It is estimated that 40% of the global population faces severe water scarcity. Recycling sewage and seawater desalination have become the key to solving this problem [1-4]. Currently, the traditional desalination methods are dominated by membrane separation technology [5,6] and thermal separation dominates [7], while the emerging methods of capacitive deionization, membrane method, and thermal method are still in the research stage [8]. However, these technologies are faced with the disadvantages of membrane pollution and high cost.

Capacitive deionization (CDI) is considered a clean, efficient, low-cost, and promising seawater desalination technology [9-14]. In contrast to membrane technologies such as reverse osmosis (RO), forward osmosis (FO), and membrane distillation (MD), CDI technology can achieve regeneration through charging and discharging electrodes [15-17]. The properties of electrode materials determine the principle. At present, carbon-based materials mainly correspond to the principle of electric double layer [18], which mainly include one-dimensional carbon nanotube material [19,20], two-dimensional graphene material [21,22], three-dimensional activated carbon material [23], etc. However, the structure and composition of these materials result in unsatisfactory CDI performance [24]. Studies have found that under the same electrode area, the capacitance of pseudocapacitance material can reach 10~100 times of electric double layer materials [25-29]. Therefore, layered bimetallic hydroxide (LDHs), MXenes [30-32], metal-organic framework (MOFs) [33,34], and other pseudocapacitive materials [35] have been used by more and more people because of their excellent physical and chemical properties.

MOFs quickly gained attention due to their highly designable structure, high electron mobility, high specific surface area, and highly adjustable porosity, and have been widely used to prepare porous carbon precursors and template materials [36-42]. However, since MOFs are unstable in aqueous solution [43,44]. Some researchers try to pyrolysis MOFs in an oxygen-free atmosphere [45] to apply them to aqueous environments. Currently, HKUST-1-derived carbon materials have been applied and studied in the fields of catalysis and adsorption, but there are few applied types of research in the field of CDI, especially the explanation and exploration of its mechanism.

Herein, HKUST-1 was synthesized by solvothermal method, and a MOFs-derived porous carbon/Cu@Cu₂O material (HDC-X) containing more complex porosity and more active sites was prepared by simple pyrolysis at different temperature. This material was used as cathode material to conduct a desalting experiment with activated carbon (AC) anode. The cyclic stability of the materials was tested, and the desalting mechanism was discussed by comparing the desalting properties of carbonized materials at different temperatures.

Deionized water was used in all experiments. All drugs were analytically graded or above. They can be used directly without additional processing. Specific information on material preparation, characterization methods, experimental analysis methods, electrode preparation, and device construction is included in the supporting information (Fig. S1 in Supporting information). The material synthesis process is shown in Fig. 1a.

After carbonization at different temperatures, the derived carbon material and the precursor HKUST-1 show different characteristics in morphology. The crystal size of HKUST-1 is between 5 and 10 µm, with an octahedral shape and clear edges and corners (Fig. 1b). After pyrolysis at 350℃ (Fig. 1c), the precursor skeleton maintains a complete regular octahedral shape, but many small spheres appear on the surface. After pyrolysis at 1100℃, some small spheres and pits appeared on the surface of the collapsed HKUST-1 skeleton (Fig. 1d). This is because the copper in the precursor melts at high temperatures, balls up and partially falls off when it is re-cooled. HRTEM results showed that Cu accumulated and agglomerated on the surface of HKUST-1 crystal at 350℃ (Fig. 1e). The high temperature causes the copper to redistribute randomly in the precursor frame. Part of the resolidified copper sphere falls off from the surface of the precursor, leaving a vacancy on the crystal surface, and the overall skeleton retains the porous structure of the MOF precursor and exposes more active sites. HKUST-1 has good hydrophilicity, but it is unstable in water. The porous carbon/Cu@Cu₂O derived from MOFs after pyrolysis shows hydrophobicity (Fig. S2 in Supporting information), which significantly improves its poor water stability. In summary, although pyrolysis reduces the hydrophilicity of the material, the precursor skeleton is retained and more pores are exposed at unsaturated Cu sites, which is conducive to capacitive deionization.

XRD analysis can further determine the composition of the material and the crystal structure of each component. In the X-ray diffraction pattern of HKUST-1 (Fig. 1f), the diffraction peaks at 11.6°, 19.0°, and 25.9° are obvious, corresponding to the crystal planes (222), (440), and (731) of HKUST-1 [46], respectively. After pyrolysis, prominent diffraction peaks appear at 74.13°, 50.43° and 43.3°, which correspond to plane (222), (220) and (111) of Cu (JCPDS No. 04–0836) [47], respectively. A weak Cu₂O peak was also observed in the XRD pattern (Fig. S3 in Supporting information), and the diffraction peak at 36.4° corresponded to the (111) crystal plane of Cu₂O. It can be found that higher temperature makes the crystallinity of Cu and Cu₂O higher.

According to thermogravimetric analysis (TGA) results, when the temperature reaches 350℃, the first stage of water loss of HKUST-1 stops, and the thermogravimetric loss gradually stabilizes around 500℃ (Fig. 1g). From the proportion of CuO (Fig. S4 in Supporting information), it can be calculated that the mass proportion of Cu in HDC-1100 is 71.66%, which is a considerable proportion. In the MOFs-derived porous carbon/Cu@Cu₂O composite material, Cu does not contribute specific surface area, but provides corresponding active sites and other properties such as electrical conductivity.

The pore size distribution and specific surface area of the material were calculated by the isothermal adsorption and desorption curve of nitrogen. The isotherms of HDC-350 and HDC-1100 exhibit type-IV curve. The rise is obvious at a higher relative pressure (P/P₀ ≥ 0.8), showing the H1-type hysteresis line, indicating that there are more mesoporous in the material structure (Fig. 2a). The specific surface area calculation (Table S1 in Supporting information) shows that high temperature causes partial collapse of organic skeleton structure. However, with the increase of pyrolysis temperature, SSA becomes larger, and the structure changes from micropore to the coexistence of micropore and mesoporous (Fig. 2b). The high surface area and complex pore structure enhance the ion adsorption capacity.

FTIR is used to analyze the types of functional groups in materials. The HKUST-1 precursor has a large number of groups (Fig. 2c). However, after carbonization, the surface functional groups of the materials show great differences (Figs. 2d and e). HDC-1100 has a broad absorption peak between 3500–3200 cm⁻¹, similar to that of the precursor, which is the stretching vibration of -OH, but the strength is somewhat reduced. The characteristic peaks of oxygen-containing functional groups appear at 1100 cm⁻¹ (C-O) and 1727 cm⁻¹ (C=O). These oxygen-containing functional groups can introduce additional pseudo-capacitors that contribute to the removal of Na⁺ [48].

XPS analysis is helpful to elucidate the morphology and chemical bond changes of surface elements. The XPS spectra of the two materials are similar, as shown in Fig. 2f, the material has obvious peak signals at 932.54 eV, 531.44 eV, and 284.81 eV, corresponding to the typical characteristic peaks of Cu 2p, O 1s, and C 1s [49]. Figs. 2g and h XPS peak fitting analysis of Cu 2p shows that Cu presents different forms of zero valences, monovalence, and bivalence during pyrolysis, and exists in the form of Cu⁰, Cu₂O, and CuO, respectively. However, compared with the content of Cu⁰, the content of Cu₂O and CuO is very small or mostly exists in the amorphous state. This is consistent with the previous analysis results of the XRD pattern [50,51]. The obvious Cu⁰ peak of HDC-1100 indicates that a large amount of Cu resolidifies on the crystal surface after melting, which is consistent with the previous analysis. XPS analysis of O 1s can indicate various oxygen-containing functional groups in the material. Although higher temperatures cause the loss of oxygen atoms, there are still plenty of oxygen-containing functional groups in the material (Figs. S5a and c in Supporting information). These functional groups can provide part of the additional pseudocapacitance, and also contribute to the contact between materials and inorganic salts, which has a positive effect on the electro-adsorption [46,48]. In addition, the presence of oxygen-containing functional groups can prevent Cu agglomeration and make Cu uniformly distributed in the material, which improves the removal rate. The appearance of π-π^* bond (290.0 eV) proves the effective binding of HKUST-1 to H₃BTC (Figs. S5b and d in Supporting information). More thorough carbonization makes HDC-1100 form more conjugate structures and constitutes more π-π^* bonds. This allows for more free electrons in the material, improving its electrical conductivity [52].

Cyclic voltammetry (CV) and galvanostatic charge-discharge (GCD) were used to test the electrode materials. The CV curve of HDC-350 is not an obvious rectangular shape, but an olive shape, showing a weak redox peak in the complete CV curve, and the voltage difference between the two peaks is very small (Fig. 3a). Under the same scan rate conditions, the redox peak of HDC-1100 is more obvious. And the current intensity of the CV curve is stronger, indicating that in the pseudocapacitive mechanism, the redox reaction plays a more important role (Fig. 3b). This indicates that the HDC-X electrode is a pseudocapacitance electrode material, as noted in the results of CV curve analysis. Within 50 cycles, the CV curve of HDC-350 did not show obvious shape change, which indicated that the initial capacity of the HDC-350 electrode was always maintained during the long cycle with almost no attenuation (Fig. 3c). In the same way, after 50 cycles, the shape of the CV curve remains unchanged, and the HDC-1100 electrode maintains good capacitance stability (Fig. 3d). When the current densities are 100, 200, 500 and 1000 mA/g, the specific capacities of HDC-350 electrodes are 10.34, 9.71, 8.75 and 7.11 mAh/g, respectively (Fig. S6a in Supporting information). As the current density increases, the specific capacity of the material decreases continuously. At low current densities, the charging and discharging time will be longer, and there will be more active sites to store electrons so that the electrochemical capacity will be higher at low current densities. In addition, the high current density will cause water decomposition, resulting in energy waste and electrode damage. The specific capacities of HDC-1100 electrodes are 16.67, 12.51, 10.86 and 8.52 mAh/g when the current density is 100, 200, 500 and 1000 mA/g, respectively (Fig. S6b in Supporting information). Compared with HDC-350, under the same current density, HDC-1100 has a larger specific capacity, which is one of the reasons why HDC-1100′s desalination efficiency is stronger than HDC-350. At different current densities, the GCD patterns of HDC-350 show similar triangular shapes, and no obvious redox platform appears, indicating that the redox reaction occurs on the surface of the material rather than the bulk phase, which is neither the behavior of the battery nor the behavior of the electric double layer principle (Fig. 3e). There is no obvious redox platform in the GCD curve of HDC-1100, and it is not a completely symmetrical triangle, indicating that the HDC material is neither an electric double layer material nor a battery material, but a pseudocapacitor material (Fig. 3f). When the current density is 200 mA/g, there is almost no difference between the GCD diagram of the two materials after 20 cycles of testing and the GCD diagram of the first test, indicating that the electrode material is stable (Figs. S6c and d in Supporting information).

The power supply was set to the constant current mode, and the voltage window was set to −1.2 V ~ 1.2 V. The desalting performance of HDC-350 and HDC-1100 was tested at 20, 40, and 60 mA/g current densities. The average desalination capacities of HDC-350 at different current densities are 37, 16.1 and 13.8 mg/g, respectively (Fig. 4a). The desalting capacity decreases with the increase of current density. This is because it takes less time for the system to rise to the cut-off voltage as the current density increases for the same voltage window. In a short period, some of the sodium ions in the NaCl solution and chloride ions have no time to move to the surface of the electrode material, resulting in a low adsorption capacity of the electrode material. Higher current density results in a larger diffusion limit and fewer active sites available, resulting in a lower capacity. At the same time, due to the short contact time, the ions moving to the electrode surface do not have enough time to move to the interior through the channel of the material. They are desorbed back into the solution, and the internal diffusion takes longer time than the surface adsorption, which also explains the phenomenon that the desalination rate is fast when the current density is large in Fig. 4a. Higher current density means faster charge transfer, resulting in a higher desalination rate. Similar to the results of the HDC-350 electrode, the desalination capacity of HDC-1100 gradually decreased with the increase of current density, but the desalination rate gradually increased (Fig. 4b).

Besides the current density, the setting of the voltage window also makes a great difference in the entire desalination process. Under the current density of 40 mA/g, when the voltage window is set to −1.0 V ~ 1.0 V, the desalination capacity of HDC-350 is only 6.8 mg/g. When the voltage window is increased to −1.4 V ~ 1.4 V, the desalination capacity expands nearly three times to 20.4 mg/g (Fig. 4c). The same pattern was observed for the HDC-1100, where the desalting capacity increased with the increase of applied voltage (Fig. 4d). The overall charge and discharge time will be improved when the voltage increases, which facilitates the movement of sodium ions and chloride ions in the solution to the surface or even inside the electrode material. As the amount of charge increases, more of the charge on the electrode is involved in the desalination process, which means that there is a stronger attraction to the ions with different charges in the solution, and the driving force for the movement of chloride ions and sodium ions to the electrode is stronger. In addition, the internal resistance of the device also consumes part of the charge, so there is less desalination capacity in the low-voltage window. The difference is that the desalination capacity of the HDC-1100 electrode is stronger, which is about twice that of the HDC-350 electrode under the same conditions. This is because the specific surface area, mesoporous number, and pseudocapacitance strength of the HDC-1100 electrode are better than those of the HDC-350 electrode. And it may be due to the melting and resolidification of copper in HKUST-1 at 1100 ℃, which attenuates the agglomeration of Cu on the crystal surface during pyrolysis at lower temperatures, reduces the blockage of pores, and increases the number of active sites. Besides, compared with the lower current density, the energy consumption almost doubles when the current density rises to 60 mA/g (Figs. S7a and b in Supporting information). Excessive voltage will lead to electrode polarization and water decomposition, which will consume additional energy and is not conducive to the expansion of the process.

The long cycle life of electrode materials is another important indicator that affects their application in the field of desalination. The desalination long-cycle performance of HDC-350 was tested, and it was cycled for 100 cycles under the conditions of −1.2 V ~ 1.2 V, 40 mA/g. Within 100 cycles, the desalination capacity of HDC-350 was stable, maintained between 13 and 19 mg/g, which indicated that the HDC-350 electrode had good desalination stability (Fig. 4e). In the long-cycle experiment, the HDC-1100 electrode maintained a relatively stable desalination capacity within 20 cycles (Fig. 4f), but there was a certain decay trend, which was related to the obvious redox peak of its constant current charge and discharge. The pseudocapacitance behavior of HDC-1100 is more prominent, so HDC-1100 has less cycling stability than the HDC-350 electrode. At higher temperatures, more Cu is exposed, which is more prone to redox reactions, and also makes the structure more fragile.

According to the above experimental results, the desalting mechanism is summarized as follows. The cathode is a porous carbon/Cu@Cu₂O electrode derived from MOFs, and the anode is an AC electrode. In the charging process, under the action of the potential difference between the two ends, Na⁺ moves to the cathode through the cation exchange membrane. HDC-1100 has the property of pseudocapacitance and has a suitable pore structure and adsorption site. The Na⁺ in the solution can be effectively adsorbed, and the activated carbon electrode will embed Cl⁻ in the pore of the activated carbon according to the electric double layer model, to reduce the internal conductivity of the device and achieve the purpose of desalting. When Cu⁺ or Cu²⁺ ions are leached in the process of charging and discharging at high potential, they are immediately captured by the cation exchange membrane at the cathode, which makes the ion concentration at the cathode much higher than that of the bulk solution, resulting in higher ion migration efficiency. This may be the main reason for the improved desalination capacity of HDC-1100 after cycling.

In this study, a composite of porous carbon/Cu@Cu₂O was synthesized, and the pyrolyzed material improved the unstable properties of HKUST-1 precursor in water. Through the comparison of materials at different pyrolysis temperatures, it is found that the material HDC-1100 at high-temperature pyrolysis has a larger specific surface area, more active sites, and more abundant oxygen-containing functional groups. Pseudocapacitance is the main mechanism of desalination. Its strong desalination ability and simple preparation method make it a promising CDI electrode material. It can be seen that MOF-derived carbon materials can play a greater role in the future and be applied to more fields.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

This research is supported by The National Natural Science Foundation of China (Nos. 22276137, 52170087). We are also thankful to the anonymous reviewers for their valuable comments to improve this manuscript.

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.cclet.2023.108362.