First, natural balsa wood was cut into thin slices (2.5 cm × 2.5 cm × 0.15 cm for the tangential, radial, and longitudinal directions, respectively). The thin slices were subjected to ultrasonic treatment for 30 min, followed by drying in an oven at 40 °C. The dried wood slices were then immersed in 300 mL sodium chlorite-acetate solution with a concentration of 2 wt% and pH of 4.6, and reacted in an oven at 80 °C for 2 h. Afterward, the wood slices were washed three times with ultra-pure water and subjected to ultrasonic treatment, then dried again in the convection oven at 40 °C to obtain pretreated wood (PW).

Initially, the obtained PW was immersed in a solution containing 60 mg of ferric chloride tetrahydrate that dissolved in 35 mL deionized water, and heated at 80 °C for 12 h, with the wood slices removed every 2 h for vacuum drying. After 12 h, the wood slices were freeze-dried, yielding PW loaded with ferric chloride, referred to as FeCl₃-PW. It was then placed in a tube furnace, where 0.5 g boric acid and 3 g melamine were added into a ceramic boat at the upstream. The temperature was increased to 700, 800, and 900 °C at a rate of 5 °C·min^-1 and held for 2 h under Ar atmosphere. After cooling to room temperature, the catalyst was placed in 1M HCl and reacted at 80 °C for 2 h. It was then rinsed with deionized water until neutral and vacuum-dried to obtain Fe₃C@BNC-700, Fe₃C@BNC-800, and Fe₃C@BNC, respectively. The preparation of Fe₃C@NC was performed by omitting the addition of boric acid during the calcination in the tube furnace, while keeping all other processes unchanged. The preparation of U-Fe₃C@BNC involved using natural balsa wood that did not undergo lignin removal, with all other processes remaining the same.

PW was directly placed in a tube furnace, with 0.5 g boric acid and 3 g melamine into a ceramic boat at the upstream. The furnace was heated at a rate of 5 °C·min^-1 to 900 °C and maintained for 2 h to obtain the B, N-doped porous carbon (BNC).

More detailed information, including materials characterization and electrochemical measurements, is presented in the Supplementary Materials.

Figure 1A illustrates the preparation process of the Fe₃C@BNC catalyst and the optical images of each stage [Supplementary Figure 1]. Firstly, 1.5 mm thick cross-sectional wood slices were pretreated with sodium chlorite to partially remove lignin. Scanning electron microscopy (SEM) images [Supplementary Figure 2] confirm that the wood slices retain their layered porous structure, including vessels, tracheids, and pits. However, parts of the middle lamellae in the cell corners disappear [Supplementary Figure 2B], and the cell walls become rough and uneven. Quantitative analysis revealed a significant reduction in lignin content within the balsa wood, decreasing from 22% to 4.3% after treatment [Supplementary Figure 3]. The significant removal of lignin in the wood cell wall increases the mesoporosity of the derived carbon; this mesoporosity enhancement improves overall carbon porosity, thereby promoting efficient mass transport. Moreover, this pretreatment exposes more cellulose on the wood surface, which may promote a higher degree of graphitization during carbonization, thereby enhancing material conductivity. The abundant hydroxyl groups in cellulose effectively anchor Fe³⁺, which serves as catalytic species and nucleation sites for CNT growth. CNTs are uniformly distributed across the porous carbon matrix, further enhancing material conductivity [Figure 1B and C]. Transmission electron microscopy (TEM) images [Figure 1D] confirm the microstructure of the Fe₃C@BNC catalyst, indicating the in situ formation of CNTs integrated with wood-derived carbon, consistent with the SEM results. Fe₃C NPs are uniformly distributed within CNTs, which extend outward, thereby enhancing electron transport and providing additional active sites. High-resolution TEM (HRTEM) images [Figure 1E] confirm a lattice spacing of 0.21 nm, corresponding to the (121) plane of Fe₃C. Notably, Fe₃C NPs are fully encapsulated by a ~3 nm thick CNT layer. High-angle annular dark-field scanning TEM (HAADF-STEM), energy-dispersive X-ray spectroscopy (EDS) elemental mapping images [Figure 1F] and EDS spectra [Supplementary Figure 4] indicate the uniform distribution of Fe, C, N, and B elements within Fe₃C@BNC. This synthesis strategy effectively prevents metal aggregation, which maximizes active site availability and enhances catalytic activity. For comparison, the SEM images of Fe₃C@NC display a similar morphology [Supplementary Figure 5]. Conversely, the BNC sample exhibits significant shrinkage of vessels and tracheids owing to high-temperature carbonization [Supplementary Figure 6]. In the absence of Fe³⁺, CNTs do not form on the BNC surface.

To further investigate the structure and composition of Fe₃C@BNC, X-ray diffraction (XRD) patterns on Fe₃C@BNC, Fe₃C@NC, and BNC were analyzed [Figure 2A]. The peaks at 22.5° and 43.5° correspond to the (002) and (101) crystal planes of wood-derived carbon, respectively. A distinct diffraction peak at 26.0° in Fe₃C@BNC and Fe₃C@NC confirms the successful synthesis of CNTs, consistent with previous reports^[42]. Additionally, peaks between 42.1° and 46.5° correspond to Fe₃C (PDF#00-003-0989), confirming the formation of Fe₃C NPs during pyrolysis. The Raman spectra of Fe₃C@BNC, Fe₃C@NC, and BNC catalysts exhibit characteristic D (1,350 cm^-1) and G (1,580 cm^-1) bands of carbon materials. The BNC sample features an I_D/I_G ratio of 1.07, which decreases to 0.93 for Fe₃C@NC. This reduction is mainly attributed to the significant presence of Fe₃C, which facilitates the formation of highly graphitized CNTs. With the incorporation of B atoms into the hybrid catalyst, the I_D/I_G ratio of Fe₃C@BNC increases to 1.01, indicating a higher defect level due to additional heteroatom doping. The presence of a 2D peak in the graphitized samples further confirms CNT contributions on the surface [Figure 2B].

The specific surface area and pore structure are crucial for reactant adsorption and mass transportation during catalysis. To evaluate these properties, Brunauer-Emmett-Teller (BET) specific surface area analysis was conducted. The Fe₃C@BNC catalyst exhibits a surface area of 411.6 m²·g^-1, which surpassed those of the other samples [Figure 2C], with a predominant pore size of ~4 nm [Supplementary Figure 7]. A comparison of Fe₃C@BNC with a sample without lignin removal [Supplementary Figure 8] further confirms that lignin removal contributes to an increased surface area. Cross-sectional conductivity measurements of the wood-derived carbon materials indicate improved conductivity with increasing carbonization temperature [Supplementary Figure 9]. Fe₃C@BNC exhibits the highest conductivity of 97.2 S·m^-1, which is thrice that of BNC. Rapid electron mobility is crucial for improving intrinsic catalytic performance, particularly in proton-coupled electron transfer electrocatalytic processes.

To further investigate the chemical states and bonding configurations of Fe₃C and B, N-doped CNTs, X-ray photoelectron spectroscopy (XPS) analysis was performed. The XPS survey spectrum of Fe₃C@BNC confirms the presence of B, O, N, C, and Fe elements [Supplementary Figure 10A]. High-resolution peaks were analyzed to determine the chemical states of the as-prepared samples. The peaks at 284.8, 286.0, and 288.3 eV correspond to the C–C/C=C bonds, C–N bonds, and C–O bonds, respectively [Supplementary Figure 10B]. Notably, the peak at 283.8 eV in the Fe₃C@BNC spectrum corresponds to C–Fe/C–B bonds, indicating successful B doping. The high-resolution N 1s spectrum of the Fe₃C@BNC catalyst [Figure 2D] can be deconvoluted into five components: N-B (398.2 eV), pyridinic N (398.8 eV), pyrrolic N (399.9 eV), graphitic N (401.3 eV), and oxidized N (404.0 eV). In contrast, Fe₃C@NC lacks N–B bonding, as evidenced in Supplementary Figure 10C. A higher content of pyridinic N facilitates O₂ molecule adsorption and O–O bond cleavage, which enhances the catalytic activity of ORR. The high-resolution B 1s spectrum confirms successful B doping [Figure 2E]^[43-45], with fitted B–N bonds indicating that B atom incorporation alters the electronic structure of CNTs. Previous studies have shown that pyridinic N–B configurations reduce the free energy of the rate-determining step in ORR intermediate formation, thereby enhancing electrocatalytic performance^[46]. The atomic percentages of B and N in Fe₃C@BNC and Fe₃C@NC are presented in Supplementary Table 1. The high-resolution Fe 2p_3/2 spectrum exhibits a peak at 707.0 eV, corresponding to the Fe–C bond from iron carbide, consistent with XRD and HRTEM analyses. Additional peaks at 707.8 and 709.5 eV are assigned to Fe²⁺ and Fe³⁺, resulting from the partial oxidation of iron [Figure 2F]^[47].

To investigate the electrocatalytic activity of the as-prepared Fe₃C@BNC, linear sweep voltammetry (LSV) was performed to evaluate its ORR performance. The LSV curves [Figure 3A] recorded using a rotating disk electrode at 1,600 rpm indicate that Fe₃C@BNC exhibits an onset potential (E_onset = 0.90 V) and a half-wave potential (E_1/2 = 0.83 V) comparable to those of commercial Pt/C. Moreover, Fe₃C@BNC achieves a higher limiting current density (J_L = 6.1 mA·cm^-2). The excellent catalytic performance of Fe₃C@BNC is mainly attributed to the regulatory effect of internal Fe₃C NPs on the carbon shell, which significantly alters the electron density of the carbon layer^[41,48]. The interfacial hybridization between the carbon shell and the metallic core enhances the electrochemical activity of the carbon surface. Additionally, boron doping plays a crucial role in further improving catalytic activity. A comparative analysis reveals that Fe₃C@BNC outperforms Fe₃C@NC, highlighting the critical role of the tailored electronic structure in the carbon layer and Fe₃C for catalytic reactions. The study examined the impact of precursor concentration and wood structure regulation on the performance of the catalyst, identifying 0.01 M Fe³⁺ as the optimal concentration [Supplementary Figure 11]. Excessively high concentrations lead to the formation of larger metal aggregates, which are unfavorable for the formation of catalytic CNTs. On the other hand, too low concentration results in an insufficient number of catalytic sites, thereby reducing activity. Furthermore, the removal of lignin aids in anchoring more Fe³⁺, which in turn promotes uniform growth of CNTs and enhances conductivity [Supplementary Figure 12]. The impact of synthesis temperature on Fe₃C@BNC performance was investigated [Supplementary Figures 13 and 14]. The results reveal that electrocatalysts carbonized at 900 °C exhibit a significantly higher activity than those carbonized at 800 and 700 °C. This improvement is likely due to higher temperatures promoting the graphitization of wood-derived carbon with CNTs, which enhances electrical conductivity and facilitates electron transfer. CNTs exhibit a greater tendency to interact with Fe₃C NPs, further promoting the catalytic reaction. Figure 3B displays the Tafel plots of different samples. Fe₃C@BNC exhibits a Tafel slope of 108 mV·dec^-1, which is closely similar to that of Pt/C (107 mV·dec^-1), and significantly lower than other samples, confirming its superior reaction kinetics. The cyclic voltammetry (CV) curves of Fe₃C@BNC in oxygen- and nitrogen-saturated electrolytes [Supplementary Figure 15] display a distinct reduction peak at 0.8 V, corresponding to ORR. A comparison of CV curves in oxygen-saturated electrolytes for different samples [Figure 3C] reveals a more positive reduction peak for Fe₃C@BNC, indicating its superior ORR performance. To provide detailed insights into the ORR kinetics of Fe₃C@BNC [Supplementary Figure 16], the rotating ring-disk electrode technique was used to determine the electron transfer number (n) and hydrogen peroxide yield [Figure 3D]. Fe₃C@BNC maintains an n value greater than 3.9 from 0.2 to 0.7 V, with an H₂O₂ yield below 3% over a wide potential range. This suggests that the ORR process on the Fe₃C@BNC electrode involves a highly selective four-electron transfer pathway, which facilitates the efficient conversion of H₂O to OH^-. The Fe₃C@BNC electrocatalyst exhibits a low electrochemical impedance of 35.9 ohms, contributing to its high ORR activity [Supplementary Figure 17 and Supplementary Table 2]. The electrochemical active surface area of the catalyst was measured through CV tests at various scan rates [Supplementary Figure 18]. Compared with BNC (1.82 mF·cm^-2), Fe₃C@BNC exhibits a significantly higher C_dl value (8.70 mF·cm^-2). This indicates that CNTs reduce interfacial resistance and enhance the contact area between Fe₃C@BNC and the electrolyte through the carbon coating layer, thereby accelerating the reaction process. The durability of the catalyst was assessed through accelerated degradation tests. The i–t results reveal that after 16,000 s, Fe₃C@BNC and Pt/C retain ~90.3% and 62.8% of their initial current, respectively [Figure 3E].

OER is a crucial electrode process during the charging of ZABs. To evaluate the OER Fe₃C@BNC, polarization curves were recorded using a 1 × 1 cm² wood-derived carbon electrode as a self-supporting substrate, with a scan rate of 1 mV·s^-1. Fe₃C@BNC achieves a current density of 10 mA·cm^-2 at a low overpotential of 200 mV [Figure 3F], which is significantly lower than that of RuO₂. In contrast, the reference sample of Fe₃C@NC exhibits an overpotential of 320 mV, similar to that of RuO₂. When the catalyst operates in a 0.1 M KOH electrolyte, the overpotential of Fe₃C@BNC is 250 mV [Supplementary Figure 19]. Durability analysis further confirms the excellent stability of Fe₃C@BNC, with only an 18 mV decrease in overpotential after 1,000 CV cycles [Figure 3G]. This remarkable stability for both ORR and ORE is attributed to the protective CNT outer layer, which prevents electrolyte-induced etching of the internal Fe₃C, thereby enhancing its potential for ZAB applications. To investigate the structural stability of Fe₃C@BNC, various characterizations were performed. The XRD pattern of the post-reaction catalyst [Supplementary Figure 20] confirms that the Fe₃C structure remains intact. SEM observations reveal that the wood-derived carbon surface remains uniformly covered with abundant CNTs even after prolonged reaction, indicating no significant structural degradation [Supplementary Figure 21A and B]. Post-reaction TEM images of Fe₃C@BNC [Supplementary Figure 21C] further indicate that Fe₃C remains uniformly dispersed in the carbon matrix. The measured interplanar spacing of 0.168 nm corresponds to the (023) crystal plane of Fe₃C [Supplementary Figure 21D], while the outer layer of Fe₃C NPs remains encapsulated by CNTs. This stable encapsulation prevents the inner Fe₃C from being etched by strong alkaline solutions during reactions, thereby enhancing catalyst stability. HAADF–STEM and EDS mapping confirm the uniform distribution of Fe, C, N, and B elements in Fe₃C@BNC [Supplementary Figure 21E], consistent with the pre-reaction results.

To evaluate the dual-functional oxygen electrocatalytic performance, the potential gap (ΔE) between ORR E_1/2 and OER E_j=10 was calculated from LSV curves. The Fe₃C@BNC electrode exhibits a ΔE of 0.65 V, which is significantly lower than that of Pt/C+RuO₂ [Figure 3H]. Additionally, the ΔE values of wood-derived dual-functional oxygen electrocatalysts were summarized to assess the activity of Fe₃C@BNC [Figure 3I and Supplementary Table 3]^{[22,25,26,49-51]}. Among other samples, Fe₃C@BNC exhibits the lowest ΔE value. Beyond heteroatom doping, the enriched CNTs on the wood-derived carbon play a crucial role in enhancing material conductivity, facilitating efficient electron transfer, and accelerating reaction kinetics. These features establish a strong basis for the application of Fe₃C@BNC in ZABs.

To illustrate the practicality of Fe₃C@BNC, liquid ZABs were assembled using Fe₃C@BNC as the air cathode. The schematic of the ZAB system is shown in Figure 4A. For comparison, a ZAB with commercial Pt/C and RuO₂ catalyst was constructed. Fe₃C@BNC-based ZABs exhibit an open-circuit voltage (OCV) of 1.46 V [Figure 4B]. Additionally, the LAND testing system confirms that ZABs maintain an OCV of 1.46 V for over 80 min, consistent with multimeter measurements and close to the theoretical value. The Pt/C+RuO₂ ZABs exhibit a lower OCV of 1.35 V. Polarization curves and corresponding power density measurements reveal that Fe₃C@BNC ZABs achieve a peak power density of 114 mW·cm^-2, which exceeds the 98 mW·cm^-2 of Pt/C-based ZABs [Figure 4C]. This superior performance can be attributed to the abundant active sites generated by heteroatom doping and the enhanced electron transport facilitated by CNTs within the wood-derived carbon.

To assess the rate performance of Fe₃C@BNC ZABs, the battery voltage was monitored at varying current densities. Even as the current density increases from 2 to 20 mA·cm^-2, Fe₃C@BNC ZABs maintain higher discharge voltage than Pt/C+RuO₂ ZABs. As the current density decreases to 2 mA·cm^-2, the voltage successfully recovers to its initial value [Figure 4D]. Moreover, Fe₃C@BNC ZABs exhibit a specific capacity of 804.5 mA·h·g^-1 at 4 mA·cm^-2 [Figure 4E], exceeding the 729.3 mA·h·g^-1 of Pt/C+RuO₂ ZABs, indicating superior discharge capacity. In practical applications, this high performance enables Fe₃C@BNC ZABs to power a light-emitting diode light, with two batteries connected in series [Figure 4F]. To assess the cycling stability of ZABs, constant current charge and discharge tests were conducted at 2 mA·cm^-2 [Figure 4G]. Notably, Fe₃C@BNC ZABs maintain stable operation for over 420 h. A comparison of the discharge voltage gap at ~50 and 420 h [Supplementary Figure 22] reveals only a slight increase from 0.91 to 0.93 V, indicating a stable discharge platform even after 400 h. In contrast, the charge–discharge curve of Pt/C+RuO₂ ZABs indicates a diffusion trend after only 270 h. The designed Fe₃C@BNC exhibits excellent stability, similar to other carbon-coated catalysts, and benefits from electronic modulation between Fe₃C NPs and the B, N co-doped carbon shell. This synergy effectively optimizes the charging and discharging processes in ZABs, highlighting Fe₃C@BNC as a promising air electrode for ZAB applications.