a 1 M Na₂SO₄ solution flowed through the SE layer at the same rate in a single pass. Potentials were measured using a two-electrode setup and manually adjusted for 100% accuracy.

Solid-state electrolyte cell

The cathode side was continuously supplied with a mixture of O₂ gas and water flow for 2e⁻-ORR, while the anode side was circulated with 0.1 M H₂SO₄ solution for water oxidation and proton supply. In the middle chamber, 1 M Na₂SO₄ solution flowed through in the SE layer to introduce the cation effects, maintaining ion conduction and a neutral environment at the catalyst/PEM interface, thus promoting H₂O₂ generation. The H₂O₂ molecules produced at the cathode were efficiently carried away by the O₂ and DI water flow. Meanwhile, the generated protons from water oxidation at the anode moved into the SE layer to compensate for the charge.

Data availability

The data supporting this article have been included as part of the ESI.†

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was financially supported by National Natural Science Foundation of China (22466010), Guizhou Provincial Basic Research Program (Natural Science) ZK[2023]47 and Key Program ZK[2025]075, Innovation and Entrepreneurship Project for overseas Talents in Guizhou Province [2022]02, Specific Natural Science Foundation of Guizhou University (X202207), and Science and Technology Department of Guizhou Province (grant no. Platform & Talents [2019]5607). The authors would like to thank Shiyanjia Lab (https://www.shiyanjia.com) for the XPS and TEM analysis. The authors also would like to SCI-GO (https://www.sci-go.com) for the HRTEM analysis and the computing support of Chengdu DiYiYuanLi Technology Co., Ltd and the State Key Laboratory of Public Big Data at Guizhou University.

Notes and references

1. S. C. Perry, D. Pangotra, L. Vieira, L.-I. Csepei, V. Sieber, L. Wang, C. Ponce de León and F. C. Walsh, Nat. Rev. Chem., 2019, 3, 442–458.
2. R. Hage and A. Lienke, Angew. Chem., Int. Ed., 2006, 45, 206–222.
3. B. S. Lane and K. Burgess, Chem. Rev., 2003, 103, 2457–2473.
4. S. Gao, Y. Jin, K. Ge, Z. Li, H. Liu, X. Dai, Y. Zhang, S. Chen, X. Liang and J. Zhang, Adv. Sci., 2019, 6, 1902137.
5. J. Ma, N. A. Choudhury and Y. Sahai, Renewable Sustainable Energy Rev., 2010, 14, 183–199.
6. J. M. Campos-Martin, G. Blanco-Brieva and J. L. Fierro, Angew. Chem., Int. Ed., 2006, 45, 6962–6984.
7. S. Siahrostami, A. Verdaguer-Casadevall, M. Karamad, D. Deiana, P. Malacrida, B. Wickman, M. Escudero-Escribano, E. A. Paoli, R. Frydendal, T. W. Hansen, I. Chorkendorff, I. E. L. Stephens and J. Rossmeisl, Nat. Mater., 2013, 12, 1137–1143.
8. J. Gao, H. bin Yang, X. Huang, S.-F. Hung, W. Cai, C. Jia, S. Miao, H. M. Chen, X. Yang and Y. Huang, Chem, 2020, 6, 658–674.
9. B.-H. Lee, H. Shin, A. S. Rasouli, H. Choubisa, P. Ou, R. Dorakhan, I. Grigioni, G. Lee, E. Shirzadi and R. K. Miao, Nat. Catal., 2023, 6, 234–243.
10. J. Qi, Y. Du, Q. Yang, N. Jiang, J. Li, Y. Ma, Y. Ma, X. Zhao and J. Qiu, Nat. Commun., 2023, 14, 6263.
11. T. H. Jeon, B. Kim, C. Kim, C. Xia, H. Wang, P. J. J. Alvarez and W. Choi, Energy Environ. Sci., 2021, 14, 3110–3119.
12. C. Xia, Y. Xia, P. Zhu, L. Fan and H. Wang, Science, 2019, 366, 226–231.
13. P. Cao, X. Quan, X. Nie, K. Zhao, Y. Liu, S. Chen, H. Yu and J. G. Chen, Nat. Commun., 2023, 14, 172.
14. X. Zhang, X. Zhao, P. Zhu, Z. Adler, Z.-Y. Wu, Y. Liu and H. Wang, Nat. Commun., 2022, 13, 2880.
15. H. Li, P. Wen, D. S. Itanze, Z. D. Hood, S. Adhikari, C. Lu, X. Ma, C. Dun, L. Jiang and D. L. Carroll, Nat. Commun., 2020, 11, 3928.
16. R. D. Ross, H. Sheng, Y. Ding, A. N. Janes, D. Feng, J. Schmidt, C. U. Segre and S. Jin, J. Am. Chem. Soc., 2022, 144, 15845–15854.
17. S. Siahrostami, S. J. Villegas, A. H. Bagherzadeh Mostaghimi, S. Back, A. B. Farimani, H. Wang, K. A. Persson and J. Montoya, ACS Catal., 2020, 10, 7495–7511.
18. S. Ding, B. Xia, M. Li, F. Lou, C. Cheng, T. Gao, Y. Zhang, K. Yang, L. Jiang, Z. Nie, H. Guan, J. Duan and S. Chen, Energy Environ. Sci., 2023, 16, 3363–3372.
19. Y. Tian, D. Deng, L. Xu, M. Li, H. Chen, Z. Wu and S. Zhang, Nano-Micro Lett., 2023, 15, 122.
20. X. Huang, M. Song, J. Zhang, T. Shen, G. Luo and D. Wang, Nano-Micro Lett., 2023, 15, 86.
21. Q. Hu, K. Gao, X. Wang, H. Zheng, J. Cao, L. Mi, Q. Huo, H. Yang, J. Liu and C. He, Nat. Commun., 2022, 13, 3958.
22. Y. Zhao, H. Wang, J. Li, Y. Fang, Y. Kang, T. Zhao and C. Zhao, Adv. Funct. Mater., 2023, 33, 2305268.
23. L. Jing, W. Wang, Q. Tian, Y. Kong, X. Ye, H. Yang, Q. Hu and C. He, Angew. Chem., Int. Ed., 2024, 63, e202403023.
24. N. Wang, Y.-S. Zhang, D.-D. Wei, H.-M. Duan, L.-M. Mo and H.-Y. Wang, J. Mater. Chem. A, 2023, 11, 4162–4169.
25. X. Sun, J. Yang, X. Zeng, L. Guo, C. Bie, Z. Wang, K. Sun, A. K. Sahu, M. Tebyetekerwa, T. E. Rufford and X. Zhang, Angew. Chem., Int. Ed., 2024, e202414417.
26. M. Yan, Z. Wei, Z. Gong, B. Johannessen, G. Ye, G. He, J. Liu, S. Zhao, C. Cui and H. Fei, Nat. Commun., 2023, 14, 368.
27. L. Liu, L. Kang, J. Feng, D. G. Hopkinson, C. S. Allen, Y. Tan, H. Gu, I. Mikulska, V. Celorrio and D. Gianolio, Nat. Commun., 2024, 15, 4079.
28. S. Li, L. Chen, J. Wang, T. Liu, D. Li, Z. Yang, X. Xiao, C. Chu and B. Chen, Environ. Sci. Technol., 2024, 58, 10072–10083.
29. P. Guo, B. Liu, F. Tu, Y. Dai, Z. Zhang, Y. Xia, M. Ma, Y. Zhang, L. Zhao and Z. Wang, Energy Environ. Sci., 2024, 17, 3077–3087.
30. X.-K. Gu, J. S. A. Carneiro, S. Samira, A. Das, N. M. Ariyasingha and E. Nikolla, J. Am. Chem. Soc., 2018, 140, 8128–8137.
31. J. Yuan, X. Liu, Y. Tang, Y. Zeng, L. Wang, S. Zhang, T. Cai, Y. Liu, S. Luo and Y. Pei, Appl. Catal., B, 2018, 237, 24–31.
32. M. Eguchi, M. Momotake, F. Inoue, T. Oshima, K. Maeda and M. Higuchi, ACS Appl. Mater. Interfaces, 2017, 9, 35498–35503.
33. X. Li, C. Li, Y. Xu, Q. Liu, M. Bahri, L. Zhang, N. D. Browning, A. J. Cowan and J. Tang, Nat. Energy, 2023, 8, 1013–1022.
34. J. Geng, S. Ji, M. Jin, C. Zhang, M. Xu, G. Wang, C. Liang and H. Zhang, Angew. Chem., Int. Ed., 2023, 62, e202210958.
35. C. Santoro, P. Bollella, B. Erable, P. Atanassov and D. Pant, Nat. Catal., 2022, 5, 473–484.
36. X. Yang, Y. Tian, S. Mukherjee, K. Li, X. Chen, J. Lv, S. Liang, L. K. Yan, G. Wu and H. Y. Zang, Angew. Chem., Int. Ed., 2023, 62, e202304797.
37. G. Ye, S. Liu, K. Huang, S. Wang, K. Zhao, W. Zhu, Y. Su, J. Wang and Z. He, Adv. Funct. Mater., 2022, 32, 2111396.
38. S. Chen, T. Luo, X. Li, K. Chen, J. Fu, K. Liu, C. Cai, Q. Wang, H. Li and Y. Chen, J. Am. Chem. Soc., 2022, 144, 14505–14516.
39. Y. Wang, Y. Zhou, Y. Feng and X. Y. Yu, Adv. Funct. Mater., 2022, 32, 2110734.
40. X. Lei, Q. Tang, Y. Zheng, P. Kidkhunthod, X. Zhou, B. Ji and Y. Tang, Nat. Sustain., 2023, 6, 816–826.
41. G. Yang, J. Zhu, P. Yuan, Y. Hu, G. Qu, B.-A. Lu, X. Xue, H. Yin, W. Cheng and J. Cheng, Nat. Commun., 2021, 12, 1734.
42. R. D. Ross, H. Sheng, A. Parihar, J. Huang and S. Jin, ACS Catal., 2021, 11, 12643–12650.
43. Y. Li, J. Chen, Y. Ji, Z. Zhao, W. Cui, X. Sang, Y. Cheng, B. Yang, Z. Li, Q. Zhang, L. Lei, Z. Wen, L. Dai and Y. Hou, Angew. Chem., Int. Ed., 2023, 62, e202306491.
44. S. Nayak, I. J. McPherson and K. A. Vincent, Angew. Chem., Int. Ed., 2018, 130, 13037–13040.
45. J. Wei, D. Xia, Y. Wei, X. Zhu, J. Li and L. Gan, ACS Catal., 2022, 12, 7811–7820.
46. C. Xia, P. Zhu, Q. Jiang, Y. Pan, W. Liang, E. Stavitski, H. N. Alshareef and H. Wang, Nat. Energy, 2019, 4, 776–785.
47. G. Kresse and J. Furthmüller, Phys. Rev. B: Condens. Matter Mater. Phys., 1996, 54, 11169–11186.
48. J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1996, 77, 3865–3868.
49. S. Ehrlich, J. Moellmann, W. Reckien, T. Bredow and S. Grimme, ChemPhysChem, 2011, 12, 3414–3420.
50. I. C. Man, H.-Y. Su, F. Calle-Vallejo, H. A. Hansen, J. I. Martínez, N. G. Inoglu, J. Kitchin, T. F. Jaramillo, J. K. Nørskov and J. Rossmeisl, ChemCatChem, 2011, 3, 1159–1165.
51. J. K. Nørskov, J. Rossmeisl, A. Logadottir, L. Lindqvist, J. R. Kitchin, T. Bligaard and H. Jónsson, J. Phys. Chem. B, 2004, 108, 17886–17892.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4ee05796a

---