Pushing the limit of layered transition metal oxides with heterolattice oxygen-mediated redox for capacitive deionization

Abstract

The use of transition metal oxides to achieve capacitive deionization (CDI) via salt adsorption is based mainly on cation electrochemistry. Activating anionic (oxygen) redox chemistry can enable additional salt adsorption on transition metal oxides, but most conventional lattice oxygen‒metal configurations require high voltages (>4 V) for activation and are prone to lattice oxygen loss. Here, we propose a heterolattice oxygen-mediated redox mechanism to activate oxygen (O_2p) redox at <2 V by constructing a V₂O₅/V₂CO_2p heterostructure. Unlike the synthetic strategy based on excess Li/Na, we develop a barrier strategy based on an oxidative nucleophilic reaction using V₂CF_x as a precursor to induce the formation of heterolattice oxygen in V₂O₅/V₂CO_2p heterostructures. Consequently, ultrahigh CDI performance is achieved, including a salt adsorption capacity of 185.8 mg g⁻¹ at 1.4 V and a salt adsorption rate of 12.1 mg g⁻¹ min⁻¹, which exceed those of reported other faradaic materials. Further mechanistic studies reveal that the induced O_2p electrons that dominate the Fermi level provide an additional pathway for electron movement, activating additional oxygen redox processes and forming a sodium-rich vanadate (Na₄V₂O₇)/V₂CO_2p heterostructure. This strategy provides insights into the development of high-performance CDI materials with oxygen redox based on lattice oxygen‒metal configurations.

Introduction

Saline water desalination is a promising and efficient solution for maintaining fresh water supplies, and the rapid development of desalination technologies has been promoted over the past few decades^1,2. Capacitive deionization (CDI), which is a reversible electrochemical process based on the formation of electrical double layer (EDL) or redox reactions under an external direct voltage, has emerged as a rising star in the desalination community^3,4. The electrode material is undoubtedly the core of the CDI technique and plays a decisive role in the desalination performance^5,6. Generally, mainstream CDI electrodes are made up of porous carbons; however, they usually have low desalination capacities, and the poor co-ion effect in carbon-based materials usually limits improvement of the desalination performance, which slows commercialization of the CDI technique^7,8. In response, transition metal (TM) oxide (TMO) electrodes based on insertion, alloying and conversion mechanisms, which allow higher desalination capacities, have been developed. In particular, TMO electrodes have higher ion selectivity to avoid the obvious common-ion effect, making them more suitable for high-salinity water desalination. In addition, TMO electrodes have a particularly prominent ability to reduce ion migration resistance in high-ionic-strength flows, which can reduce energy consumption^9,10,11. However, their full reaction in practical applications leads to volume expansion, which can easily shorten the cycle life and reduce the rate performance. The inherently low ion diffusion/electron conduction also leads to slow reaction kinetics, making achievement of device applications difficult. Therefore, a completely different redox system beyond the central metal-mediated redox process is required.

Lattice oxygen-mediated redox is believed to provide an important supplement to the high energy storage capacity of TMOs and is generally activated by nonbonding oxygen orbitals at high voltages¹². However, the evolution of unstable oxygen, such as local oxygen vacancies and oxygen dimers, during the electrochemical process can cause irreversible oxygen redox¹³. In addition, the deep decomposition of these unstable oxygen species at high voltages not only drives O–O dimerization but also causes a series of problems, such as lattice distortion, extracellular migration of transition metal ions, and structural transformation, leading to voltage hysteresis, voltage decay, and slow oxygen redox kinetics¹⁴. To address this issue, several strategies to suppress the decay and design high-performance redox materials have been proposed, including introducing concentration gradients of Li (Na) or O, increasing the ratios of Li (Na)/TM and O/TM, and introducing atomic substituents (cation interlayer substitution) or cation vacancies¹⁵. However, intrinsic lattice oxygen-mediated redox reactions occur at high voltages >4 V, which greatly limits low-voltage-driven electrochemical devices (<2 V), such as aqueous energy storage systems and CDI systems. Directly constructing heterolattice oxygen to ensure the occurrence of oxygen redox, especially at low driving voltages, and deeply understanding the relationship between anion redox triggering and reversibility are challenging. The design of layered TMOs (TMO_2p) with heterolattice oxygen activity is believed to enable activation of reversible oxygen redox reactions to achieve high energy densities. Compared with A-rich-TMOs, which require a local Li/Na excess environment, these layered TMO_2p can capture target alkali metal ions in the electrolyte without introducing additional Li/Na sources, thus avoiding the problem of TM out-of-plane migration.

In this work, the intercalant-induced in situ synthesis method is employed to introduce heterolattice oxygen to V₂O₅ by forming a V₂O₅/V₂CO_2p heterostructure. Notably, the spatial confinement effect of the solvent cage formed by the stable tetramethylammonium cation (TMA⁺) intermediate greatly slows the oxidative nucleophilic reaction, which is key to the successful introduction of heterolattice-type O_2p and the final preparation of the V₂O₅/V₂CO_2p heterostructure. As shown in Supplementary Fig. 1, compared with commercial V₂O₅, which has slow diffusion and a large volume change, V₂O₅/V₂CO_2p, with excitation of O_2p, has the potential for rapid diffusion of Na⁺ and the formation of Na_4+xV₂O₇/V₂CO_2p heterostructures with abnormal Na storage during electrochemical processes. The electronic configuration change and charge distribution at the V₂O₅/V₂CO_2p heterointerface are conceptualized by density functional theory (DFT). Based on this, the notable Na binding energy of V₂O₅/V₂CO_2p and the contribution of induced heterolattice oxygen to fast charge transport are revealed. As expected, the V₂O₅/V₂CO_2p electrode delivers a high desalination capacity (185.8 mg g⁻¹) and an impressive desalination rate (12.1 mg g⁻¹ min⁻¹). Moreover, the structural evolution and charge transfer process of V₂O₅/V₂CO_2p during charge/discharge are systematically monitored to demonstrate its rapid kinetics and high-capacity ion storage for CDI. The design and strategy for this heterostructure will provide a direction for other materials and contribute to their application in the field of energy storage.

Results

Conceptual understanding and synthesis of the V₂O₅/V₂CO_2p heterostructure

To determine the electronic states of O ions of V-based Na-rich materials with localized Na-excess environments, DFT calculations were performed for several Na_xV_yO_z species. As shown in Supplementary Fig. 2, model systems of V₂O₅ (Na-free), Na₂V₂O₅ and Na₄V₂O₇ were initially established. The projected densities of states (PDOSs) of the O 2p state in the three oxygen environments are shown in Fig. 1a–c. Although the PDOS of O does not obviously change in the 0Na/2V and 2Na/2V configurations, it gradually shifts upward near the Fermi level (E_F) when four Na atoms are close to the oxygen (Fig. 1c). Importantly, the PDOS of O ions coordinated with four Na and two V atoms becomes much larger between 0 and −1.5 eV relative to E_F. This result suggests that the localized Na-excess environment induces a significant increase in the unstable electronic state of the O ion near E_F, while O 2p provides an additional pathway for electron movement, which in turn increases the capacity by triggering additional redox processes. Furthermore, as shown in Supplementary Fig. 3, between 0 and −2.0 eV, the PDOS produced by the O 2p state is much greater than that produced by the V 3d state, indicating that O 2p greatly improves electron transfer. With respect to 0Na/2V and 2Na/2V, the lower distribution of the V 3d band in the energy band structure (Supplementary Fig. 3a, b) increases the possibility of O₂ release and corresponds to irreversible redox processes with poor anionic utilization. The 4Na/2V configuration exhibits a higher O 2p state near E_F (Supplementary Fig. 3c), whereas the elevation of the V 3d band suppresses the tendency for O₂ release, suggesting that the localized Na-excess environment increases anionic utilization and improves redox reversibility. The calculations show that the O 2p states of all the species dominate the top of the valence band, implying that oxygen can successfully participate in the oxidation process.

Fig. 1: Effect of the local atomic environment on the electronic states of O ions in Na_xV_yO_z.

PDOSs of the O 2p orbitals of O atoms in Na_xV_yO_z with coordination of a zero Na and two V, b two Na and two V, and c four Na and two V. Insets are the coordination modes of the O ion. d Predicted stable configurations for V₂O₅/V₂CO_2p. e PDOS of O atoms of V₂O₅/V₂CO_2p. f XRD patterns at different 2θ values, where the arrows show the shift of the V₂CT_x (002) peak during the V₂O₅/V₂CO_2p synthesis. g SEM images of V₂O₅/V₂CO_2p (30%) at different magnifications. h Atomic schematic of the expected formation mechanism for the V₂O₅/V₂CO_2p heterostructure.

We designed layered V-oxides (V₂O₅/V₂CO_2p) with heterolattice oxygen activity that can activate reversible oxygen redox. As shown in Supplementary Fig. 4, intercalation of TMA⁺ as a surface modifier is achieved by introducing tetramethylammonium hydroxide into the V₂CF_x MXene precursor. In this process, the ‒F terminal functional groups are replaced by many strongly negatively charged ‒O terminals. Afterward, a V_xO_y/V₂CO_2p heterostructure rich in heterolattice oxygen is formed in situ via the solvothermal method, in which part of V₂C is slightly oxidized to form V_xO_y. Finally, a highly crystallized V₂O₅/V₂CO_2p heterostructure is achieved by annealing V_xO_y/V₂CO_2p. As shown in Fig. 1d, the predicted stable configuration of the V₂O₅/V₂CO_2p heterojunction was preliminarily constructed. Many O₁ (O_2p) terminal functional groups exist at the V₂O₅/V₂CO_2p heterointerface, O_2p is partially replaced by N atoms, and the V atom derived from V₂O₅ coordinates with the surrounding O₁ (V₂CO_2p) and O₂ (V₂O₅) to form V₂O₅, which matches the conceptualization of the in situ formation of the V₂O₅/V₂CO_2p heterostructure very well. Furthermore, the local distribution characteristics of electrons in the V₂O₅/V₂CO_2p heterostructure were investigated via the electron localization function (ELF) (Supplementary Fig. 5). According to the total DOS (TDOS) (Supplementary Fig. 6a), the rich electronic states at E_F result in high conductivity in V₂O₅/V₂CO_2p. In addition, a band gap does not appear for the V₂O₅/V₂CO_2p heterostructure, which suggests that the V₂O₅/V₂CO_2p heterostructure has strong intrinsic electron transport behavior, thus guaranteeing Na⁺ transport at high rates. Moreover, the PDOS shown in Supplementary Fig. 6b reveals that the abundant electron state density at the Fermi level is mainly contributed by O₁, verifying that the strong electronegativity of O₁ positively contributes to the conductivity of the V₂O₅/V₂CO_2p heterostructure. As shown in Fig. 1e, O_2p electrons are dominant at the Fermi level, especially for O₁, leading to nonhybridization of heterolattice oxygen^11,16, from which electrons can be removed for charge compensation.

X-ray diffraction (XRD, SmartLab, Japan) was performed, as shown in Fig. 1f. The typical V₂CF_x structure is clearly obtained after etching V₂AlC since the classical reflection peak at 7.2° corresponding to the (002) plane is observed^17,18,19. Upon the introduction of the intercalation agent, tetramethylammonium hydroxide (TMAOH), the peak shifts to a lower angle. Remarkably, the movement of the (002) plane is strengthened from V₂CO_x-TMA (10%) to V₂CO_x-TMA (50%) by adjusting the concentration of TMAOH. The polycrystalline V_xO_y/V₂CO_2p heterostructure was subsequently obtained by oxidizing V₂CO_x-TMA via a solvothermal reaction. The typical reflection peak associated with V₂O₅ is detected for V_xO_y/V₂CO_2p, suggesting that V_xO_y has been completely converted into V₂O₅. Notably, the optimum concentration of TMAOH is 30% since the reflection peak related to V₂CF_x is the strongest in V₂O₅/V₂CO_2p (30%). In V₂O₅/V₂CO_2p (10%), the terminal functional group ‒F is less replaced by O, which cannot guarantee the formation of an appropriate amount of TMA⁺ solvent cage. In contrast, the excessive TMAOH in V₂O₅/V₂CO_2p (50%) leads to enrichment of the surface terminal functional group ‒O, which accelerates the oxidation reaction. Supplementary Fig. 7a shows a scanning electron microscopy (SEM, Hitachi SU5000 Japan) image of V₂CF_x, which displays a typical accordion-like multilayer nanostructure. Supplementary Fig. 7b shows that for V₂CO_x-TMA (30%), the interlayer spacing significantly expands. This enlarged interlayer spacing is beneficial for the in situ formation of V_xO_y. Supplementary Fig. 7c and 7d shows SEM images of V_xO_y/V₂CO_2p (30%). The accordion-like multilayered nanostructure remains, while the surface of V_xO_y/V₂CO_2p (30%) is covered by some tiny V_xO_y nanoparticles. This result is attributed to the in situ reaction of O and V on the V₂CF_x surfaces. SEM images of V₂O₅/V₂CO_2p are shown in Fig. 1g. The interlayer spacing of V₂O₅/V₂CO_2p shrinks after annealing because of the removal of terminal functional groups on V₂CF_x at high temperatures. The high-magnification SEM image also shows that the highly crystallized V₂O₅ nanoparticles are closely arranged on the surface of V₂CO_2p. This phenomenon could ensure fast charge transfer at the heterointerface and enhance the conductivity of the V₂O₅/V₂CO_2p electrode. In addition, SEM images of V₂O₅/V₂CO_2p (10%) and V₂O₅/V₂CO_2p (50%) are displayed in Supplementary Fig. 8. Both of these materials exhibit a similar behavior to that of V₂O₅/V₂CO_2p (30%).

Based on the above results, the synthesis mechanism of the V₂O₅/V₂CO_2p heterostructure shown Fig. 1h is proposed. Preoxidation is expected to commence from the terminal functional group ‒F on the V₂CF_x surface. When annealed at a high temperature under an air atmosphere, oxygen is adsorbed and diffuses on the surface of V₂CF_x and then replaces the ‒F terminal group. When the oxygen density increases, V₂CF_x can be converted to V₂CO_x. Regarding the V‒O bond, the electrons tend to combine with the O atoms rather than the V atoms. Therefore, the electron-deficient intermediate V atom is more vulnerable to electron attack than the O atom and finally completes the oxidation reaction. In this process, the air provides a sufficient oxygen supply to allow full oxidation of V₂CF_x to form orthogonal V₂O₅ crystals. This conclusion is further confirmed by the XRD and SEM results shown in Supplementary Fig. 9 and Supplementary Fig. 10, respectively. V₂O₅ with high crystallinity is obtained by directly annealing V₂CF_x in air, while the reflection peak of V₂CF_x is not detected. Moreover, a nanorod structure forms due to the growth of V₂O₅ along the (001) plane. In addition, owing to the action of the alkaline intercalation agent TMAOH, –OH groups replace –F groups at the terminal of V₂CF_x, forming V₂C(OH)_x. In the case of excessive –OH groups, the –OH groups deprotonate to form O-based terminals with strong electronegativity. Under alkaline conditions, the deprotonated V–O– part is stabilized by forming an intermediary (V–O–TMA⁺) with TMA⁺, resulting in less electron localization on the V atom. This electron-deficient intermediate V atom is not easily attacked by electrons of the O atom in O₂. In addition, the coordinated TMA⁺ can help capture H₂O molecules, forming a large solvent cage to surround the V–O– part. Because of the poor electrophilicity of the intermediate and the steric hindrance effect, the solvent cage hinders entry of H₂O/O₂ and slows the oxidation process, thus ensuring the formation of the V₂O₅/V₂CO_2p heterostructure²⁰. These results illuminate that the alkaline intercalation agent TMAOH plays a decisive role in the oxidation process when constructing the V₂O₅/V₂CO_2p heterostructure.

Characterization of the V₂O₅/V₂CO_2p heterostructure

Figure 2a, b shows high-resolution transmission electron microscopy (HRTEM, FEI Talos F-200S) images of V₂O₅/V₂CO_2p at different magnifications. The V₂O₅/V₂CO_2p nanoparticles are clearly tightly connected, generating a nanosheet-like structure. The V₂O₅/V₂CO_2p microstructures were further analyzed in depth using aberration-corrected scanning transmission electron microscopy (STEM, Thermo Scientific, Themis Z). As shown in Fig. 2c, the lattice spacings of 0.32 nm (green) and 0.63 nm (yellow) correspond to the (110) and (004) crystal planes of V₂O₅ and V₂CO_2p, respectively. Most importantly, the lattice stripes marked indicate the in situ growth of V₂O₅ on V₂CO_2p. Moreover, high-angle annular dark-field STEM (HAADF-STEM) images (Fig. 2d) were obtained to refine the atomic arrangement of the V₂O₅/V₂CO_2p heterointerfacial connection. The alternating atomic layers of V₂O₅ (green) and V₂CO_2p (yellow) demonstrate coherent intergrowth of structural domains with different orientations of V₂O₅ and V₂CO_2p (marked with white arrowheads), which provides strong evidence for the in situ growth of V₂O₅/V₂CO_2p heterostructures. In addition, the HAADF image in Supplementary Fig. 11 combined with the element mapping images illustrate the even distributions of V, C, O and N in V₂O₅/V₂CO_2p.

Fig. 2: Characterization of basic structure and morphology of V₂O₅/V₂CO_2p.

a, b HRTEM images of V₂O₅/V₂CO_2p at different magnifications. c STEM and d HAADF‐STEM images of V₂O₅/V₂CO_2p. e O K-edge XAS spectrum of V₂O₅/V₂CO_2p. f EPR spectra of V₂O₅/V₂CO_2p. g Comparison of the adsorption energies (ΔE, eV) for a Na atom. XPS spectra of h O 1s, i V 2p and j N 1s. k CDD analysis of V₂O₅/V₂CO_2p, the yellow and blue regions represent charge accumulation and depletion, respectively. l Top view of the CDD. The inset in c is enlarged area marked by the white dash line. The green and yellow lines in c and d correspond to the (110) and (004) crystal planes of V₂O₅ and V₂CO_2p, respectively. The dash lines in h, i and j show the shift of the peak during the V₂O₅/V₂CO_2p synthesis.

Importantly, X-ray absorption near-edge structure spectroscopy (XANES) of V₂O₅/V₂CO_2p was explored to provide stronger evidence for the formation of O 2p. As shown in Fig. 2e, at the O K-edge, the peaks between 525 and 535 eV are caused by the O 1s orbital electron leaping to the V 3d‒O 2p (t_2g) and V 3d‒O 2p (e_g) states, respectively, corresponding to π and σ interactions^21,22,23. The d peak near 542.7 eV belongs to the transition from the O 1s nuclear energy level to the O 2p energy level with hybridization of the V 4s and V 4p states, which indicates that the V‒O bond has an obvious covalent character. The X-ray absorption spectroscopy (XAS) results are in agreement with the X-ray photoelectron spectroscopy (XPS) results, which further proves the existence of O _2p in V₂CO_2p. Electron paramagnetic resonance (EPR, Bruker EMXplus-6/1) was used to detect oxygen vacancies (Fig. 2f), in which a signal at g = 2.006 originating from the oxygen vacancies is observed for all the samples. Interestingly, the signal is strong in V₂O₅/V₂CO_2p (30%), implying the presence of massive amounts of oxygen vacancies. The presence of oxygen vacancies enhances the adsorption of Na⁺, leading to fast kinetics. Signals from oxygen vacancies are observed at g = 2.005 for all V_xO_y/V₂CO_2p samples and are more intense than those of V₂O₅/V₂CO_2p (Supplementary Fig. 12). This behavior may originate from the increase in crystallinity, in which the deep defect levels induced by oxygen vacancies are largely eliminated, leading to a reduction in the number of oxygen vacancies. To reveal the interactions between the different terminal functional groups and Na⁺ during the evolution of V₂O₅/V₂CO_2p, DFT calculations were adopted (Supplementary Fig. 13a-c). The adsorption energies of the optimized structures are shown in Fig. 2g. Compared with V₂CO_x-TMA (−3.39 eV), the adsorption energy of V₂CF_x is much lower (−2.98 eV) (Supplementary Fig. 13d-f), confirming that the Na atom has a stronger affinity with the −O terminal functional group than with the −F group. This behavior is attributed mainly to the strong interaction between the electron-rich Lewis base and the electron-deficient Lewis acid Na⁺. Furthermore, the adsorption energy of the Na atom at the V₂O₅/V₂CO_2p heterointerface is calculated as −4.14 eV, whereas it is −4.60 eV and −3.97 eV for the top and bottom Na atoms, respectively (Supplementary Fig. 14). This finding may be due to V₂O₅ and V₂CO_2p having strong covalent bonds (V (V₂O₅)–O_2p–V (V₂CO_2p)), which result in a smaller interfacial gap. Therefore, the adsorption of V₂O₅ and V₂CO_2p on Na atoms at the heterojunction interface would be competitive, thus weakening the adsorption energy at the interface.

To analyze the chemical bonding information, XPS (Thermo Escalab 250Xi) spectra of the samples were collected. In Fig. 2h, the deconvoluted O 1s XPS spectra show four characteristic peaks at 530.2, 530.7, 531.3 and 532.2 eV, which are attributed to the oxygen species in V‒O (V₂O₅), the peroxo-related species (O₂²⁻)^24,25,26, the oxygen in the terminal functional group C–V−O_x, and the –OH terminal functional group/surface-adsorbed H₂O (C–V–(OH)_x), respectively. The characteristic peaks of C–V–(OH)_x and V–O appear in V₂CO_x-TMA, which originate from the substitution of terminal functional groups –F/–O on the surface of V₂CF_x by –OH groups and the micro-oxidation of V₂CF_x caused by the presence of H₂O, respectively. Furthermore, the characteristic peak of O₂^2- emerges in V_xO_y/V₂CO_2p, which is attributed to the intercalation agent TMAOH. In addition, O₂^2- is still retained in the V₂O₅/V₂CO_2p heterostructure after annealing. O₂^2- has been reported to have strong anionic reactivity, which ensures triggering of V₂O₅/V₂CO_2p oxygen redox. Compared with V_xO_y/V₂CO_2p, the characteristic peaks of O₂^2- and V –O shift to the lower energy region in V₂O₅/V₂CO_2p, indicating that the O atoms related to O₂^2- and V –O gain electrons, which is consistent with the Bader charge analysis results. Figure 2i shows the high-resolution V 2p XPS spectrum. Usually, V mainly exists in the chemical states of +2, +3 and +4 in V₂CF_x^27,28. The intensity of the peak associated with V⁴⁺ increases in V₂CO_x-TMA, confirming that TMAOH enables V to reach a high chemical state. Furthermore, V⁵⁺ is detected in the spectra of the V₂O₅/V₂CO_2p heterostructure. Notably, the deconvoluted characteristic peak of V gradually shifts to the higher energy region during the conversion of V₂CF_x to the V₂O₅/V₂CO_2p heterostructure, which demonstrates that V loses electrons during this process. The high-resolution N 1s XPS spectra can be deconvoluted into three characteristic peaks at 395.9, 399.8 and 401.6 eV, as depicted in Fig. 2j, which are derived from lattice substitution (LS), surface functional site substitution (FS) and surface absorption (SA). Among them, the proportion of SA in V_xO_y/V₂CO_2p is greater than that in the other samples^29,30. For V₂O₅/V₂CO_2p, LS-N becomes dominant due to the incorporation of SA-N into the MXene. Moreover, the characteristic peak shifts to the lower energy region, suggesting that N obtains electrons, which is consistent with the DFT analysis results. N doping of V₂CO_2p increases the electrostatic repulsion between the layers, which effectively alleviates restacking of the MXene layer, thereby promoting penetration of the electrolyte. In addition, the group vibrations, specific surface areas (SSAs) and pore size distributions of the samples were analyzed by Fourier transform infrared (FTIR, WQF-520A) spectroscopy and N₂ adsorption‒desorption measurements (Supplementary Figs. 15–17 and Supplementary Table 1).

Furthermore, the electronic configuration change and charge distribution at the V₂O₅/V₂CO_2p heterointerface were investigated based on the charge difference density (CDD), as shown in Fig. 2k. Obviously, charge carriers accumulate in the V₂O₅/V₂CO_2p heterostructure, and local electrons are redistributed and delocalized, which effectively regulates the electron configuration and proves the strong charge interaction between V₂O₅ and V₂CO_2p. Notably, electrons accumulate at the interface, especially around the O₁ atoms, which is attributed to the formation of O₁ with strong electronegativity after the deprotonation induced by TMAOH. Additionally, the specific electron transfer was determined by Bader charge analysis, as shown in Supplementary Table 2. According to these results, V atoms in V₂O₅ and V₂CO_2p can be inferred to release more electrons that flow to O₁ atoms. Moreover, the electrons can be inferred to be transferred through the V (V₂O₅)-O_2p (O₁)-V (V₂CO_2p) interface bonds. Moreover, V₂O₅ contributes 3.55 electrons to V₂CO_2p, which promotes strong interface electron coupling between V₂O₅ and V₂CO_2p. Therefore, the V₂O₅/V₂CO_2p heterostructure is conducive to improving the migration/transport dynamics of Na⁺ ions and electrons. Figure 2l shows the CDD image from the top view. Charge accumulation around the N atom is clearly observed, which verifies that N doping facilitates electron transfer.

Electrochemical properties

Figure 3a shows the cyclic voltammetry (CV) curves of the V₂O₅/V₂CO_2p, commercial V₂O₅ and annealed-V₂CF_x electrodes. They all exhibit a pair of reversible redox characteristic peaks at a scanning rate of 1.0 mV s⁻¹ in a three-electrode system. An oxidation peak clearly appears at ~0.4 V in the anode scan, which is related to the oxidation of V⁴⁺ to V⁵⁺ during the Na⁺ extraction process. The corresponding reduction peak is observed at approximately −0.2 V in the cathodic scan, which represents the conversion of V⁵⁺ to V⁴⁺ accompanied by Na⁺ intercalation. Notable redox reversibility of the V₂O₅/V₂CO_2p electrode is therefore achieved. According to the peak height (i_pc and i_pa), interpeak spacing (E_pc and E_pa) and symmetry of the oxidation and reduction peaks in the CV curves^31,32, the reversibility of the reaction associated with the electroactive substance on the electrode surface and the kinetics of the electrode process can be explored. Compared with commercial V₂O₅ (ΔE_p = 0.63) and annealed-V₂CF_x (ΔE_p = 0.65), V₂O₅/V₂CO_2p realizes a much narrower redox potential gap (ΔE_p = 0.59), a higher current density and a lower resistance. Furthermore, the |i_pc|/|i_pa| values of the commercial V₂O₅, annealed-V₂CF_x and V₂O₅/V₂CO_2p electrodes are 1.04, 0.98 and 1.01, respectively, indicating superior reversibility. Generally, the electrochemical surface area (ECSA) is an indicator of the electrochemically active area, i.e., the effective area involved in the electrochemical reaction. Therefore, we estimated the ECSA based on the CV curve. As shown in Supplementary Fig. 18, the V₂O₅/V₂CO_2p electrode has the highest ECSA (3.16 cm²) compared with commercial V₂O₅ (0.30 cm²) and annealed-V₂CF_x (2.84 cm²). Figure 3b shows the specific capacitance obtained from the CV curve. The specific capacitance of V₂O₅/V₂CO_2p is the highest among all the electrodes at each fixed scan rate. The galvanostatic charge‒discharge (GCD) curves of the three samples are shown in Fig. 3c. They all display an almost triangular shape, and the voltage plateau is consistent with the corresponding CV curve. According to the GCD curves, the specific capacitances of the commercial V₂O₅, annealed-V₂CF_x and V₂O₅/V₂CO_2p electrodes are calculated to be 339 F g⁻¹, 172 F g⁻¹ and 84 F g⁻¹, respectively (Supplementary Fig. 19). In addition, when the current density is varied from 0.8 to 5.0 A g^-1, the specific capacitance of V₂O₅/V₂CO_2p is the highest. Figure 3d shows the electrochemical impedance spectroscopy (EIS) spectra of the three electrodes in the frequency range of 0.01 ~ 10⁶ Hz. The fitting results are shown in Supplementary Table 3. V₂O₅/V₂CO_2p has the smallest R_s and R_ct and a larger linear slope, thus featuring the highest conductivity. This behavior is attributed to the coupling of heterostructures generating a heterogeneous interface with a built-in electric field, which accelerates the interface charge transfer, facilitates transport and diffusion of ions in the V₂O₅/V₂CO_2p electrode, and provides a more effective deionization energy.

Fig. 3: Electrochemical and kinetic analysis of V₂O₅/V₂CO_2p.

a CV curves in 1 M NaCl solutions at 1 mV s⁻¹ within the potential window of −0.4 to 0.8 V. b Specific capacitance (F g⁻¹) with respect to the scan rate (mV s⁻¹). c GCD curves in 1 M NaCl solutions at 0.5 A g⁻¹. d EIS spectra of V₂O₅/V₂CO_2p, commercial V₂O₅ and annealed-V₂CF_x electrodes. e CV curves at different scan rates. f Relationships between peak currents and the scan rate. g Capacitive contribution at 1.0 mV s⁻¹ and h capacitive contribution at various scan rates for the V₂O₅/V₂CO_2p electrode.

To clarify the origin of the fast kinetics of the V₂O₅/V₂CO_2p electrode, the contributions of capacitance control and diffusion control to the capacity were quantitatively analyzed. Figure 3e shows the CV curve of V₂O₅/V₂CO_2p scanned in the range of 1.0~100 mV s⁻¹. An increase in the scan rate is accompanied by an increase in the polarization effect, which causes a shift in the oxidation characteristic peak to a high potential and a shift in the reduction characteristic peak to a low potential. In general, the capacitance effect relationship is given by Eq. (1)^33,34

$$i={{av}}^{b}$$

(1)

where i and v represent the response current and the scan rate, respectively. The values of a and b are determined by plotting the curves of log (i) versus log (v). When b is equal to 0.5, diffusion dominates the reaction kinetics. When b is equal to 1.0, the capacitance storage process dominates the reaction kinetics. The b values calculated for the anode peak and the cathode peak are shown in Fig. 3f and are 0.65 and 0.64, respectively. This result indicates that the electrochemical process of V₂O₅/V₂CO_2p is synergistically determined by the capacitive reaction and the diffusion reaction. Moreover, the capacitance contribution ratio can be quantitatively analyzed based on the correlation between the capacitance effect k₁v and the diffusion effect k₂v^1/2, as shown in Eq. (2)^33,34

$$i\left(V\right)={k}_{1}v+{k}_{2}{v}^{1/2}$$

(2)

Taking the scan rate of 1.0 mV s⁻¹ as an example (Fig. 3g), the capacitance contribution ratio of the V₂O₅/V₂CO_2p electrode is 23%. In addition, the capacitance contribution ratio of V₂O₅/V₂CO_2p when the scan speed was increased from 1.0 mV s⁻¹ to 100 mV s⁻¹ was calculated. The statistical results are shown in Fig. 3h. With increasing scan rate, the proportion of the capacitance contribution gradually increases, and the control dynamics become more prominent, dominating the high-rate performance.

CDI performance evaluation

Desalination experiments were carried out in an intermittent recirculation system (Supplementary Fig. 20) with a NaCl solution of a certain concentration as the target brine, which was continuously driven by a peristaltic pump through the CDI module at a constant flow rate. A direct current power supply was connected between the electrodes of the CDI module to form an electric field, resulting in a decrease in the conductivity and current. A conductivity monitor and an electrochemical workstation were used to synchronously record the changes in the conductivity and current in the circulating solution. The CDI performance of the V₂O₅/V₂CO_2p anode was evaluated by combining a V₂O₅/V₂CO_2p//activated carbon (AC) CDI system with an intermittent recirculation device. Figure 4a, b shows the changes in the conductivity and current, respectively, in real time at voltages of 0.8, 1.0 and 1.2 V in a 1000 μS cm⁻¹ NaCl solution. Essentially, both the conductivity and current sharply decrease once the direct voltage is activated. After 60 min, they almost do not change further, or the variations are very small, indicating that the electrode becomes saturated due to the equilibrium of the adsorbed charged ions. During the salinization process, a voltage of -0.8 V was applied to the CDI module. Figure 4c shows the Ragone Kim–Yoon plot curve of the V₂O₅/V₂CO_2p//AC system in a 1000 μS cm⁻¹ NaCl solution at a voltage of 0.8~1.2 V. The optimal desalination capacity increases from 58.9 mg g⁻¹ to 77.0 mg g⁻¹ when the working voltage is varied from 0.8 V to 1.2 V. Moreover, the desalination rate gradually increases with increasing voltage. The highest desalination rate of 4.9 mg g⁻¹ min⁻¹ is obtained at a voltage of 1.2 V, which is very competitive among the reported CDI electrodes. Furthermore, in the initial stage of salinization, an ultrahigh salinization rate of 5.8 mg g⁻¹ min⁻¹ is obtained. The high desalination rate and salinization rate demonstrate that the V₂O₅/V₂CO_2p//AC system has notable Na⁺/Cl⁻ reversible intercalation/deintercalation kinetics. Figure 4d shows the effect of voltage on the desalination performance. Apparently, V₂O₅/V₂CO_2p (30%) has the highest salt removal capacity (SRC). When the applied voltage is 0.8~1.4 V, the desalination capacity increases from 58.9 mg g⁻¹ to 185.8 mg g⁻¹ and reaches the maximum value at 1.4 V. In constant-voltage operation mode, high electrode potentials lead to more charge transfer to the active material. With the accumulation of charge in the electrode, the ions in the solution have a stronger driving force, thus increasing the adsorption capacity. The conductivity changes of the V₂O₅/V₂CO_2p (10%) and V₂O₅/V₂CO_2p (50%) electrodes in the CDI system are shown in Supplementary Fig. 21. Moreover, the water recovery (WR, %) and volumetric energy consumption (E_v, Wh m^-3) have become important indicators for evaluating the superiority of CDI systems. As shown in Fig. 4e, the WR values for the V₂O₅/V₂CO_2p//AC system reach 64%, 63% and 64% at voltages of 0.8, 1.0 and 1.2 V, respectively. Correspondingly, E_v is 14.16, 26.29 and 36.31 Wh m⁻³, respectively, which is very impressive among the reported electrodes. In Supplementary Fig. 22 and Supplementary Fig. 23, the conductivity changes of the commercial V₂O₅//AC and V₂CO_x-TMA//AC CDI systems at voltages of 0.8, 1.0 and 1.2 V in a 1000 μS cm⁻¹ NaCl solution are respectively shown, suggesting a conductivity change trend similar to that of the V₂O₅/V₂CO_2p//AC CDI system. Figure 4f shows the SRCs of the V₂O₅/V₂CO_2p, commercial V₂O₅ and V₂CO_x-TMA electrodes under different voltages. Among all the systems, V₂O₅/V₂CO_2p has the highest desalination capacity, reaching 77.0 mg g⁻¹ at a voltage of 1.2 V. The desalination ability of the V₂O₅/V₂CO_2p//AC system under different working voltages in KCl, NaCl, LiCl and ZnCl₂ solutions with an initial concentration of 529 mg L⁻¹ was tested, as shown in Supplementary Fig. 24 and Fig. 4g. The V₂O₅/V₂CO_2p//AC system has the highest removal ability for Na⁺, followed by Li⁺ > K⁺ > Zn²⁺. Figure 4h shows the statistics of the desalination capacity of various electrodes in the CDI device. Notably, among all the target materials, the V₂O₅/V₂CO_2p//AC system exhibits the best desalination ability under certain experimental conditions (Supplementary Table 4). Supplementary Fig. 25 shows the desalination stability and capacity retention rate of the V₂O₅/V₂CO_2p//AC and commercial V₂O₅//AC CDI systems in a 1000 μS cm⁻¹ NaCl solution at 1.2 V. After 50 cycles, the desalination capacity of V₂O₅/V₂CO_2p//AC is maintained at 75.5%. Notably, the capacity retention rate of commercial V₂O₅//AC is only 34.9% at the fifth cycle. In addition, the ion concentrations after desalination/salination were examined via inductively coupled plasma‒optical emission spectrometry (ICP‒OES, Agilent, 5110, USA) at different voltages. As shown in Supplementary Table 5, the amount of V^δ+ released by commercial V₂O₅ is nearly ~5 times greater than that by V₂O₅/V₂CO_2p. This result suggests the exceptional structural stability of V₂O₅/V₂CO_2p. Moreover, in the V₂O₅/V₂CO_2p//AC system, the concentration of V^δ+ after desalination is much lower than 0.01 mg L⁻¹, which satisfies the drinking water sanitary standard (GB5749-2006) very well. In addition, Supplementary Fig. 26 shows digital photographs of the anode and cathode after desalination by the commercial V₂O₅//AC and V₂O₅/V₂CO_2p//AC CDI systems. Both the anode and cathode of the V₂O₅/V₂CO_2p//AC CDI system are obviously clearer, confirming its good structural stability and notable cycling stability.

Fig. 4: CDI performance and rate of V₂O₅/V₂CO_2p.

a Conductivity and b current transient of the V₂O₅/V₂CO_2p//AC-based CDI system. c CDI Ragone Kim‒Yoon plots versus voltage. d SRC (mg g^-1) with respect to the voltage (V) in different systems. e WR and E_v of the V₂O₅/V₂CO_2p//AC-based CDI system. f SRC with respect to the voltage (V) in different systems. g SRC of the V₂O₅/V₂CO_2p//AC-based CDI system in different saline solutions. h Comparison of SRCs, in which data were collected from published works, as shown in Supplementary Table 5.

Mechanistic insights into the superior CDI performance of the V₂O₅/V₂CO_2p heterostructure

XRD, XPS, HRTEM and DFT calculations were further employed to explore the desalination mechanism of the V₂O₅/V₂CO_2p heterostructure in detail. As shown in Fig. 5a, nine different charge/discharge potentials were selected from the GCD curves within the potential window of −0.4~0.8 V. Figure 5b, c shows the XRD patterns corresponding to each point. When the first charge is from the open-circuit potential (OCP) to 0.4 V, a characteristic peak appears near a 2θ of 31.6°, corresponding to the sodium-rich Na₄V₂O₇ phase (PDF#28-1180). As the charging potential increases, the peak strengthen and shifts in the low-angle direction. This behavior is attributed mainly to the insertion of Na⁺. During the discharge process, the peak representing Na₄V₂O₇ gradually shifts to a high angle, and the corresponding intensity gently weakens, which results from the deinsertion of Na⁺. More importantly, this characteristic peak still exists when the sample is completely discharged to −0.4 V, implying that irreversible V₂O₅ is transformed into reversible sodium-rich Na₄V₂O₇, thereby forming the Na₄V₂O₇/V₂CO_2p heterostructure. A second charging was further carried out to 0.8 V. Unsurprisingly, compared with the c point (first charge to 0.8 V), the Na₄V₂O₇ characteristic peak at the i point (second charge to 0.8 V) becomes stronger and has a larger negative displacement, suggesting that more Na⁺ ions are embedded and confirming the high reversibility of Na₄V₂O₇. Notably, the trends of the intensity and position changes of the Na₄V₂O₇ characteristic peaks are more obvious in the h to i process than in the a to c process representing the Na₄V₂O₇ to Na_4+xV₂O₇ (Na⁺-rich state) transition, whereas the h to i process (second charge process) is ascribed to the Na_4-xV₂O₇ (Na⁺-barren state) to Na_4+xV₂O₇ (Na⁺-rich state) transition. Figure 5d shows the contour map of the XRD characteristic peak of Na₄V₂O₇, which is highly consistent with the XRD results.

Fig. 5: Desalination mechanism of the V₂O₅/V₂CO_2p.

a First and second charge/discharge profiles and selected voltage states of V₂O₅/V₂CO_2p. b, c XRD patterns of V₂O₅/V₂CO_2p at different charge/discharge stages within 2θ ranges of 5° to 80° and 31° to 33°, in which the arrows show the shift of the Na₄V₂O₇ peak during the charging and discharging processes. d 2D contour plots of the XRD patterns. HRTEM images of the V₂O₅/V₂CO_2p electrode after e the first charge to 0.8 V, f the first discharge to −0.4 V and g the second charge to 0.8 V, the inset on the right is enlarged area marked by the dash line. h Schematic of the working mechanism for the charge/discharge processes of V₂O₅/V₂CO_2p.

To further highlight the sodium-rich storage characteristics of the V₂O₅/V₂CO_2p heterostructure, Supplementary Fig. 27 shows the XRD patterns of the commercial V₂O₅ and annealed-V₂CF_x electrodes under different charge‒discharge potentials. During the first charge process from the OCP to 0.4 V, the characteristic peak of Na₄V₂O₇ is not observed, and only the characteristic peak of Na₂V₂O₅ is detected³⁵. In addition, the characteristic peaks related to Na₄V₂O₇ are not detected even when charging to 0.8 V, further demonstrating the sodium-rich storage characteristics of the V₂O₅/V₂CO_2p heterostructure. Figure 5e-g shows HRTEM images at different charge/discharge potentials. The enlarged HRTEM image in Fig. 5e shows lattice spacings of 0.65 nm and 0.44 nm, which correspond to the (004) plane of V₂CO_x and the (008) plane of Na_4+xV₂O₇, respectively. Moreover, both lattice spacings are significantly higher than the theoretical values of 0.63 nm for the (004) plane of V₂CO_x and 0.40 nm for the (008) plane of Na₄V₂O₇ due to the intercalation of a large number of Na⁺ ions. The (004) plane of V₂CO_x and (008) plane of Na_4+xV₂O₇ are easily captured when the first discharge process reaches -0.4 V, as shown in Fig. 5f, and even when the second charge process tends toward 0.8 V, as shown in Fig. 5g. Therefore, the abnormal capacity enhancement of the V₂O₅/V₂CO_2p heterostructure can be attributed to the following aspects. (1) A sodium-rich Na₄V₂O₇/V₂CO_2p heterostructure is formed after the initial insertion of Na⁺. (2) The Na₄V₂O₇/V₂CO_2p heterostructure has a good elastic interlayer space for the insertion/deinsertion of Na⁺. (3) The Na₄V₂O₇/V₂CO_2p heterostructure is robust. (4) In the subsequent charge/discharge processes, the Na₄V₂O₇/V₂CO_2p heterostructure has notable Na⁺ enrichment and reversible conversion of the barren state (Na_4-xV₂O₇/V₂CO_2p ↔ Na₄V₂O₇/V₂CO_2p ↔ Na_4+xV₂O₇/V₂CO_2p). Based on these results, the charge/discharge mechanism of the V₂O₅/V₂CO_2p heterostructure is schematically proposed, as shown in Fig. 5h. In the initial stage of the charge process, Na⁺ is inserted into the V₂O₅/V₂CO_2p heterostructure, forming a sodium-rich Na₄V₂O₇/V₂CO_2p heterostructure due to the induced O_2p state. As the charge process continues, more Na⁺ can be inserted into the Na₄V₂O₇/V₂CO_2p heterostructure, resulting in the appearance of an interlayer Na⁺-rich region in the Na_4+xV₂O₇/V₂CO_2p state. Afterward, many Na⁺ ions are released when the discharge process begins. Compared with Na₄V₂O₇/V₂CO_2p, the Na⁺ barren region formed by Na_4-xV₂O₇/V₂CO_2p in the interlayer can provide more storage sites for the subsequent capture of more Na⁺.

The origin of the fast Na⁺ transport kinetics of the V₂O₅/V₂CO_2p heterostructure was further revealed by XPS. Figure 6a–c and Supplementary Fig. 28 show the XPS spectra taken under the different charge/discharge states shown in Fig. 5a. In the Na 1s spectra shown in Supplementary Fig. 28, a characteristic peak appears at 1071.2 eV when the sample is first charged to 0.4 V compared with the original state, suggesting the formation of Na–O and Na–V bonds in V₂O₅/V₂CO_2p. As the charge process proceeds, the Na 1s characteristic peak shifts to the high-energy region since Na acts as an electron-donating unit providing electrons. During the discharge process, the Na 1s characteristic peak exhibits a negative shift in terms of the binding energy, indicating that Na obtains electrons during the removal process. Figure 6a shows the corresponding O 1s spectra, and the intensity of the characteristic peak of the O₂²⁻ substances decreases after charging. The intensity of the characteristic peak of O²⁻ (V–O) increases, indicating that some O₂²⁻ are converted to O²⁻. The characteristic peaks of O₂²⁻ and O²⁻ then recover during the discharge process^25,26. Therefore, O₂²⁻ has highly reversible activity, which guarantees the high reversibility of Na₄V₂O₇/V₂CO_2p. Additionally, the V–O characteristic peak shifts to a lower energy region during the charge process compared to the initial state, indicating that O acts as an electron acceptor that captures electrons during Na⁺ capture, whereas Na and V act as electron-donating units providing electrons. The V–O characteristic peak positively shifts during discharge, indicating that O loses electrons during Na⁺ removal. This result is consistent with the Bader charge analysis results (Supplementary Table 6). In the high-resolution V 2p XPS spectra (Fig. 6b), the characteristic peaks of V⁵⁺ and V⁴⁺ undergo a reversible shift and a peak intensity change during the charge/discharge process, indicating that the valence state of V undergoes good reversible conversion. The peaks of V for V⁵⁺ and V⁴⁺ negatively shift during the charge process and then positively shift during the discharge process, confirming that d orbital electrons of V are obtained in the sodiation process and that d orbital electrons are lost in the Na⁺ removal process. During the charge process, V stores the charge, and the O₂²⁻ anion at the interface offers a position for storing Na⁺. During the discharge process, both the charge and Na⁺ are released. The results show that the strongly electronegative O₂²⁻ anion can be used as an additional electron storage site for Na⁺ adsorption, thus increasing the amount of Na⁺ absorbed. Figure 6c shows the high-resolution N 1s XPS spectra. The N 1s peak shifts to the lower binding energy region during the charge process because N acts as an electron donor providing additional electrons to adjacent V and O during the Na⁺ storage process. The N 1s peak returns to the initial position at the end of the discharge process, indicating good charge reversibility. The above results show that V guarantees good charge reversibility, the induced oxide-related substance O₂²⁻ can accommodate additional Na⁺, and N can provide additional storage sites for electrons, thus achieving rapid transport dynamics of Na⁺³⁶.

Fig. 6: The origin of the fast Na⁺ transport kinetics of the V₂O₅/V₂CO_2p.

XPS profiles for V₂O₅/V₂CO_2p; from top to bottom, the spectra collected for the pristine sample and the samples charged to 0.4 V, 0.6 V, and 0.8 V, discharged to 0.5 V, 0.2 V, 0 V, −0.2 V, and −0.4 V, and charged for a second time to 0.8 V. a O 1s, b V 2p and c N 1s. d–f Na⁺ diffusion pathways and corresponding migration activation energies.

DFT calculations were performed to further explore the coordination structure of Na₄V₂O₇/V₂CO_2p and reveal its effect on Na⁺ storage. Supplementary Fig. 29 shows the equilibrium structure of Na₄V₂O₇/V₂CO_2p. On the heterogeneous interface of Na₄V₂O₇/V₂CO_2p, part of the terminal functional group O_2p is shared by Na₄V₂O₇ and V₂CO_2p. The observed CDD results (Supplementary Fig. 30) show electron depletion for the Ad-Na atom (adsorbed Na) and electron accumulation on the O_2p atom, which clearly demonstrate the strong charge transfer between the two atoms, indicating the enhanced adsorption of Na⁺. To reveal the sodiation kinetics, the migration energy barriers for Na atom diffusion in Na₄V₂O₇/V₂CO_2p were studied, as shown in Fig. 6d–f. Compared with path 2, the Na atom diffusion starting point in path 1 has multiple O_2p active sites. The migration energy barrier results show that the energy barrier for Na atom diffusion through path 1 is lower (1.5 eV), which further conceptually verifies that the induced O_2p state positively contributes to improvement of the sodiation kinetics.

Discussion

In summary, a heterolattice oxygen-containing V₂O₅/V₂CO_2p heterostructure with sodium-rich storage characteristics as an efficient electrode for CDI was successfully prepared by the intercalant-induced in situ synthesis method. Owing to the space confinement effect of the solvent cage formed by the stable TMA⁺ intermediate, the oxidative nucleophilic reaction is limited, and the structural stability of the highly conductive V₂O₅/V₂CO_2p heterostructure is ensured. The introduction of O_2p and N doping induce redistribution of the charge at the V₂O₅/V₂CO_2p heterojunction interface, resulting in strong interface coupling, which greatly improves the electronic conductivity. In addition, DFT calculations confirm that the Na atom has the highest affinity for the heterojunction interface. Most importantly, the induced O_2p electrons dominate the Fermi level, which provides an additional pathway for electron movement, thereby increasing the capacity by triggering an additional redox process to form a sodium-rich vanadate heterostructure (Na₄V₂O₇/V₂CO_2p). As a conceptual verification, these characteristics endow the V₂O₅/V₂CO_2p heterostructure with notable desalination capacity and rate performance, which are significantly better than those of the commercial V₂O₅ electrode. In addition, the structural evolution and electron transfer process of the V₂O₅/V₂CO_2p electrode during the charge‒discharge process and the mechanism of the fast kinetics were deeply characterized and revealed. These findings provide opportunities for preparing layered sodium-rich metal oxide electrodes and ultimately promote their commercial application in the field of CDI.

Methods

Synthesis of V₂CF_x

V₂CF_x was synthesized by using the solvothermal method. Two grams of LiF (99%) was dissolved in 40.0 mL of HCl (36.0~38.0%), and then, 2.0 g of the V₂AlC (325 mesh, 99%) precursor was added and stirred for 30 min. Afterward, the mixed solution was transferred to a Teflon-lined stainless steel reactor and heated at 120 °C for 90 h. After natural cooling to room temperature, the sample was centrifuged and washed with deionized (DI) water until the pH approached 7. Finally, black V₂CF_x powder was obtained by drying at 60 °C for 12 h.

Synthesis of V₂CO_x-TMA

V₂CF_x (0.5 g) was added to a tetramethylammonium hydroxide (TMAOH, 97%) solution and vigorously stirred for 18 h at room temperature. The precipitate was subsequently centrifuged and washed at least three times with DI water. Finally, it was dried overnight at 60 °C. The black powder was denoted as V₂CO_x-TMA. Moreover, the V₂CO_x-TMA materials prepared with 10%, 30% and 50% tetramethylammonium hydroxide solutions were denoted as V₂CO_x-TMA (10%), V₂CO_x-TMA (30%) and V₂CO_x-TMA (50%), respectively.

Synthesis of V_xO_y/V₂CO_2p and V₂O₅/V₂CO_2p

A total of 0.2 g of V₂CO_x-TMA (10%, 30%, and 50%) was separately added to 70 mL of DI water and vigorously stirred for 2 h. Then, the mixture was transferred into a Teflon-lined stainless steel reactor and heated at 180 °C for 12 h. V_xO_y/V₂CO_2p (10%), V_xO_y/V₂CO_2p (30%) and V_xO_y/V₂CO_2p (50%) powders were obtained by centrifuging, washing three times with DI water and drying at 60 °C for 12 h to obtain the precipitate. Finally, the as-prepared V_xO_y/V₂CO_2p was annealed at 350 °C for 2 h under an air atmosphere to obtain a yellow V₂O₅/V₂CO_2p powder.

Electrochemical and desalination measurements

Both electrochemical measurement electrodes and ion-removal CDI electrodes were prepared by the following methods^37,38. Graphite paper was cut into 60 × 10 mm² and 70 × 70 mm² geometries for electrochemical and CDI tests, respectively. Then, the target samples, conductive carbon black, and polyvinylidene fluoride were mixed in a mass ratio of 8: 1: 1, during which a small amount of N-methyl-2-pyrrolidone (NMP, >99.0%) was added for mixing in order to form a mucilage. The mixed slurry was then uniformly coated onto the graphite substrate using a scraper (50 μm in thickness). The effective area of coating was 10 × 20 mm² and 60 × 60 mm² for electrochemical electrodes and CDI electrodes, respectively. Finally, the electrodes were dried in an oven at 70 °C for 12 h for spare parts. The electrochemical properties of the prepared electrodes were investigated by linear scanning voltammetry (LSV), differential capacitance curves, CV, GCD and EIS using a three-electrode system in 1.0 M NaCl solution using an electrochemical workstation (CHI 660E). The three-electrode system consisted of a working electrode, a platinum counter electrode and an Ag/AgCl reference electrode.

The ion removal performance of the electrodes was tested on a continuous cycle CDI system. The CDI cell consisted of AC anode, anion exchange membrane, sealing ring, separator, cation exchange membrane and target electrode cathode. The volume of the target brine was 40 mL. A constant-current peristaltic pump drove the brine circulation at a rate of 20 mL min^-1. The electrochemical workstation (PARSTAT 3000A-DX, Princeton Applied Research, USA) provided a stable working voltage. Meanwhile, the conductivity meter monitored the change of brine conductivity in the vessel in real time. The desalination capacity of the active substance for unit mass of salt ions was calculated according to Eq. (3).

$$\varGamma=\frac{({G}_{0}-{G}_{t})V}{2m}$$

(3)

where Γ (mg g⁻¹): desalting capacity; G₀, G_t (μS cm⁻¹) electrical conductivity of the salt solution at the initial and saturation stages; V (L): total volume of the salt solution; m (g): effective mass of the active material.

The charge efficiency (Ʌ) is quantitatively used to determine the utilization rate of electric energy, which is calculated by Eqs. (4) and (5):

$$\Lambda=\frac{(\Gamma {{\rm{\times }}}F)}{M\varSigma }$$

(4)

$$\varSigma=\frac{\int {idt}}{m}$$

(5)

where F: Faraday constant (96485 C g⁻¹); M (58.44 g mol⁻¹): molar mass of NaCl; I: current; m (g): effective mass of the active material.

The water recovery (WR, %) is obtained according to Eqs. (6)–(8):

$${V}_{d}={\int }_{\Delta {t}_{d}}{Qdt}$$

(6)

$${V}_{c}={\int }_{\Delta {t}_{c}}{Qdt}$$

(7)

$${WR}=\frac{{V}_{d}}{{V}_{d}+{V}_{c}}$$

(8)

where V_d (V): volume of desalinated water; Q (mL min⁻¹): flow rate; Δt_d (s): desalination durations; V_c (V): volume of the regenerated water; Δt_c (s): the time to collect the regenerated water.

The volumetric energy consumption (E_v) of the CDI device is defined as the following Eqs. (9)–(11):

$${E}_{{in}}={\int }_{\varDelta {t}_{{cycle}}}{IVdt},{{\rm{w}}}{{\rm{h}}}{{\rm{ere}}}\; {{IV}} \, > \, 0$$

(9)

$${E}_{{out}}={\int }_{\varDelta {t}_{{cycle}}}{IVdt},{{\rm{w}}}{{\rm{h}}}{{\rm{ere}}}\; {{IV}} \, < \, 0$$

(10)

$${E}_{v}=\frac{{E}_{{in}}-\eta {E}_{{out}}}{{V}_{d}}$$

(11)

in which, IV (W): the product of the current and voltage in the desalination process; t_cycle (min): the total time under dynamic-steady state (DSS) during desalination; E_in (J): the total energy input in the DSS cycle; E_out (J): the total recoverable energy of the CDI unit in the desalination process; E_v (Wh m^-3): volumetric energy consumption. η is a part of E_out; which actually represents the next charging stage of recycling and reuse. Theoretically, the η approximately equals to 1.

Theoretical calculations

All DFT calculations were performed with the Vienna Ab Initio Package (VASP 6.3.2) in the generalized gradient approximation (GGA) using the Perdew–Burke–Ernzerhof (PBE) formula^39,40,41. The ion nucleus was described by the projection enhanced wave (PAW) potential, and the valence electrons were taken into account using a plane wave basis set with a kinetic energy cutoff of 520 eV^42,43. The use of Gaussian smearing and a width of 0.2 eV allows partial occupancy of the Kohn-Sham orbitals. When the energy change was less than 10⁻⁵ eV, the electron energy was considered to be self-consistent. Geometric optimization was considered convergent when the energy change was less than 0.02 eV Å⁻¹. The vacuum gap in the direction perpendicular to the structural plane was 18 Å. The weak interaction was described by the DFT + D3 method using empirical corrections in the Grimme’s scheme^44,45. Geometry optimization was performed via a gamma centered k-point. To consider the strong correlation effects of transition metal in structure, all calculations were carried out by using the spin-dependent GGA plus Hubbard correction U method, and the effective Ueff parameters are 4.0 eV for V^46,47,48.

Reporting summary

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