C–Na–O electrostatic interactions boost the kinetics of coal-derived hard carbon anodes for high-performance sodium-ion batteries†

Tianqi Xu^abc, Zongxu Yao^abc, Wei Jiang^abc, Yaxin Chen^d, Chenmin Liu^abc, Yinshuang Guan^abc, Zhiqiang Tang^abc and Liang Dong*^abc
aJiangsu Key Laboratory for Clean Utilization of Carbon Resources, China University of Mining and Technology, Xuzhou 221116, China. E-mail: dongl@cumt.edu.cn
^bInternational Joint Laboratory of Minerals Efficient Processing and Utilization, Ministry of Education, China University of Mining and Technology, Xuzhou 221116, China
^cSchool of Chemical Engineering & Technology, China University of Mining and Technology, Xuzhou 221116, China
^dSchool of Materials Science and Physics, China University of Mining and Technology, Xuzhou, 221116, China

First published on 3rd July 2025

---

---

1. Introduction

Sodium-ion batteries (SIBs) have emerged as one of the most promising alternatives to lithium-ion batteries (LIBs) for large-scale energy storage applications, owing to their intrinsic safety, excellent cost-effectiveness, and the natural abundance of sodium resources.^1,2 Among various candidate anode materials, hard carbon has attracted widespread attention due to its relatively high theoretical capacity, wide availability of precursors, and tunable microstructure, making it a strong contender for next-generation SIBs.^3,4 In recent years, coal—an inexpensive and widely distributed carbon-rich feedstock—has been considered a viable precursor for hard carbon materials.⁵ However, the direct pyrolysis of coal typically results in products with excessive graphitic disorder, low specific surface area, and poor pore structure, which severely restrict their sodium storage capabilities.^6,7

In recent years, co-pyrolysis of biomass with coal has emerged as an effective strategy for tailoring the microstructure of coal-derived hard carbon. For example, studies by Aboyade et al.,⁸ Ding et al.,⁹ and Wu et al.¹⁰ have shown that the incorporation of biomass during pyrolysis can significantly improve the pore structure and morphology of the resulting carbon materials. Further, Chen et al.¹¹ revealed that sucrose can form ester-linked cross-linked structures with lignite at around 400 °C, thereby enhancing the structural stability of the carbon precursor. Zhou et al.¹² also demonstrated that glucose, rich in hydroxyl groups, can graft onto lignite during low-temperature carbonization, suppressing gas release and impurity formation, reducing structural defects, and ultimately improving the electrochemical performance of coal-derived hard carbon. Despite these benefits, co-pyrolysis of coal with glucose can only moderately enhance the initial coulombic efficiency (ICE), which typically remains in the range of 82–84%, falling short of achieving high-ICE hard carbon materials. Against this backdrop, researchers have begun exploring the use of inorganic sodium sources to further regulate ICE.^13,14 Among them, sodium carbonate (Na₂CO₃) has attracted considerable attention due to its ability to improve pyrolysis efficiency, alter decomposition kinetics, and activate intrinsic mineral matter in coal. Previous studies have confirmed that Na₂CO₃ can accelerate volatile release and reduce the activation energy of coal pyrolysis,¹⁵ as reported by Valluri et al.,¹⁶ Liu et al.,¹⁷ and Yang et al.¹⁸ Additionally, Zhao et al.¹⁹ showed that sludge carbonized in a molten KOH/Na₂CO₃ system exhibited significant improvements in specific surface area and carbon content. However, whether Na₂CO₃ can synergistically promote carbon structural ordering and Na⁺-active site formation during the co-pyrolysis of coal and glucose—and how it regulates the sodium storage mechanism at the atomic level—remains largely unexplored and warrants further investigation.

Herein, we propose a dual-additive co-pyrolysis strategy that combines glucose and Na₂CO₃ to engineer the multi-scale structure of coal-derived hard carbon. Specifically, glucose, owing to its high oxygen content and propensity for low-temperature carbonization, contributes to the introduction of oxygen-containing functional groups and adjusts the degree of disorder in the carbon skeleton. These modifications are expected to improve Na⁺ diffusion pathways. Concurrently, Na₂CO₃ serves as a synergistic modifier with two key roles: on one hand, Na⁺ can intercalate into the interlayer spaces of carbon materials at high temperatures, thereby expanding the interlayer spacing and improving the Na⁺ kinetics. On the other hand, the alkaline environment generated upon Na₂CO₃ decomposition may catalyze the coal–glucose co-carbonization reaction, thereby promoting the development of micro- and mesoporous structures. This integrated co-pyrolysis route enables simultaneous tuning of pore structure, surface chemistry, and promotes the formation of C–Na–O Structure, ultimately offering a promising pathway for constructing high-capacity, long-cycle-life coal-based hard carbon anodes for advanced sodium-ion batteries.

2. Experimental

2.1 Materials synthesis

Hard carbon precursors were synthesized from Jilangde bituminous coal (Xinjiang, China) with a particle size of less than 0.075 mm. In the standard preparation procedure, raw coal (RC) was blended with glucose at mass ratios of 1:0, 1:0.5, 1:1, and 1:2, followed by homogenization using a planetary ball mill at 300 rpm for 2 hours. The resulting mixtures were denoted as BCG0, BCG0.5, BCG, and BCG2, respectively. For sodium-modified precursors, RC was first co-dispersed with Na₂CO₃ in deionized water at a mass ratio of 9:1 (RC:Na₂CO₃), then oven-dried at 80 °C to remove moisture. The dried composite was subsequently mixed with glucose (final RC:glucose:Na₂CO₃ = 9:9:1) using the same milling parameters. This ternary precursor was labeled as BCGNa0.1. All precursor samples were subjected to a two-step pyrolysis process in a tubular furnace under flowing high-purity argon (99.99%). The first-stage carbonization was conducted at 300 °C for 2 hours to stabilize the intermediate structures, followed by a second-stage high-temperature treatment at 1300 °C for 2 hours with a heating rate of 2 °C min⁻¹. The final carbon materials were designated as CG0, CG0.5, CG, CG2, and CGNa0.1, corresponding to their respective precursors.

To evaluate the necessity and effect of acid leaching on sodium storage performance, a modified preparation route was employed. Raw coal (RC) was first demineralized through sequential acid treatment. The procedure involved initial digestion in 5 M HCl solution (solid–liquid ratio 1:30, stirred for 12 hours), followed by etching in 5 M HF under the same conditions to remove residual inorganic minerals. The acid-treated samples were thoroughly rinsed with deionized water until a neutral pH was achieved, and subsequently oven-dried at 105 °C to obtain deashed coal (CC). The CC was then combined with glucose at the same mass ratios used for the BCG series (i.e., 1:0, 1:0.5, 1:1, and 1:2), followed by identical ball milling and pyrolysis procedures under argon atmosphere. The resulting hard carbon products were labeled as CCG0, CCG0.5, CCG, and CCG2. This parallel preparation route allowed for a systematic comparison between acid-washed and unwashed precursors in terms of their structural features and electrochemical properties.

2.2 Materials characterization

A comprehensive characterization protocol was employed to investigate the multi-scale structural and chemical features of the coal-derived precursors. Thermogravimetric analysis (TGA) coupled with differential scanning calorimetry (DSC) was conducted on a NETZSCH STA449F5 system to evaluate thermal decomposition behavior and cross-linking evolution during pyrolysis. Elemental compositions were measured using a Flash 2000 elemental analyzer operating in dual modes for CHNS and oxygen detection. The interactions among C, O, and Na were investigated using Time-of-Flight Secondary Ion Mass Spectrometry (TOF-SIMS, PHI nanoTOFII Time-of-Flight SIMS). Crystalline structure and graphitization degree were analyzed by X-ray diffraction (XRD, Bruker D8 Advance, Cu Kα, λ = 1.5406 Å), while carbon skeleton configurations were examined using solid-state ¹³C cross-polarization/magic angle spinning nuclear magnetic resonance (CP/MAS NMR, Bruker Avance Neo 600 MHz). Functional group compositions were identified by Fourier-transform infrared spectroscopy (FTIR, Shimadzu IRTracer-100, 4 cm⁻¹ resolution). Surface chemical states were characterized by X-ray photoelectron spectroscopy (XPS, ESCALAB 250Xi), with the C 1s peak calibrated to 284.8 eV. Morphological and microstructural features were observed via field-emission scanning electron microscopy (FE-SEM, TESCAN GAIA3 XMH) and high-resolution transmission electron microscopy (HRTEM, Tecnai G2 F20, 200 kV). Structural defects were assessed by Raman spectroscopy (Horiba LabRAM HR, 532 nm laser). Porosity characteristics, including specific surface area and pore size distribution, were evaluated using nitrogen adsorption–desorption isotherms based on the Brunauer–Emmett–Teller (BET) method and analyzed on Micromeritics ASAP 2460 system. The ultra-micropore structures were investigated using small-angle X-ray scattering (SAXS, Xeuss 2.0).

2.3 Electrode fabrication and electrochemical characterization

Electrochemical evaluations were carried out using CR2032-type coin cells assembled under rigorously controlled conditions. The working electrodes were fabricated by homogenizing coal-derived hard carbon (90 wt%), Super P conductive carbon black (5 wt%), and polyvinylidene fluoride (PVDF, 5 wt%) in N-methyl-2-pyrrolidone (NMP) to form a uniform slurry, which was then cast onto carbon-coated copper foil using a doctor-blade technique. The electrodes were dried and punched into circular disks with an active material loading of approximately 2.5 mg cm⁻². Sodium metal foil was used as both counter and reference electrode, and Whatman GF/D glass fiber served as the separator. The electrolyte consisted of 1 M sodium hexafluorophosphate (NaPF₆) dissolved in diethylene glycol dimethyl ether (diglyme). Cell assembly was conducted in an argon-filled glovebox (H₂O/O₂ < 1 ppm) to ensure an inert atmosphere. The initial capacity evaluation was conducted using constant current charge–discharge tests at a current density of 20 mA g⁻¹ within a voltage window of 0.01–3.0 V (vs. Na⁺/Na) on CT2001 battery testing system (Wuhan Land Electronics Co., Ltd, China). Subsequent rate and long-cycle performance tests were also carried out within the same voltage range of 0.01–3.0 V. Cyclic voltammetry (CV) measurements were conducted on a DH7003B electrochemical workstation (Donghua Analytical Instruments) over the same voltage range with scan rates from 0.1 to 1.0 mV s⁻¹. Full-cell configurations were constructed using optimized CGNa0.1 as the anode and sodium nickel iron manganese oxide (NFM-Na) as the cathode, operated within a voltage range of 1.5–3.8 V. All electrochemical measurements were conducted at 25 ± 1 °C in a thermostatically controlled environment.

2.4 Computational details

Before performing pyrolysis kinetics and sodium-ion diffusion simulations, a systematic molecular modeling workflow was established to construct and validate representative coal-based structures. Initially, molecular model of RC was constructed based on element analysis (Table 1), solid-state NMR spectral data (Fig. S1 and Table S1†) and XPS analysis (Fig. S2 and Table S2†). The structural accuracy of the model was verified by comparing simulated ¹³C NMR spectra, generated using MestReNova software, with experimental results (Fig. S3†). Geometric optimization of the coal molecule was carried out in the Forcite module using the COMPASSIII force field to achieve local energy minimization under medium precision. To obtain the global energy minimum, annealing dynamics simulations were performed with thermal cycling between 300 K and 800 K (Fig. S4a†). The optimized structure was then used to build three representative systems (RC, BCG, and BCGNa0.1), each confined within a cubic simulation box (α = β = γ = 90°) and initialized at a uniform density of 0.5 g cm⁻³. These systems were subjected to density optimization under an NPT ensemble at 300 K and 0.01 GPa for 200 ps with a time step of 0.25 fs (Fig. S4b†). Subsequent reactive molecular dynamics (ReaxFF-MD) simulations were performed in LAMMPS using the ReaxFF force field²⁰ under an NVT ensemble at 1573 K for 200 ps with a time step of 0.1 fs. Computational reliability was ensured by adopting a dual-force-field strategy, where COMPASSIII was used for non-reactive geometry optimization, and ReaxFF was employed for simulating thermally induced reactive events.

---

Based on DFT, the CASTEP module in Materials Studio was employed to calculate the adsorption and diffusion of Na atoms on the carbon surface. All calculations were carried out using the generalized gradient approximation (GGA) with the PBE²¹ exchange-correlation functional. K-points on the Brillouin zone were 1 × 1 × 2, and the energy cutoff of 517 eV. To ensure the stability of the optimized geometries, the force convergence was set to be 0.05 eV·Å⁻¹.

3. Results and discussion

3.1 Structural evolution of C–Na–O during pyrolysis

Fig. 1a illustrates the preparation pathway of coal-derived hard carbon. Through the synergistic regulation of glucose and Na₂CO₃, O and Na elements were successfully introduced into the hard carbon. As shown in Table 1, the oxygen content increased from 4.95% to 6.39%, while the sodium content, determined by ICP analysis, reached 0.17%, providing the basis for the formation of C–Na–O interactions. To further elucidate the effects of Na and O incorporation on the pyrolysis behavior of coal, ReaxFF-MD simulations were performed on the RC, BCG, and BCGNa0.1 models. Fig. 1b–d depict the evolution of carbon structures during a 200 ps simulation at 1573 K. The pyrolysis products were classified into four groups based on the number of carbon atoms: C1–C5, C6–C10, C11–C15, and C16+. Specifically, C1–C5 fragments correspond to volatile species, while C6–C10 and C11–C15 represent intermediate products; during pyrolysis, C6–C10 fragments volatilize further, whereas C11–C15 fragments gradually condense into C16+ fragments. The product distribution provides direct insight into the carbon fragmentation and condensation processes involved in hard carbon formation. In the RC model, pronounced fluctuations in C16+ fragments indicate a high degree of structural disorder during pyrolysis. In contrast, BCG and BCGNa0.1 exhibit relatively stable C16+ fragments throughout the simulation, suggesting a more ordered structural evolution.

Furthermore, Fig. 1e reveals that the CG0 sample exhibits highly disordered carbon layers, which facilitate Na⁺ adsorption and enhance capacity in the high-voltage region. The introduction of glucose reduces the edge disorder of the carbon layers, while the addition of Na₂CO₃ promotes the formation of abundant nanopores during pyrolysis, driven by the release of CO₂. Na₂CO₃ decomposes at approximately 850 °C into Na₂O and CO₂, followed by further reaction at higher temperatures to form metallic Na.²² The resulting Na deposits onto the hard carbon surface within the sealed ceramic boat, facilitating the formation of C–Na–O interactions. To further confirm the existence of C–Na–O interaction structures in the coal-derived material, TOF-SIMS measurements were performed. As shown in Fig. 2, several characteristic secondary ion fragments were detected in the CGNa0.1 sample, including [NaCO₃]⁻ (m/z = 82.974), [NaCO₄]⁻ (m/z = 98.987), [NaHCO₄]⁻ (m/z = 99.985), [NaC₃CO₄]⁻ (m/z = 122.969), [Na₃CH₂O₅]⁻ (m/z = 162.959), and [Na₃CO₃]⁺ (m/z = 128.972). These fragments are consistent with the decomposition patterns of sodium carboxylate species, indicating stable interactions between Na and the C/O framework. Based on the detection of fragments such as [Na₃CH₂O₅]⁻, it is further inferred that Na may exist in both chemically bonded and physically adsorbed forms with C and O. Notably, Na₂CO₃ plays a critical role in both accelerating the pyrolysis of coal (Fig. S5†) and promoting the formation of an ordered microstructure, leading to the development of hard carbon with an interlayer spacing of up to 0.41 nm. This structural feature is beneficial for enhancing Na⁺ diffusion kinetics and rate performance. In addition, SEM-EDS and TEM-EDS analyses (Fig. S6†) reveal that the introduced Na is uniformly distributed within the material, further confirming the significant impact of Na incorporation on the electrochemical performance.

The structural evolution of CG0, CG, and CGNa0.1 samples was systematically examined, as illustrated in Fig. 3. The XRD patterns of the carbonized samples (Fig. 3a) show broad (002) and (100) diffraction peaks, characteristic of amorphous carbon structures. Compared with the samples before carbonization (Fig. S8a†), the characteristic peaks associated with glucose completely vanished after carbonization, confirming the complete decomposition of glucose during thermal treatment. Additionally, no sodium oxide diffraction peaks were detected in the CGNa0.1 sample, implying that the introduced sodium was embedded into the hard carbon matrix through electrostatic interaction. Moreover, the addition of glucose and Na₂CO₃ effectively suppressed the retention of mineral phases in coal. Raman analysis (Fig. 3b) indicated changes in carbon structural disorder, with the CG0 sample showing a highly disordered carbon layer structure and an AD/AG ratio of 3.31. Glucose doping greatly reduced this ratio, reflecting a clear decrease in defect density. Following the addition of Na₂CO₃, the AD/AG ratio of CGNa0.1 was close to that of CG, further highlighting the dominant role of glucose in controlling carbon layer ordering (Table S3†). In addition, The in-plane crystallite size L_a^23,24 was calculated based on the A_D1/A_G ratio using the modified Tuinstra–Koenig relation:

La = Cλ/(AD1/AG)	(1)

The constant C_λ in the equation was determined according to the method reported in ref. 25, with a value of 4.67. Based on this, the in-plane crystallite sizes L_a of CG0, CG, and CGNa0.1 were calculated to be 1.41 nm, 2.00 nm, and 2.08 nm, respectively. FTIR analysis (Fig. 3c) revealed that glucose doping led to intensity variations in the C–O–C vibration band (975–1197 cm⁻¹), indicating structural changes induced by oxygen incorporation. The addition of Na₂CO₃ further adjusted the vibration pattern, likely due to the impact of C–Na–O electrostatic interactions on the C–O–C bond vibrations. Overall, these findings demonstrate that the combined effects of glucose and Na₂CO₃ effectively tuned the microstructural ordering and modified the C–O–C bonding environment, thus providing a structural basis for improved sodium storage performance.

XPS analysis (Fig. 3d–f) reveals notable modifications in surface chemical composition. Compared to CG0, both CG and CGNa0.1 exhibit increased sp²-hybridized carbon and decreased sp³-hybridized carbon content, indicating enhanced graphitization and reduced structural disorder, which aligns well with the Raman results. Interestingly, CGNa0.1 shows higher absolute amounts of both sp²C and sp³C species compared to CG, which may be attributed to lattice distortions induced by Na incorporation; however, the sp²C/sp³C ratio is nearly close (Table S4†), suggesting that their degrees of order are similar. The O 1s spectra further confirm the enrichment of C–O–C functional groups in CG and CGNa0.1 (Table S5†), consistent with the FTIR results. In addition, the Na 1s spectrum of the CGNa0.1 sample (Fig. 3f) shows a prominent peak at approximately 1071.5 eV, attributed to Na, and another peak appears at around 1073.2 eV, which is ascribed to the C–Na–O interaction. Nitrogen adsorption–desorption isotherms (Fig. 3g) exhibit typical Type II characteristics, with CGNa0.1 achieving the highest BET surface area (7.02 m² g⁻¹) compared to CG0 (1.70 m² g⁻¹), reflecting its more developed pore structure. The pore size distribution (Fig. 3h) shows that the CG sample mainly exhibits a macroporous structure centered around 40–80 nm, which contributes to enhanced capacity in the high-voltage region. In contrast, CGNa0.1 demonstrates a significantly increased surface area and microporosity, which facilitate ion transport and electrolyte accessibility, thereby improving its electrochemical performance. To further investigate the microporous structures of the three samples, we compared the SAXS curves (q range: 0.06–1 Å) of CG0, CG, and CGNa0.1, and analysed their pore size distributions in the 0.6–2 nm range, as shown in Fig. 3i. The results indicate that CGNa0.1 contains a large number of ultramicropores, further confirming its abundant nanoporous architecture.

3.2 Electrochemical performances and sodium storage mechanism

The electrochemical performance of CG0, CG, and CGNa0.1 was systematically evaluated, as illustrated in Fig. 4. Fig. 4a shows the initial charge–discharge curves of CG0, CG, and CGNa0.1. Among them, CG exhibits the highest discharge capacity (353.1 mA h g⁻¹), while CGNa0.1 demonstrates the highest initial coulombic efficiency (90.46%). The dQ/dV analysis of the three curves is presented in Fig. 4b. Compared with CG0 and CG, CGNa0.1 shows lower peak intensity but a smoother and more symmetric profile, indicating a milder and more reversible Na⁺ insertion process. This also reflects enhanced structural stability and interfacial reaction stability, contributing to its excellent overall electrochemical performance. To further analyse the sodium storage mechanism during the initial discharge process, the discharge capacity in Fig. 4a was divided into slope and plateau regions to explore the role of different structural features. The sloping region is primarily associated with Na⁺ adsorption on defect sites and disordered surfaces, while the plateau region corresponds to Na⁺ intercalation into the carbon matrix and pore filling, as shown in Fig. S7a.†²⁶ The reversible capacities of CG0, CG, and CGNa0.1 were 228.8, 264.6, and 299.8 mA h g⁻¹, respectively, with corresponding initial coulombic efficiencies (ICEs) of 73.7%, 74.9%, and 90.46%. The incorporation of sodium carbonate significantly improves the initial efficiency, indicating enhanced reversibility of sodium storage. CG0 exhibits a high sloping-region capacity of 154.4 mA h g⁻¹, which is attributed to its highly disordered structure that facilitates extensive Na⁺ adsorption. Similarly, CG delivers a sloping capacity of 158.8 mA h g⁻¹, which, despite its more ordered structure, can be attributed to the presence of abundant macropores that provide additional adsorption sites.²⁷ In contrast, CGNa0.1 exhibits a lower adsorption capacity of 139.2 mA h g⁻¹, owing to its reduced macroporosity, higher structural order, and partial occupation of active sites by Na⁺ derived from Na₂CO₃. Notably, CGNa0.1 exhibits a “gentle slope” feature in the charging curve within the voltage range of 0.5–0.9 V, which is primarily attributed to the Na⁺ desorption process induced by the weakened C–Na–O interactions. This process contributes to the observed increase in charging capacity. Moreover, we analysed the second-cycle charge–discharge curves of CG0, CG, and CGNa0.1, as shown in Fig. S7b.† The results show that the overall profile of CGNa0.1 remains largely unchanged compared to the first cycle, indicating that the initially formed SEI film is relatively stable and has minimal impact on its subsequent electrochemical behaviour. The high ICE of CGNa0.1 can be attributed to the introduction of C–Na–O electrostatic interactions, abundant closed-pore structure, and expanded interlayer spacing, all of which effectively reduce irreversible sodium loss and promote efficient desodiation.

Fig. 4c compares the rate performance of CG0, CG, and CGNa0.1, with detailed charge–discharge profiles of CGNa0.1 shown in Fig. 4d. Within the low current density range of 20–30 mA g⁻¹, both CG and CGNa0.1 maintained stable capacities. However, CG exhibited significant capacity decay at higher current densities (100–500 mA g⁻¹), followed by a sharp drop at 500–1000 mA g⁻¹. In contrast, CG0 displayed a continuous decline in capacity across the entire range from 20 to 1000 mA g⁻¹. All samples showed good rate recovery when the current density was returned to 200 mA g⁻¹, indicating structural resilience. For CGNa0.1, the characteristic desodiation “gentle slope” in the 0.5–0.9 V range, observed under 20–30 mA g⁻¹, gradually disappeared at higher current densities. This phenomenon suggests that with increasing current, the contribution from intercalation and pore-filling mechanisms decreases, while the disturbance to surface-adsorbed Na⁺ during extraction is reduced. The weakened interaction leads to the disappearance of the “gentle slope” in the charging curve, reflecting changes in sodium storage kinetics under high-rate conditions. Fig. 4e presents the cyclic voltammetry (CV) curves of CGNa0.1 during the first three cycles. A slight irreversible capacity near 0.5 V is observed in the initial cycle, while the subsequent cycles show well-overlapped profiles, indicating good electrochemical reversibility and cycling stability.

To further investigate the cause of capacity degradation of CG0 under high current densities, cycling tests were performed at 200 and 500 mA g⁻¹ for 100 cycles, as shown in Fig. 4f. The severe capacity fading of CG0 is primarily attributed to structural degradation and surface cracking, which was confirmed by post-cycling SEM observations. Fig. 4g shows that CGNa0.1 maintains 86.3% of its capacity after 800 cycles at 200 mA g⁻¹, demonstrating excellent long-term cycling stability. Moreover, we investigated the effects of precursor mixing ratios and intrinsic mineral components on sodium storage performance (Fig. 4h, S7c and d†). It was observed that, in addition to CG0 (8:1:1) and CCG0 (8:1:1), both CGx (8:1:1) and CCGx (8:1:1) exhibited a distinct “gentle slope” in the charge profile between 0.5–0.9 V, similar to CGNa0.1. However, this feature was absent in CG (90:5:5), suggesting that the 8:1:1 slurry ratio improves electrode conductivity and structural integrity, thereby facilitating Na⁺ insertion and increasing capacity. When the discharge capacity exceeds a threshold (∼400 mA h g⁻¹), the formation of C–Na–O interactions is promoted, which benefits the reversibility of desodiation. Nevertheless, the increased proportion of conductive agent and binder leads to a significant reduction in initial coulombic efficiency. In contrast, CGNa0.1, with the introduction of an external sodium source, effectively lowers the capacity threshold for C–Na–O formation, which improves the initial efficiency despite a slight capacity loss. Furthermore, as shown in Fig. 3a and S8a and b†, minerals such as kaolinite and pyrite in raw coal are converted into electrochemically active iron oxides and SiO₂ during pyrolysis, while no obvious impurity peaks are detected in acid-treated CCG0, indicating effective mineral removal. Compared to CG0, acid treatment improves the initial coulombic efficiency from 66.6% to 72.0%, but reduces the reversible capacity from 255.1 to 244.5 mA h g⁻¹. Although these mineral-derived phases can contribute to capacity, they primarily induce irreversible reactions. Overall, CGNa0.1 achieves the best comprehensive performance among all samples by balancing capacity and efficiency without relying on environmentally harmful acid treatments or complex recovery processes.

The Na⁺ storage behavior of coal-derived hard carbon was investigated through kinetic analysis using cyclic voltammetry (CV) curves of CG0, CG, and CGNa0.1 at various scan rates (Fig. 5a, S9a and b†). The relationship between peak current (i) and scan rate (v) was evaluated using the equation i = av^b, where a and b are constants related to the charge storage mechanism.²⁸ The value of b was obtained by plotting log(i) versus log(v); b = 0.5 corresponds to a diffusion-controlled process, while b = 1.0 indicates a capacitive-dominated behavior,²⁹ as shown in Fig. 5b, S9c and d.† In the plateau region, the b values for CG0, CG, and CGNa0.1 were 0.530, 0.422, and 0.358, respectively, confirming that Na⁺ storage in this region is predominantly diffusion-controlled. In contrast, the sloping regions exhibited b values of 0.924, 0.941, and 0.913, respectively, suggesting that capacitive contributions dominate in these voltage ranges.³⁰ Furthermore, with increasing scan rates, CGNa0.1 exhibited the highest fitting correlation (R² = 0.9852) for the reduction peak P, indicating superior kinetic stability and consistent Na⁺ storage behavior under varying scan rates.

In in situ Raman spectra (Fig. 5c), as Na⁺ enters the hard carbons, it affects the intensity and position of the G and D bands.^30,31 From open-circuit voltage to 0.5 V, the positions of the G and D bands remain nearly constant, suggesting that Na⁺ exhibits a high diffusion coefficient, corresponding to the adsorption stage within the sloping voltage region. As the battery is discharged down to 0.2 V, the D band peak position begins to shift, and a new peak emerges between the D and G bands, which is likely related to side reactions such as electrolyte decomposition and solvation effects occurring during the discharge process.³² With continued discharge, the D band intensity gradually diminishes and eventually disappears, indicating the filling of defect sites in the electrode material by Na⁺. During charging, the newly emerged peak between the D and G bands progressively weakens, suggesting that the side reactions induced during discharge are at least partially reversible; meanwhile, the D and G band intensities show partial recovery, further confirming the reversible nature of Na⁺ storage behavior in the hard carbon material.

Electrochemical impedance spectroscopy (EIS) was employed to further investigate the electrochemical kinetics of CG0, CG, and CGNa0.1 electrodes (Fig. 5d). The equivalent circuit model used for fitting (Fig. S10†) includes multiple resistive and capacitive components. The high-frequency intercept represents the internal resistance (R_S), encompassing the ohmic resistance of the separator, electrode, and current collector. The first semicircle corresponds to the charge transfer resistance (R_ct) and the double-layer capacitance (C_dl) at the electrode/electrolyte interface. The second semicircle reflects surface film resistance (RSEI) and its associated capacitance (CSEI), related to SEI layer formation. The low-frequency Warburg element (W) represents the diffusion resistance of Na⁺ ions in the bulk of the active material.³³ The extracted parameters (R_S, R_ct, and W) for all three samples are summarized in the table within Fig. 5d. Notably, CGNa0.1 exhibited the lowest R_S value, indicating excellent electronic conductivity. Additionally, its R_ct was also the lowest among the samples, suggesting faster charge transfer kinetics and more favorable interfacial electrochemical reactions.

The Na⁺ storage mechanism of CGNa0.1 was further elucidated via the galvanostatic intermittent titration technique (GITT), as shown in Fig. 5e. The Na⁺ diffusion coefficient (D) was calculated using the following equation.³⁰

	(2)

In the formula, τ denotes the pulse duration, M_B represents the molar ratio of carbon, S signifies the electrode surface area, m_B denotes the mass of the active material, V_M represents the molar volume of carbon, ΔE_τ signifies the instantaneous total voltage of a constant current cell at different time points, and ΔE_S represents the difference between stable potentials. Based on the GITT results in Fig. 5e, the Na⁺ storage process of CGNa0.1 can be divided into three distinct voltage regions: 0.8–0.6 V corresponding to Na⁺ adsorption on surface sites, 0.6–0.1 V attributed to intercalation into graphitic layers, and <0.1 V representing Na⁺ filling into closed pores and nanovoids.

To further investigate the electrochemical kinetics of Na⁺ in CGNa0.1 during cycling, galvanostatic charge–discharge tests were conducted at a current density of 200 mA g⁻¹. EIS measurements were performed at the 10th, 59th, and 258th cycles, and the Na⁺ diffusion coefficients were calculated using eqn (3) and (4).³⁴ The corresponding results are presented in Fig. 6. Fig. 6a presents the reversible capacity of CGNa0.1 over 258 cycles at 200 mA g⁻¹ and 60 cycles at 500 mA g⁻¹. Overall, the material exhibits good cycling stability, although some fluctuation appears after the EIS2 measurement. As shown in Fig. 6b–d, from the 10th to the 59th cycle, R_SEI and R_ct decrease significantly while R_S remains relatively stable, indicating a steady activation process with minimal capacity fluctuation. However, from the 59th to the 258th cycle, R_ct increases from 2.02 Ω to 3.04 Ω, and the R_SEI fitting signal weakens, with both parameters showing instability—likely contributing to the increased capacity fluctuation at later stages. Nevertheless, the Na⁺ diffusion coefficient increases from 10⁻¹⁰ cm² s⁻¹ to 10⁻⁹ cm² s⁻¹ and remains stable, indicating consistently favorable Na⁺ diffusion kinetics throughout the cycling process.

DNa+ = 0.5[(RT/(An2F2σωC))]2	(3)

Z′ = Rtotal + σωω−1⁄2	(4)

where, R is the universal gas constant, T is set to 298 K, A represents the electrode area (1.13 cm²), and n is taken as 0.21 based on the charge transfer analysis from DFT calculations.

To further assess the practical electrochemical performance of CGNa0.1, a coin-type full cell was assembled using commercially available O3-type layered sodium oxide (O3-NFM-Na) as the cathode and CGNa0.1 as the anode. As shown in Fig. 7b, a stepwise formation protocol was employed to activate both electrodes: the cell was initially charged to 3.0 V at 20 mA g⁻¹, followed by charging to 3.4 V at 50 mA g⁻¹, and finally to 4.0 V at 100 mA g⁻¹. This gradual activation strategy enables full utilization of electrode materials and promotes stable cycling performance. The cycling stability of the O3-NFM-Na//CGNa0.1 full cell at 20 mA g⁻¹ is presented in Fig. 7c. After 40 cycles, the cell maintained a stable charge capacity of 2.34 mA h, slightly exceeding the capacity recorded in the first cycle post-formation, suggesting improved electrode compatibility and enhanced sodium-ion transport kinetics following full activation. Fig. 7d illustrates the rate capability of the full cell under various current densities ranging from 20 mA g⁻¹ to 200 mA g⁻¹. The cell exhibited a slight capacity increase at moderate rates, followed by a gradual decline at higher current densities. Notably, when the current density was decreased from 200 mA g⁻¹ to 40 mA g⁻¹, the charge capacity recovered significantly—from 0.54 mA h to 1.47 mA h—demonstrating excellent rate reversibility and structural stability. These results collectively highlight the strong compatibility, kinetic resilience, and practical application potential of the O3-NFM-Na//CGNa0.1 full cell configuration for sodium-ion energy storage systems.

3.3 C–Na–O interactions enhanced Na⁺ diffusion mechanism

The adsorption behavior and interaction characteristics of Na on the carbon surface were investigated, and the corresponding results are presented in Fig. 8. Literature reports indicate that the most energetically favorable adsorption site for Na lies directly above the center of the carbon ring (Top site),³⁵ consistent with the Top site illustrated in Fig. 8. Upon oxygen incorporation, Na is adsorbed via synergistic interactions with carbon and oxygen atoms, leading to the formation of a C–Na–O configuration. To assess the thermodynamic stability of different configurations, formation energies were calculated and summarized in Table 2. The formation energies of the C⋯Na–O and C–Na–O configurations are −3.56 eV and −3.61 eV, respectively, both lower than that of the Top site, suggesting that the C–Na–O configurations are thermodynamically more stable and thus more favorable for Na adsorption. These findings are in good agreement with previously reported studies.

---

The electronic characteristics of the C–Na–O structure were further analyzed, with detailed results presented in Table 3 and Fig. S11.† As shown in Fig. S11,† the O and C atoms carry Mulliken charges of −0.58e and −0.51e, respectively. Upon Na adsorption at the Top site, O and C gain 0.04e and 0.03e in Mulliken charge, respectively, suggesting that their presence influences the Na adsorption behavior at the carbon ring center. Further taking into account the interaction between C, O, and Na, in the C⋯Na–O configuration, Na donates 0.07e, O accepts 0.21e, and C donates 0.01e. Because the influence of C on Na is relatively weak, this configuration results in a net charge accumulation of 0.13e. In comparison, in the C–Na–O configuration, Na donates 0.19e, and O and C gain 0.19e and 0.07e, respectively, leading to an overall charge-neutral configuration with minimal disturbance to the surrounding environment. Moreover, to further elucidate the electronic properties of the structure, we calculated the partial density of states (PDOS) of Na, as shown in Fig. 9. The presence of pronounced free electron states near the Fermi level indicates the metallic nature of Na in the structure,³⁵ which is particularly prominent in the C–Na–O configuration. This metallic characteristic enhances the electronic conductivity of the material,³⁶ thereby improving its cycling stability. Therefore, the C–Na–O configuration exhibits superior electronic stability, which is expected to better promote Na storage and diffusion.

---

The diffusion behavior of Na within the carbon matrix was categorized into three distinct modes: surface diffusion, edge diffusion, and bulk diffusion. To gain deeper insight, two representative diffusion pathways extending from the surface to the interior were defined, as illustrated in Fig. 10a. In Pathway 1, Na diffuses via the C⋯Na–O configuration, whereas in Pathway 2, diffusion occurs through the C–Na–O configuration. The corresponding energy barriers for two pathways are also presented in Fig. 10a. During surface diffusion, the energy barriers for two pathways are calculated to be 0.93 eV and 0.75 eV, respectively, suggesting that Na can spontaneously form both C⋯Na–O and C–Na–O configurations on the Carbon surface, with the latter offering a lower diffusion barrier and thus promoting more favorable surface transport. In the case of edge diffusion, both configurations encounter higher diffusion barriers, and the C–Na–O pathway, being more strongly influenced by surrounding C atom, exhibits a significantly higher energy barrier than the C⋯Na–O configuration. In comparison, bulk diffusion remains energetically unfavorable, with energy barriers of 1.71 eV in Pathway 1 and 1.75 eV in Pathway 2, indicating that disruption of the C⋯Na–O and C–Na–O structures requires external perturbation due to their high intrinsic stability.

To further evaluate the effect of Na⁺ doping, Na⁺ derived from Na₂CO₃ were introduced to construct C⋯Na1⁺—O and C–Na2⁺—O configurations, as shown in Fig. 10b. The promoted bulk diffusion barriers were then recalculated. As a result, the barrier in Pathway 1 decreased significantly from 1.71 eV to 1.05 eV, which can be attributed to the relatively weak influence of C on the Na1⁺ in the C⋯Na1⁺—O configuration, while the diffusing Na still experiences C interaction, resulting in a slightly higher barrier. In contrast, the C–Na2⁺—O structure maintains charge neutrality and causes minimal perturbation to the surrounding framework, and the diffusing Na experiences weaker C interaction, leading to an even lower diffusion barrier of only 0.11 eV in Pathway 2. To verify whether the optimized diffusion energy barrier of 0.11 eV is consistent with the experimentally fitted sodium-ion diffusion coefficients, eqn (5) was used to calculate the corresponding diffusion coefficient. The result was 1.38 × 10⁻¹⁰ cm² s⁻¹, which was in good agreement with the experimental data fitted from EIS1 (2.25 × 10⁻¹⁰ cm² s⁻¹), further reinforcing the correlation between experimental observations and theoretical calculations.

D = D0e(−((Eb/(kBT))))	(5)

where, D₀ is set to 1.0 × 10⁻⁸. Furthermore, as shown in Fig. 10c, both Na1⁺ and Na2⁺ exhibit a tendency to dissociate during the diffusion process, with Na1⁺ exhibiting a more pronounced detachment behavior, as evidenced by a Z-coordinate difference of up to 1.03. These findings demonstrate that the introduction of the C–Na–O configuration significantly enhances Na⁺ diffusion kinetics within the carbon framework, while the C⋯Na–O structure tends to induce desodiation of adsorbed Na during diffusion, thereby contributing to an improved initial coulombic efficiency.

4. Conclusions

In this study, a green and acid-free dual-regulation strategy was proposed by co-pyrolyzing Xinjiang bituminous coal with glucose and sodium carbonate to construct high-performance coal-derived hard carbon anodes for sodium-ion batteries. This approach enables the synergistic regulation of multiscale structural parameters, including pore structure, interlayer spacing, surface functionalities, and in situ Na⁺ doping. The combined action of glucose and Na₂CO₃ effectively promotes carbon structural ordering, expansion of interlayer spacing, and enrichment of closed pores. The optimized sample, CGNa0.1, delivers excellent electrochemical performance, with a high reversible capacity of 299.8 mA h g⁻¹, an initial coulombic efficiency of 90.46%, and a capacity retention of 217 mA h g⁻¹ after 800 cycles at 0.2 A g⁻¹. In situ Raman spectroscopy reveals that Na storage during discharge leads to defect filling (as evidenced by the gradual disappearance of the D band), while electrolyte decomposition and related side reactions introduce new peaks between the D and G bands. Nevertheless, the overall sodium storage behavior of hard carbon is highly reversible. Density functional theory (DFT) calculations indicate that the synergistic interaction between carbon and oxygen facilitates Na adsorption, with the C⋯Na–O and C–Na–O configurations exhibiting low formation energies of −3.56 eV and −3.61 eV, respectively. Furthermore, the C⋯Na–O configuration assists in the desodiation of adsorbed Na during diffusion, thereby enhancing the initial coulombic efficiency, while the C–Na–O configuration effectively promotes Na⁺ diffusion kinetics. This work provides a scalable and sustainable strategy for designing high-performance coal-based hard carbon anodes and offers new mechanistic insights into sodium storage behavior.

Data availability

The data that support the findings of this study are available from the corresponding author Prof. Dong upon reasonable request.

Author contributions

Tianqi Xu: data curation, investigation, methodology, software, writing – original draft. Zongxu Yao: formal analysis, investigation. Wei Jiang: data curation, investigation. Yaxin Chen: supervision. Chenmin Liu: investigation, writing – review & editing. Yinshuang Guan: investigation. Zhiqiang Tang: investigation. Liang Dong: supervision, project administration, methodology, funding acquisition.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

The work was supported by the Jiangsu Key Laboratory for Clean Utilization of Carbon Resources Research Project (BM2024007) and the Fundamental Research Funds for the Central Universities, the Graduate Innovation Program of China University of Mining and Technology and the Postgraduate Research & Practice Innovation Program of Jiangsu Province (KYCX24_2892, KYCX23_2817).

Notes and references

1. X. Lin, S. Zhang, M. Yang, B. Xiao, Y. Zhao, J. Luo, J. Fu, C. Wang, X. Li, W. Li, F. Yang, H. Duan, J. Liang, B. Fu, H. Abdolvand, J. Guo, G. King and X. Sun, Nat. Mater., 2025, 24, 83–91 CrossRef CAS PubMed .
2. A. Yao, S. M. Benson and W. C. Chueh, Nat. Energy, 2025, 10, 404–416 CrossRef .
3. C. Wu, Y. Yang, Y. Zhang, H. Xu, X. He, X. Wu and S. Chou, Chem. Sci., 2024, 15, 6244–6268 RSC .
4. X.-X. He, L. Li, X. Wu and S.-L. Chou, Adv. Mater., 2025, 2506066 CrossRef .
5. Z. Li, Y. Shang, Z. Zhu, J. Chang and K. Zhao, Clean Coal Technol., 2024, 30, 99–117 CAS .
6. R. Ma, Y. Chen, Q. Li, B. Zhang, F. Chen, C. Leng, D. Jia, N. Guo and L. Wang, Chem. Eng. J., 2024, 493, 152389 CrossRef CAS .
7. X. Y. Wang, K. Y. Zhang, M. Y. Su, H. H. Liu, Z. Y. Gu, D. Dai, B. Li, J. W. Wang, X. Y. He and X. L. Wu, Carbon, 2024, 229, 119526 CrossRef CAS .
8. A. O. Aboyade, M. Carrier, E. L. Meyer, H. Knoetze and J. F. Görgens, Energy Convers. Manage., 2013, 65, 198–207 CrossRef CAS .
9. L. Ding, Y. Zhang, Z. Wang, J. Huang and Y. Fang, Bioresour. Technol., 2014, 173, 11–20 CrossRef CAS PubMed .
10. Z. Wu, W. Yang and B. Yang, Bioresour. Technol., 2018, 249, 501–509 CrossRef CAS PubMed .
11. H. Chen, N. Sun, Q. Zhu, R. A. Soomro and B. Xu, Adv. Sci., 2022, 9, 2200023 CrossRef CAS .
12. Z. Zhou, Z. Wang and L. Fan, Chem. Eng. J., 2024, 490, 151428 CrossRef CAS .
13. S. Lin, H. Zhang, C. Shu, W. Hua, X. Wang, Y. Zhao, J. Luo, Z. Tang, Y. Wu and W. Tang, Adv. Funct. Mater., 2024, 34, 2409628 CrossRef CAS .
14. J. Hu, L. Xu, X. Li, Q. Liang, C. Ding, Y. Li, Y. Liu and Y. Gao, J. Mater. Chem. A, 2025, 13, 3206–3235 RSC .
15. L. Xu, H. Liu, D. Zhao, Q. Cao, J. Gao and S. Wu, Int. J. Energy Res., 2018, 42, 4131–4141 CrossRef CAS .
16. S. Valluri and S. K. Kawatra, Fuel Process. Technol., 2021, 212, 106620 CrossRef CAS .
17. J. Liu and H. Teng, Guangdong Chem. Ind., 2019, 46, 12–13 Search PubMed .
18. T. Yang, X. Yang, L. Wei, X. Kai, L. Guo, R. Li and J. China, Coal Soc., 2020, 45, 3284–3292 Search PubMed .
19. Y. Zhao, Z. Yuan, R. Zhang, S. Wu, Y. Shen, L. Wang and Z. Lou, Environ. Sci. Pollut. Res., 2023, 30, 4435–4447 CrossRef CAS PubMed .
20. C. Bai, L. Liu and H. Sun, J. Phys. Chem. C, 2012, 116, 7029–7039 CrossRef CAS .
21. G. Yan, T. Xu, M. Hu, Y. Xi and C. Lan, Adv. Theory Simul., 2023, 6, 2200724 CrossRef CAS .
22. D. Sergeev, E. Yazhenskikh, D. Kobertz and M. Müller, Calphad, 2019, 65, 42–49 CrossRef CAS .
23. W. Li, D. Peng, W. Huang, X. Zhang, Z. Hou, W. Zhang, B. Lin and Z. Xing, Carbon, 2023, 204, 315–324 CrossRef CAS .
24. L. Cheng, C. Ma, W. Lu, X. Wang, H. Yue, D. Zhang and Z. Xing, Chem. Eng. J., 2022, 433, 133527 CrossRef CAS .
25. Y. Wang, Y. Zhao, R. He, H. Zhou, N. Li, Y. Teng, Y. Song and Q. Liu, J. China Coal Soc., 2023, 48, 1736–1746 CAS .
26. A. Gomez-Martin, J. Martinez-Fernandez, M. Ruttert, M. Winter, T. Placke and J. Ramirez-Rico, Chem. Mater., 2019, 31, 7288–7299 CrossRef CAS .
27. Q. Jia, Z. Li, H. Ruan, D. Luo, J. Wang, Z. Ding and L. Chen, Molecules, 2024, 29, 4331 CrossRef CAS PubMed .
28. Q. Ren, J. Wang, L. Yan, W. Lv, F. Zhang, L. Zhang, B. Liu and Z. Shi, Chem. Eng. J., 2021, 425, 131656 CrossRef CAS .
29. P. Lu, Y. Sun, H. Xiang, X. Liang and Y. Yu, Adv. Energy Mater., 2018, 8, 1702434 CrossRef .
30. G. Zhao, T. Xu, Y. Zhao, Z. Yi, L. Xie, F. Su, Z. Yao, X. Zhao, J. Zhang, W. Xie, X. Li, L. Dong and C. M. Chen, Energy Storage Mater., 2024, 67, 103282 CrossRef .
31. X. Cheng, H. Li, Z. Zhao, Y. Wang and X. Wang, New Carbon Mater., 2021, 36, 93–105 CrossRef CAS .
32. Y. Gu, E. M. You, J. D. Lin, J. H. Wang, S. H. Luo, R. Y. Zhou, C. J. Zhang, J. L. Yao, H. Y. Li, G. Li, W. W. Wang, Y. Qiao, J. W. Yan, D. Y. Wu, G. K. Liu, L. Zhang, J. F. Li, R. Xu, Z. Q. Tian, Y. Cui and B. W. Mao, Nat. Commun., 2023, 14, 3536 CrossRef CAS PubMed .
33. Y. Xu, E. Swaans, S. Basak, H. W. Zandbergen, D. M. Borsa and F. M. Mulder, Adv. Energy Mater., 2016, 6, 1501436 CrossRef .
34. J. Li, T. Ouyang, L. Liu, S. Jiang, Y. Huang and M.-S. Balogun, J. Energy Chem., 2024, 93, 368–376 CrossRef CAS .
35. Z. Wang, X. Feng, Y. Bai, H. Yang, R. Dong, X. Wang, H. Xu, Q. Wang, H. Li, H. Gao and C. Wu, Adv. Energy Mater., 2021, 11, 2003854 CrossRef CAS .
36. Z. Khanam, T. Xiong, F. Yang, H. Su, L. Luo, J. Li, M. Koroma, B. Zhou, M. Mushtaq, Y. Huang, T. Ouyang and M.-S. Balogun, Small, 2024, 20, 2311773 CrossRef CAS PubMed .

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d5ta03632a

This journal is © The Royal Society of Chemistry 2025