Introduction

The increasing global concerns regarding environmental issues, coupled with the rapid expansion of both the population and industry, have given rise to a pressing demand for freshwater resources1,2. This has, in turn, triggered an explosive surge in research focused on advanced materials for desalination technologies, which is fueled by the abundant volume and widespread availability of seawater and brackish water3,4. Capacitive deionization (CDI), in which the migration of charged ions and their subsequent storage in the electric double-layer (EDL) formed on an electrode surface are guided, has emerged as a promising solution for desalinating low-salinity water because of its notable advantages, including low energy consumption, scalability, high water utilization efficiency, and minimal secondary environmental impact5,6,7. Porous carbonaceous electrodes, including activated carbon (AC), carbon aerogels, graphene, and metal-organic framework-derived carbon, have garnered substantial attention for their efficacy in desalination because of their abundant resource availability and cost effectiveness8,9. Unfortunately, the efficacy of these carbonaceous electrodes is impeded by inadequate conductivity and poor electrochemical stability, leading to limited adsorption capacity and degraded cycling life10,11,12.

Among the various modification methods used to increase the performance and electrochemical stability of carbonaceous materials in desalination processes, heteroatom doping has garnered widespread attention for several compelling reasons: (1) dopants can create additional active sites, introduce defects, and enhance the electrical conductivity; (2) heteroatom doping significantly improves the structural integrity and stability of carbonaceous materials, extending their operational lifespan; (3) this doping allows precise tuning of the surface chemistry, thereby improving interactions with specific ions or molecules; and (4) certain heteroatoms, such as oxygen and nitrogen, can increase the hydrophilicity of carbonaceous materials13,14,15,16. In particular, nitrogen (N), often referred to as the doping “star,” stands out as the most effective and feasible heteroatom for improving the electrochemical performance of carbonaceous materials8,17,18,19. In addition, in our pioneering research conducted in 2015, we revealed that nitrogen doping markedly enhances the electrochemical and desalination performance of reduced graphene oxide (rGO), and this groundbreaking discovery established the foundational principles for utilizing heteroatom doping to significantly improve the CDI performance20. As shown in Fig. 1 (with detailed sample information provided in Supplementary Table 1), the introduction of N into a carbonaceous framework (producing N-doped carbon (NC)) can indeed induce a high salt adsorption capacity (SAC, Fig. 1a), improve the cycling life (Fig. 1b), and enhance the capacity retention (Fig. 1c) in CDI desalination.

Fig. 1: Performance comparison of NC electrodes reported in the literatures for CDI.
figure 1

a SAC, (b) cycle number, and (c) corresponding capacity retention (S1: N-doped graphitic porous carbon; S2: N-doped porous carbon composite; S3: N-enriched micro-mesoporous carbon; S4: N, S and O co-doped porous carbon; S5: N-doped rod-like porous carbon; S6: N-doped graphene; S7: N-doped layered mesoporous carbon; S8: N-doped carbon nanorods; S9: N, P co-doped hierarchical porous carbon; S10: N, S co-doped carbon; S11: N, B co-doped carbon; S12: N-doped porous carbon; S13: N-doped hierarchical porous carbon; S14: N-doped porous carbon; S15: N, P co-doped porous carbon; S16: N-doped graphene quantum dot; S17: N-doped carbon nanotubes; S18: N, P, S co-doped hollow carbon; S19: N-doped carbon hollow shells S20: N, P co-doped hierarchical carbon).

Full size image

Based on a literature review, the increased electronic conductivity due to the presence of graphitic N (N-Q) is considered to be the main reason for the modified desalination performance of NC electrodes, in which the graphitic N atoms are integrated into the carbon lattice in a manner that preserves the sp2 hybridization, thus maintaining and further enhancing the conductive network21,22,23. In addition to graphitic N, other N configurations, including pyrrolic N (N-5) and pyridinic N (N-6), also play significant roles in enhancing the desalination performance by enhancing the surface reactivity, improving the hydrophilicity, facilitating charge transfer, and stabilizing the carbon structure24,25. Although the N configurations in carbon electrodes are generally challenging to control, some pioneering works have demonstrated that adjustment of these configurations can significantly improve the electrochemical performance. For example, Ma’s study demonstrated that the incorporation of phytic acid into carbonaceous electrodes not only alters the nitrogen configurations, particularly increasing the N-6 concentration but also enhances the ion adsorption capacity and reduces the energy barrier for ion migration during electrochemical cycling26. Similarly, Zhang’s study emphasized that N-5 and N-6 enrichment through the controlled synthesis of carbon electrodes significantly boosts the specific capacitance, highlighting the crucial impact of the nitrogen configurations24.

With respect to CDI, although enhanced desalination performance has been reported for NC electrodes, the underlying mechanisms, particularly concerning the N configurations, have rarely been discussed in detail. Furthermore, the inherent complexity of the precursors used to produce NCs must also be considered. Regardless of the precursor, whether it is rGO, carbon nanotubes, biomass, metal/covalent-organic frameworks, or conductive polymers, oxygen is invariably present to some degree. However, the potential influence of these trace oxygen atoms on the desalination capabilities of NCs has not been extensively studied. More crucially, comprehending how these oxygen atoms affect various nitrogen configurations is vital to obtaining a comprehensive understanding of the mechanisms underlying the enhanced CDI performance of NC electrodes.

In this work, we engineered an NC nanosheet with trace oxygen co-doping (ONC-S), utilizing guanine as the precursor, to delve into the role of oxygen in modifying NCs to enhance their desalination performance. The controlled pyrolysis of guanine not only yields abundant N-5/N-6 configurations but also introduces a precise amount of oxygen, thereby providing a robust model for our comprehensive analysis. Remarkably, the introduction of a small amount of oxygen into an NC electrode enriched with N-5/N-6 configurations profoundly affects the charge distribution, wettability, and ion diffusion, thereby significantly improving the reaction kinetics and desalination performance. Furthermore, the interplay among the compositional tailoring, physical/chemical properties, electrochemical activity, and desalination performance of the target ONC-S was elucidated through material characterization, electrochemical testing, and theoretical calculations, revealing the significant role of oxygen in enhancing the CDI performance of NC electrodes.

Results

Material preparation and characterization

To address the challenge of fabricating target ONC-S with specific N configuration enrichment and simultaneous trace O co-doping, the selection of the precursor is crucial. Herein, we utilized guanine as the starting material to enrich the N-5 and N-6 configurations in the final sample owing to its distinctive nitrogen sites located in both five-membered and six-membered rings (Fig. 2a)27. In addition, the trace amount of oxygen present in guanine fulfills our experimental design of trace O co-doping to obtain the target ONC-S28. Owing to the presence of carbonyl and amino groups, guanine molecules are adept at establishing an enhanced hydrogen bond network in the solid state through these functional moieties, which promotes a layered arrangement of the molecules, as shown in Supplementary Fig. 129. Specifically, the carbonyl group offers a potent site for hydrogen bond acceptance, while the amino group effectively contributes as a hydrogen bond donor28. For comparative analysis, adenine was also utilized as a precursor to prepare an NC that notably lacks O co-doping30. The lack of a carbonyl group in adenine tends to lead to the formation of a bulk structure rather than a layered structure, resulting in the formation of NC grains after pyrolysis (NC-G)31.

Fig. 2: Synthesis and characterizations for ONC-S and NC-G.
figure 2

a Schematic illustration of the preparation of the target ONC-S and NC-G; TEM images with different magnifications of (b, c) ONC-S and (d, e) NC-G; (f) elemental distribution maps of ONC-S. g Raman spectra, (h) N2 adsorption/desorption isotherms, and (i) pore size distributions of ONC-S and NC-G.

Full size image

Initially, to analyze the morphological disparities between ONC-S and NC-G, transmission electron microscopy (TEM) was utilized for characterization. As depicted in Fig. 2b, c, ONC-S has a curled sheet morphology, accentuated by its nearly transparent sheets that underscore its ultrathin nature27,29. Conversely, the bulk NC-G derived from adenine presents an irregular and granular form (Fig. 2d, e). Furthermore, the elemental distribution of ONC-S (Fig. 2f) confirms the uniform dispersion of O, N, and C, thus validating the efficacy of the employed self-templating and in situ doping strategy with guanine32.

The structural characteristics of NC-G and ONC-S were comprehensively investigated via X-ray diffraction (XRD) and Raman spectroscopy. As depicted in Supplementary Fig. 2, both materials exhibit similar characteristics in their XRD patterns, with prominent peaks at ~ 26° (002 plane) and 43° (101 plane), which are indicative of a carbonaceous material18. This observation corroborates the successful conversion of guanine and adenine into carbonaceous entities, as confirmed through comparative analysis with the XRD patterns of guanine and adenine (Supplementary Fig. 3). Furthermore, Raman spectral analysis of NC-G and ONC-S (Fig. 2g) reveal distinct peaks at 1356 and 1578 cm−1, corresponding to the characteristic D band (associated with disorder and defects) and G band (representing the vibration of sp2 hybridized carbon atoms) of carbonaceous materials, respectively33. As depicted, the fitted Raman spectra clearly demonstrate that the area of the D band is noticeably larger than that of the G band. Moreover, the area ratios of the D band to the G band (AD/AG) are calculated to be 2.66 and 1.69 for ONC-S and NC-G, respectively, indicating that the simultaneous incorporation of O and N into the carbonaceous frameworks leads to a significant increase in defects34. In general, the increased number of defects resulting from O and N dual doping offers many active sites for ion adsorption35. As a result, the charge transfer processes involved in desalination mechanisms are expedited, ultimately facilitating more efficient desalination33.

Moreover, owing to the pronounced disparities in the morphology and structure, the N2 adsorption/desorption isotherms (Fig. 2h) and pore size distributions (Fig. 2i) of ONC-S and NC-G also exhibit notable distinctions. The gradual increase in the amount of adsorbed N2 and the presence of a hysteresis loop observed for ONC-S imply the existence of pores with various sizes, which is further confirmed by the pore size distribution analysis36. As delineated in Supplementary Table 2, ONC-S (180.152 m2 g−1) demonstrates a substantially greater specific surface area (SSA) than NC-G (62.158 m2 g−1), consequently increasing the ion accessibility within the pores and mitigating the overlap of EDLs, and this augmentation ultimately enhances the electrochemical performance and facilitates desalination efficacy37.

N configuration and activity

The chemical compositions of the obtained ONC-S and NC-G were further elucidated through X-ray photoelectron spectroscopy (XPS). As shown in Fig. 3a, the pronounced N 1s and O 1s peaks evident in the survey spectrum of ONC-S confirm successful dual doping17. For NC-G, a discernible N 1s peak can also be detected, and notably, the relatively subdued O 1s peak observed in the spectrum of NC-G, compared to that of ONC-S, likely stems from adsorbed water10. Moreover, the elemental contents of ONC-S and NC-G are tabulated in Supplementary Table 3. The nitrogen contents of ONC-S and NC-G are as high as 27.82 and 20.05 at.%, respectively, with the oxygen content of ONC-S (4.99 at.%) surpassing that of NC-G (0.78 at.%), revealing high N doping contents in both ONC-S and NC-G and variations in the O content due to discrepancies in doping and adsorption35. Compared with their precursors, a reduction in the nitrogen content is observed, primarily due to the escape of gaseous products during the high-temperature pyrolysis process33. In addition, the local bonding environments of the nitrogen atoms in ONC-S were discerned through the high-resolution N 1s spectrum. As depicted in Fig. 3b, the deconvoluted peaks centered at approximately 298.2, 399.7, and 400.8 eV in the high-resolution N 1s spectrum of ONC-S correspond to N-5, N-6, and N-Q, respectively, consistent with previous studies27. The contents of N-5, N-6, and N-Q in ONC-S are calculated to be 32.44%, 37.53%, and 30.03% (Fig. 3c), respectively, indicating enrichment of N-5 and N-6 (69.97%). In contrast, Supplementary Fig. 4 shows similar deconvoluted peaks in the N 1s spectrum of NC-G, along with variations in the N configurations, showing low contents of N-5 (29.57%) and N-6 (30.56%) compared with ONC-S and revealing the positive effect of trace O doping on the enrichment of N-5 and N-6.

Fig. 3: Chemical properties and ion adsorption analysis of ONC-S and NC-G.
figure 3

a XPS survey spectra of NC-G and ONC-S; (b) N 1s XPS spectrum of ONC-S; (c) ratios of various nitrogen configurations; top view of simulated models illustrating Na adsorption and corresponding adsorption energies (Eads) at (d) N-5, (e) N-6, and (f) N-Q sites in the NC model; (g) N-5, (h) N-6, and (i) N-Q sites in the ONC model; and (j) graphene model; (k) comparison of adsorption energies of various structural models.

Full size image

To compare the activities of different nitrogen configurations, density functional theory (DFT) calculations were performed for geometry optimization and to evaluate the Eads. Generally, ions are well recognized to be more likely to adsorb at sites characterized by higher adsorption energies38. Considering both kinetic and thermodynamic aspects, sodium atoms tend to be adsorbed at the N-doping interface8,33. Note that a monolayer graphene model was utilized as the carbon substrate in our calculations. In this model, carbon atoms at various sites within the substrate were systematically substituted with nitrogen atoms to simulate nitrogen doping in diverse chemical environments39. As illustrated, the adsorption energies at the N-5, N-6, and N-Q sites are calculated to be −1.38, −1.23, and −0.86 eV, respectively (Fig. 3d–f), in the NC model. In comparison, the adsorption energies at N-5, N-6, and N-Q are −1.65, −1.38, and −0.98 eV, respectively (Fig. 3g–i), in the ONC model. Furthermore, compared to the low adsorption energy for sodium atoms on graphene (−0.51 eV, Fig. 3j), the presence of N-5, N-6, and N-Q substantially enhances the adsorption energies of the carbon substrate for sodium atoms (Fig. 3k), indicating the thermodynamic favorability of ion adsorption at the N-doping interface40. Notably, the N-5 and N-6 sites exhibit higher adsorption energies than N-Q, implying a preference for sodium atoms to adsorb at the N-5/N-6 sites owing to their increased sodium affinity24. More importantly, the trace O introduction further amplifies the adsorption energies for sodium at the N-5/N-6 sites, revealing the positive effect of O co-doping in the NC model35. For a detailed investigation, the side views of these computational models related to sodium adsorption are presented in Supplementary Fig. 5. The N-5 and N-6 sites are posited to introduce more dangling bonds and defects than the N-Q sites, which increases their sodium affinity25. Considering the relatively high contents of N-5 and N-6 in ONC-S (Fig. 3c), high electrochemical and desalination performance can be expected.

Electrochemical and desalination performance

The electrochemical characteristics of the as-synthesized ONC-S and NC-G electrodes were examined via a three-electrode system immersed in a 1.0 M NaCl solution. As depicted in Fig. 4a, the cyclic voltammetry (CV) curves of all three electrodes manifest nearly symmetric and quasi-rectangular shapes devoid of discernible redox peaks within the potential range of −1.0 to 0 V (vs. Ag/AgCl electrode), which are indicative of characteristic EDL capacitor (EDLC) behavior18,41,42. Notably, the enclosed area of the CV curve for the ONC-S electrode markedly surpasses that of NC-G, suggesting an increased specific capacitance10,43. A comparison of the galvanostatic charge/discharge (GCD) profiles at 1.0 A g−1, as illustrated in Fig. 4b, reveal that the two samples exhibit roughly triangular shapes, indicating typical EDLC behavior and ideal electrochemical reversibility36,44. Importantly, the ONC-S electrode demonstrates a longer discharge duration than its counterparts, underscoring its superior specific capacitance and showing high consistency with the CV results45,46. Specifically, the specific capacitance of ONC-S (298.01 F g−1) is much greater than that of NC-G (174.74 F g−1). Importantly, the GCD profiles change with increasing current density, with shorter durations at higher current densities due to increased concentration polarization, which is common in three-electrode tests (Fig. 4c and Supplementary Fig. 6)17,37. In addition to the decreased duration, the GCD profiles of ONC-S exhibit significant symmetry across all current densities for all samples, indicating remarkable electrochemical reversibility47. Supplementary Fig. 7 and Supplementary Table 4 present a comparison of the specific capacitances of ONC-S and NC-G across various current densities. This comparison clearly demonstrates that ONC-S consistently exhibits superior capacitance at all current densities, with values of 298.01, 264.96, 185.87, 152.85, and 139.41 F g−1 at current densities of 1.0, 2.0, 5.0, 8.0, and 10.0 A g−1, respectively. The superior performance of ONC-S highlights its excellent ion adsorption capability and suggests its high potential for efficient desalination48.

Fig. 4: Electrochemical performance and desalination ability.
figure 4

a CV curves at 10.0 mV s−1 and (b) GCD profiles at 1.0 A g−1 of NC-G and ONC-S; (c) GCD profiles of ONC-S at various current densities; (d) capacitive contributions of ONC-S at 40 and 50 mV s−1; (e) ex-situ Raman spectra of the ONC-S electrode upon charging/discharging. Desalination performance: (f) plots of the SAC dynamics over time and (g) Ragone plots of ONC-S and NC-G; (h) plots of the SAC dynamics over time at various voltages and (i) SAC values at different salt concentrations for ONC-S; (j) cycling performance of ONC-S at 1.2 V in a 500 mg L−1 NaCl solution; (k) initial and final SAC values during cycling. Error bars in (i) represent the standard deviation and were calculated on the basis of three experimental data points.

Full size image

The electrochemical dynamics of the ONC-S electrode were further investigated to gain deeper insights into both its charge storage mechanism and the impact of morphology/compositional adjustments on the pseudocapacitance. Generally, the relationship between the scan rate and current response at a fixed potential can be expressed as i = avb49. As illustrated in Fig. 4d and Supplementary Fig. 8, the ONC-S electrode exhibits 81% and 92% capacitive contributions to charge storage at scan rates of 40 and 50 mV s−1, respectively. Compared with NC-G (Supplementary Fig. 9), ONC-S displays greater capacitive contributions, which is attributable to its enhanced charge transfer kinetics and heightened sodium adsorption ability, thus underscoring the crucial role of oxygen doping in increasing the electrochemical activity24,41. Furthermore, the Nyquist plots shown in Supplementary Fig. 10 provide insights into the charge transfer discrepancy between NC-G and ONC-S. These fitted plots reveal that ONC-S (0.23 Ω) has a significantly smaller equivalent charge transfer resistance than NC-G (0.41 Ω). This reduction in resistance is attributed to the increased conductivity and convenient ion diffusion induced by the trace O doping in the NC35,37. The significant capacitive contribution of the ONC-S electrode, coupled with its low charge transfer resistance, facilitates rapid charge transfer at the interface between the electrode and feedwater, thus reducing electrochemical polarization and allowing full exploitation of its electrochemical activity47. Notably, a side effect caused by ion adsorption upon charging and discharging is the deformation of the electrode material, as initially reported by Soffer and coworkers in the 1970s50. Although such electrode deformation is generally minimal compared with that observed in batteries, it offers valuable insights into the underlying adsorption mechanisms51. In the case of our target ONC-S, copious nitrogen doping, coupled with oxygen substitution, promotes numerous defects and enhances the activity, and this combination is likely to intensify the physical stress and distortions on and near the electrode surface. Moreover, when combined with the charge redistribution resulting from ion adsorption, these deformations become measurable, particularly in the doped regions52. To elucidate these changes, DFT calculations were employed to analyze microstructural variations during the charging and discharging cycles. As depicted in Supplementary Fig. 11, the pristine structures undergo notable deformation due to ion adsorption, primarily at the doped sites, which corroborates the previously mentioned structural deformation. Notably, subsequent geometric optimizations, performed after the removal of adsorbed ions, reveal that the deformed structures cannot fully revert to their original form, thereby facilitating the observation of microstructural variations. As illustrated in Fig. 4e, the initial charging phase, characterized by rapid ion adsorption, triggers a significant reduction in the intensity of the D band53. This intensity gradually recovers during the charging process, which can be attributed to dynamic structural adjustments designed to mitigate the mechanical stress concentrated in the doped areas54. In contrast, the changes in the G band are less pronounced, reflecting a defect-dominated adsorption process. During the subsequent discharging phase, the accumulation of ions increases the mechanical stress, leading to a gradual decrease in the D band intensity, which further supports the conclusion that ion adsorption predominantly occurs in the doped areas during both charging and discharging.

The desalination performances of the synthesized NC-G and ONC-S were systematically assessed within a CDI cell complemented by an AC cathode. Detailed photos of the individual components and the fully assembled CDI device are presented in Supplementary Fig. 12. Notably, as depicted in Fig. 4f, ONC-S demonstrates superior SAC (35.0 mg g−1) compared with NC-G (31.6 mg g−1) when tested in a 500 mg L−1 NaCl solution under an applied voltage of 1.2 V. This enhanced performance of ONC-S highlights the beneficial impact of oxygen and nitrogen co-doping on the desalination efficacy. Notably, the CDI cells were specifically designed with an electrode area of 32 cm2 and a mass loading of approximately 80 mg, equivalent to a specific area loading of 2.5 mg cm−2. This loading density was strategically chosen to achieve a balance between a uniform distribution of the active material on the substrate and an optimized desalination efficiency41,55. Generally, lower mass loadings could increase the SAC by facilitating charge transfer and reducing ion diffusion paths56. Nevertheless, excessively low loadings are typically detrimental, as they lead to a reduced overall desalination performance due to the sparse amount of active material57. This issue is further complicated by nonuniform coatings at loadings below 1.0 mg cm−2, as shown in Supplementary Fig. 13. Conversely, higher mass loadings may hinder ion diffusion in thicker electrodes, thus impeding effective material utilization58. To comprehensively explore the influence of the mass loading on the desalination results, a series of experiments with various loadings, i.e., 1.0, 1.5, 2.0, 2.5, 3.0, 3.5, and 4.0 mg cm−2, were conducted. The results, illustrated in Supplementary Fig. 14, indicate that while the SAC decreases with increasing mass loading, the total desalination capacity concurrently increases. An optimal balance is identified at mass loadings between 2.0 and 2.5 mg cm−2, where both high SAC and high overall desalination efficiency are achieved, supporting the rationale behind selecting 2.5 mg cm−2 as the mass loading in our experiments. In addition, in the Ragone plot diagram (Fig. 4g), ONC-S occupies the upper right corner, confirming its superiority, with a higher desalination ability and the fastest mean salt adsorption rate (SAR). Specifically, the SARs for ONC-S and NC-G are 2.33 and 1.95 mg g−1 min−1, respectively, at 1.2 V in a 500 mg L−1 NaCl solution. In general, the layered structure and rich defects originating from O and N co-doping are believed to promote charge transfer in ONC-S, leading to a higher desalination rate33.

The impact of the applied voltage on the desalination performance was explored across different voltages ranging from 0.6 to 1.2 V (Fig. 4h and Supplementary Fig. 15). As illustrated, the ONC-S electrode demonstrates SACs of 6.7, 16.1, 26.2, and 35.0 mg g−1 at 0.6, 0.8, 1.0, and 1.2 V, respectively, indicating a dependency of the SAC on the applied voltage49,57. Compared with NC-G, ONC-S consistently exhibits superior desalination capacities across all voltages (Supplementary Table 5), further highlighting the advantages obtained from its layered structure and O/N co-doping8,24. In addition, the effect of the salt concentration on the SAC was further investigated across various NaCl concentrations (250 to 1000 mg L−1), revealing that ONC-S achieves higher SACs of 27.6, 35.0, and 40.3 mg g−1 at 250, 500, and 1000 mg L−1, respectively (Fig. 4i, Supplementary Fig. 16, and Supplementary Table 6). A further increase in the concentration to 2000 mg L−1 increases the SAC to 42.2 mg g−1 because of the increased ionic strength at high NaCl concentrations (Supplementary Fig. 17)4. Notably, the ONC-S electrode exhibits superior desalination performance even in low-concentration feedwater solutions, underscoring its versatility and broad practical applications.

Furthermore, the long-term cycling stability of the ONC-S electrode, also known as the regeneration performance, was evaluated in a 500 mg L−1 NaCl solution with an applied voltage of 1.2 V (Fig. 4j), and NC-G was investigated under the same conditions (Supplementary Fig. 18). As anticipated, ONC-S demonstrates exceptional cycling stability without any noticeable performance degradation over 50 cycles. The SAC remains at 33.6 mg g−1, with an impressive SAC retention of 96.0% (Fig. 4k). In contrast, the NC-G electrode shows significant degradation, exhibiting a low SAC retention of 68.0% after 50 cycles. In addition to the cycling stability, the mean SAR value over 50 cycles (as shown in Supplementary Fig. 19) further demonstrates the superiority of ONC-S over NC-G. Furthermore, after 50 cycles, the nitrogen configurations in the ONC-S electrode (Supplementary Fig. 20) exhibit minimal changes, confirming their stability throughout cycling24,25. In summary, trace O co-doping, combined with abundant N-5 and N-6 configurations, increases the electrochemical activity of the target ONC-S in terms of the specific capacitance in three-electrode tests and the SAC in CDI, which originates from the optimized chemical/physical properties obtained from the rational design.

Enhancement mechanism

To further substantiate the effectiveness of compositional tailoring of carbonaceous frameworks, the conductivities and charge distributions of the as-prepared samples were comparatively analyzed. As illustrated in Supplementary Fig. 21, relative to the total density of states (TDOS) of graphene, both the NC and ONC with N-5 models demonstrate notable enhancements in the density of states near the Fermi level, indicative of an improved electronic conductivity59. Moreover, a deeper examination reveals that the incorporation of oxygen into NC significantly facilitates electron transfer, as reflected by the increased contribution of oxygen to the TDOS of the ONC with the N-5 model60. In addition, the electron localization function (ELF) maps across various computational models, depicted in Fig. 5a–c and Supplementary Fig. 22, reveal distinct differences at the doped sites. Notably, the doped sites exhibit higher ELF values than the undoped sites, suggesting that heteroatom doping increases the electron density, thereby modifying the reaction kinetics and favoring Na+ capture at the doped sites over the carbon bulk25,33. Furthermore, a detailed analysis reveals that electron localization is more pronounced with the co-doping of oxygen and nitrogen than with nitrogen doping alone60. This observation significantly underscores the vital role of oxygen in the carbonaceous framework. Moreover, electron localization is more evident at the N-5/N-6 sites than at the N-Q site, which aligns with previous calculations identifying the N-5/N-6 sites as the optimal active sites25. This comprehensive comparison elucidates the substantial impact of trace O doping on the electrochemical properties of NC electrodes, affirming the strategic importance of heteroatom doping for enhanced CDI performance.

Fig. 5: Charge density analysis of ONC-S and NC-G.
figure 5

ELF maps for optimized geometries of (a) graphene, (b) NC with N-5 model, and (c) ONC with N-5 model; charge density differences of sodium atoms adsorbed on (d) graphene, (e) NC with N-5 model, and (f) ONC with N-5 model, and corresponding two-dimensional maps.

Full size image

In addition to the investigation of the electronic conductivity, further charge density difference analysis of sodium atoms adsorbed on the NC with N-5 model, the ONC with N-5 model, and graphene enables not only estimation of their charge distributions but also assessment of the interaction between the substrate and sodium atom. As illustrated in Fig. 5d–f, the charge density differences of the sodium atoms adsorbed on the NC and ONC with N-5 models appear to be more distinct than those on pure graphene, indicating an altered charge distribution after heteroatom doping18,27. In addition, two-dimensional maps of their charge density differences confirm the increased sodium adsorption after O and N doping, which is evident from the increased intensity in the doped region61,62. In addition, the charge density differences of sodium atoms adsorbed on the N-6 and N-Q sites for NC and ONC are illustrated in Supplementary Fig. 23, and a more obvious tendency for localized charge distribution can be detected at the N-5/N-6 sites, contributing to the high electrochemical activity at the doped sites27. In summary, the introduction of O into NC not only further refines the electronic conductivity but also enhances the charge distribution and sodium adsorption, contributing to an improved desalination performance.

Generally, the incorporation of N and O into a carbon framework enhances the hydrophilicity of carbon materials, thereby strengthening electrode‒water interactions, increasing the surface energy, facilitating hydrogen bonding, and increasing the surface charge18,40. To verify this improvement, the contact between H2O and various substrates (graphene, the NC with N-5 model, and the ONC with N-5 model) was investigated. As shown in Fig. 6a–c, the charge density differences of H2O adsorbed on these substrates show obvious discrepancies, in which the charge distribution at the doped sites, especially for the ONC sample, is more localized, indicating dominate adsorption at the doped sites, which also demonstrates their much higher activity than that of the undoped sites33,40. Specifically, the adsorption energies of H2O on graphene, the NC with N-5 model, and the ONC with N-5 model are calculated to be −1.06, −1.32, and −1.58 eV, respectively, thus further highlighting the enhanced interaction between H2O and ONC with N-6/N-5 sites27. Furthermore, as illustrated in Fig. 6d–f, ONC-S clearly demonstrates superior wettability to NC-G and AC, with contact angles of 29°, 36°, and 50°. This discrepancy highlights the enhanced wettability resulting from the introduction of heteroatoms. Generally, the enhanced interaction between the electrode and feedwater facilitates charge transfer, thereby increasing the electrochemical performance49.

Fig. 6: Hydrophilicity and ionic diffusion analysis of ONC-S and NC-G.
figure 6

Charge density differences of H2O adsorbed on (a) graphene, (b) NC with N-5 model, and (c) ONC with N-5 model, and corresponding adsorption energies; contact angles of H2O on the surface of (d) AC, (e) NC-G, and (f) ONC-S; simulated diffusion pathways of Na+ on the surface of (g) graphene, (h) the NC with N-5 model, and (i) the ONC with N-5 model; migration energy barriers for ionic diffusion on the surface of (j) graphene, (k) the NC with N-5 model, and (l) the ONC with N-5 model.

Full size image

The preceding analysis of the charge density variance and wettability convincingly demonstrates that heteroatom doping can alter the affinity of the substrate for sodium atoms and the interactions between them, consequently influencing charge transfer. For an ideal electrode in CDI, minimizing energy barriers and ensuring convenient pathways are imperative for achieving high desalination capacity and stability during cycling11. Thus, graphene, the NC with N-5 model, and the ONC with N-5 model were selected for comparison of their ion diffusion pathways and corresponding energy barriers. Figure 6g–i depicts the Na+ diffusion pathways on the surfaces of these models from the initial adsorption sites to the final sites, while Fig. 6j–l illustrates the energy barriers during ionic migration. Notably, the energy barrier decreases as Na+ approaches the doped sites, with the specific energy barriers for graphene, the NC with N-5 model, and the ONC with N-5 model measured as 0.31, 0.24, and 0.21 eV, respectively, underscoring that co-doping of N and O effectively reduces the diffusion barrier for Na+ and promotes rapid ion transfer8,49. These findings and calculations lay the groundwork for comprehending the superior electrochemical and desalination performance of ONC-S, further verifying the significant role of trace O doping in NC electrodes.

Practical assessment compared with commercial AC

As shown above, the DFT calculations confirm the superiority of ONC over graphene model in terms of charge transfer, ion diffusion, and wettability, showing the great potential of the obtained ONC-S for practical application. Thus, taking commercial AC as a comparison, a practical assessment of the target ONC-S was performed. As illustrated in Supplementary Fig. 24, the field-emission scanning electron microscopy (FESEM) images, XRD patterns, Raman spectra, N2 adsorption/desorption isotherms, and pore size distribution characteristics of AC clearly reveal its amorphous nature and high SSA, which facilitates ion adsorption during the desalination process63,64. In the electrochemical tests, the commercial AC exhibits specific capacitances of 107.98, 83.15, 63.21, 56.27, and 55.93 F g−1 at current densities of 1.0, 2.0, 5.0, 8.0, and 10.0 A g−1, respectively (Fig. 7a), along with capacitive contributions of 75% and 83% at 40 and 50 mV s−1 (Supplementary Fig. 25), indicating typical EDLC behavior10,37. With respect to subsequent desalination, the commercial AC displays a SAC of 16.5 mg g−1 at 1.2 V in 500 mg L−1 NaCl solution (Fig. 7b). When the applied voltage decreases, the SAC also decreases. When the NaCl concentration is increased to 1000 mg L−1, the SAC increases to 18.5 mg g−1 (Fig. 7c). In addition, the SAC obviously decreases during cycling due to the poor electrochemical stability, with a capacity retention of 45.5% after 50 cycles (Supplementary Fig. 26).

Fig. 7: Practical assessment compared with commercial AC.
figure 7

a GCD profiles of commercial AC at various current densities; plots of the SAC dynamics over time for the commercial AC under (b) various voltages and (c) various NaCl concentrations; (d) comparison of the electrochemical and desalination performances of AC, NC-G, and ONC-S; (e) comparison of the desalination performance of the target ONC-S with that of other related materials reported in the literatures (LCN: lignin derived carbon nanofibers; NP-CN: N, P co-doped carbon nanoring; PC: porous carbon; NSO-PC: N, S, O co-doped porous carbon; NHCF: N-doped hierarchical porous carbon foam; NHPC: N-doped hierarchical porous carbon; HPC: hierarchical porous carbon; AC/G: activated carbon/graphene; SMC: self-assembled mesoporous carbon; HOMC: hierarchically ordered mesoporous carbon; PCR: porous carbon rod; MCS/G: microporous carbon sphere/graphene; 3DNG: 3D nanoporous graphene; ECF/G: electrospun carbon fiber/graphene; CGF: cellulose derived graphenic fibers; 3DPG: 3D porous graphene; NPCP: N doped porous carbon polyhedron; SCBFA: sugar cane bagasse fly ash; 3DGR: three-dimensional graphene; This work: O, N co-doped carbon sheets).

Full size image

Compared with those of the as-prepared NC-G and ONC-S, both the electrochemical and desalination performances of AC are mediocre. As shown in Fig. 7d, the target ONC-S shows obvious superiority over the AC in terms of the specific capacitance, capacitive contribution, SAC value, and SAC retention, highlighting the enhanced activity after O/N co-doping and the layered architecture inherited from guanine34,65. Notably, based on the studies related to the use of AC for desalination, the poor electrochemical stability, low active site density, and restricted surface area utilization should be responsible for the unsatisfactory electrochemical performance of AC57,66. Furthermore, in addition to commercial AC, the comparison of the desalination performance of the as-prepared ONC-S with those of state-of-the-art carbonaceous materials reported in the literatures (with detailed sample information provided in Supplementary Table 7) also highlights the excellent performance of ONC-S (Fig. 7e), thus demonstrating its potential for practical application. In general, the superior performance of ONC-S can be mainly attributed to the following reasons: (1) the layered structure originating from guanine provides efficient interactions with the feedwater; (2) N doping, enhanced by trace O introduction, improves sodium adsorption and hydrophilicity; and (3) heteroatom doping modifies the migration of both ions and electrons, accelerating charge transfer12,18,64.

Discussion

The remarkable compatibility of NC with CDI has attracted significant scholarly interest. However, the precise mechanisms, particularly the roles of distinct nitrogen configurations within the NC matrix, have not yet been fully elucidated. In addition, the impact of other atomic dopants on these nitrogen configurations requires further investigation. We synthesized ONC-S featuring optimally configured N-6 and N-5 sites through our expertize in the application of carbonaceous electrodes in CDI. Our comprehensive analysis reveals that the introduction of trace O significantly enhances the hydrophilicity, structural stability, and electrochemical properties of the NC electrode, boosting the electrochemical activity of the NC by increasing the charge distribution, wettability, and ion diffusion. The presence of O, in conjunction with abundant N-5 and N-6, creates synergistic effects that increase the specific capacitance, ion adsorption capacity, and cycling stability of the target electrode. Our ONC-S electrode exhibits superior performance compared with traditional NC electrodes and commercial AC, underscoring the substantial impact of O on the desalination efficiency. The insights gained from this study pave the way for future exploration of heteroatom-doped carbon materials for CDI, offering avenues for their application in various electrochemical technologies.

Methods

Materials

All reagents and chemicals were obtained from Canrd Technology Co., Ltd., Shanghai Titan Scientific Co., Ltd., Macklin Co., Ltd., and Aladdin Chemistry Co., Ltd. and were used as received without further purification. All the solutions were prepared with deionized water. The anion and cation exchange membranes used were purchased from Tokuyama Corp., Japan.

Preparation of ONC-S and NC-G

NC-G and ONC-S were synthesized through a one-step pyrolysis process. Both guanine (99%) and adenine (99%) were sourced from Shanghai Titan Scientific Co., Ltd., China. Specifically, the guanine (2.0 g) was placed in a tubular furnace, heated at a rate of 5 °C min−1 to 900 °C, and held at this temperature for 5 h in an argon atmosphere to yield ONC-S. The same method was applied to adenine (2.0 g) to obtain NC-G. The commercial AC was purchased from Canrd Technology Co., Ltd., China.

Material characterization

FESEM (Hitachi SU4800) and TEM (JEM 2010 JEOL) were employed for morphology examination and elemental analysis. Crystal structure analysis was conducted via XRD (Bruker D8 Advance) with Cu Kα radiation (40 kV and 40 mA). Raman spectra were captured by a Raman spectrometer (Renishaw) equipped with an argon ion laser (λ = 514 nm). XPS (Thermo ESCALAB 250Xi) was employed to investigate the chemical composition of the as-prepared samples. The Brunauer-Emmett-Teller SSA was determined from N2 adsorption/desorption test data obtained via the Autosorb iQ (Quantachrome, USA). The pore size distributions of ONC-S and NC-G were derived from the adsorption branch via the Barrett-Joyner-Halenda model, and the pore size distribution of AC was obtained through the DFT method.

Electrode preparation

To create a homogeneous slurry, 80 wt.% active material (NC-G, ONC-S, or AC), 10 wt.% Super P ( ≥ 99%, conductive agent, Canrd Technology Co., Ltd.), and 10 wt.% polyvinylidene fluoride (binder, Mw~ 400,000, Macklin Co., Ltd.) were blended in 1-methyl-2-pyrrolidone (NMP, 99.5%, Aladdin Chemistry Co., Ltd.). This slurry was subsequently evenly cast onto graphite substrates (1 cm2 for the three-electrode system and 32 cm2 for the CDI cell). The working electrode was then dried at 100 °C under vacuum overnight. In the electrochemical tests, the mass loading of the active material on the working electrode was approximately 2.0 mg. In the desalination tests, it was approximately 80 mg.

Electrochemical tests

The electrochemical properties of the working electrodes were assessed at room temperature via a CHI660B electrochemical analyzer (Shanghai CH Instruments Co.). This evaluation employed CV, GCD tests, and electrochemical impedance spectroscopy (EIS) in a 1.0 M NaCl ( ≥ 99.9%, Shanghai Titan Scientific Co., Ltd., China) solution. Within the three-electrode system setup, an Ag/AgCl electrode served as the reference electrode and a platinum mesh was used as the counter electrode. Notably, the electrochemical tests for the working electrode (1 cm × 1 cm) were performed with a potential range from −1.0 to 0 V (vs. Ag/AgCl electrode) and scan rates ranging from 5.0 to 50 mV s−1. GCD tests were conducted over the same voltage range at various current densities ranging from 1.0 to 10.0 A g−1. The EIS spectra were recorded over a frequency range from 105 Hz to 10−2 Hz at an amplitude of 5 mV.

For the three-electrode system, the specific capacitance (C, F g−1) of the electrode material was calculated based on the CV curves via the following formula:

$$C={\int }_{{U}_{1}}^{{U}_{2}}i{{{\rm{d}}}}V/2{mv}(\Delta U)$$
(1)

where i denotes the current (A) and ΔU refers to the potential window (V). The symbol v represents the scan rate (V s−1), and m indicates the mass loading of the active material (g).

For the specific capacitance obtained from the GCD tests, the following equation was employed:

$$C=I\times \Delta t{{{\rm{ / }}}}m\times \Delta V$$
(2)

in this context, I, Δt, m, and ΔV represent the current (A), discharge time (s), mass loading of the active material (g), and potential window (V), respectively.

The capacitive contribution based on the CV curves collected at various scan rates can be calculated according to the following equation:

$$i(U)={k}_{1}v+{k}_{2}{v}^{1/2}$$
(3)

where i(U) and v represent the total current at a given potential (A) and the scan rate (V s−1), respectively. k1 and k2 are constants, where k1 v and k2 v1/2 correspond to the capacitive process and the diffusion-controlled process, respectively.

Desalination tests

Desalination assessments were conducted using the CDI device in batch mode, in which a NaCl solution was consistently circulated through the CDI unit (volume: 70 mL, flow rate: 100 mL min−1, temperature: room temperature) by a peristaltic pump. Real-time measurement of the pump effluent concentration was executed via an ion conductivity meter (DDSJ-308A, Shanghai Leici). Within the CDI device (Supplementary Fig. 12), a pair of electrodes (NC-G/ONC-S/AC cathode and AC anode) were separated by a spacer and an anion/cation-exchange membrane. During each batch test, a constant voltage was applied to the unit cell via the CHI660B electrochemical workstation. Desalination assessments were performed on NaCl solutions with various initial concentrations (250, 500, and 1000 mg L−1) under different voltages (0.6, 0.8, 1.0, and 1.2 V). The changes in the NaCl concentration were monitored with an online conductivity meter (Precision and Scientific Instrument, DDS-308), building upon the relationship between the conductivity and concentration established in our previous study. Specifically, the definitions for the SAC (Γ, mg g−1) and SAR (v, mg g−1 min−1) were established through the following equations:

$${{\rm{SAC}}}=({{\rm{C}}}_{0}-{{\rm{C}}})\times {{\rm{V/m}}}$$
(4)
$${\mbox{SAR}}=\varGamma /t$$
(5)

where C0 and C represent the initial and final concentrations of the NaCl solution (mg L−1), respectively; V corresponds to the volume of the NaCl solution (L); m indicates the total mass of the active material for both electrodes (g), and t represents the desalination time (min).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.