Three-dimensional network of graphene for electrochemical capacitors and capacitive deionization

As global climate change intensifies, sustainable development has risen to the forefront as a critical global concern.¹ Dependence on traditional energy resources, especially fossil fuels, has resulted in substantial greenhouse gas emissions, exacerbating global warming and environmental pollution.² In this context, carbon neutrality has become an important goal.³ Traditional renewable energy sources, such as solar and wind energy, have their limitations due to their fluctuating nature, making it challenging to provide a steady energy supply.⁴ To address this issue, developing energy storage technologies has become a key research direction. Energy storage technologies can store energy as it’s generated and release it when needed, achieving a reliable and stable energy supply. Additionally, energy storage technologies can help address the temporal and spatial mismatch of energy sources, making renewable energy more reliable and accessible.⁵ Among energy storage technologies, supercapacitors with high-performance have garnered attention. Compared to traditional batteries, supercapacitors offer advantages such as high power density, rapid charge-discharge rates, and a long lifespan.⁶ These characteristics make supercapacitors excellent at energy storage and release, fulfilling the specific demands of applications like electric vehicles, smart grids, and renewable energy systems.⁷ 

At the same time, climate change has led to frequent droughts along with an unstable water supply, while human activities have exacerbated water scarcity.⁸ The ocean is abundant in water resources, but its high salt content renders it unsuitable for direct use in human life and agriculture. Therefore, the conversion of seawater into freshwater to make it a useable resource holds significant importance.⁹ Until now, desalination technologies mainly include distillation and membrane separation methods. Although these technologies are effective, they have high energy consumption, complex facilities, and high costs. Therefore, there is a pressing need for ongoing technological innovation aimed at enhancing energy efficiency and cost reduction.¹⁰ Compared to traditional desalination technologies such as distillation, reverse osmosis, nanofiltration, and capacitive deionization (CDI), they present the advantages of low energy consumption, a simple and compact device structure, easy and controllable operation, no secondary pollution, and strong mobility.¹¹ Furthermore, this technology can be combined with renewable energy sources, such as solar and wind energy, to lower energy consumption and enhance the overall environmental friendliness and sustainability of the desalination process.¹² Therefore, CDI technology can not only be applied on a large scale in coastal areas but is also very suitable for brackish water desalination in remote areas, exhibiting promising development and application prospects.

In supercapacitors and capacitive desalination technology, materials are one of the most critical factors. Carbon materials, as one of the most common materials, are widely used as electrode materials due to their high conductivity, diverse structures, excellent chemical stability, and low cost.¹³ Various forms of carbon materials have been utilized, including zero-dimensional (0D) carbon spheres,¹⁴ one-dimensional (1D) carbon fibers (CF) and carbon nanotubes (CNT),¹⁵ two-dimensional (2D) graphene¹⁶ and carbon nanosheets,¹⁷ and three-dimensional (3D) carbon foam¹⁸ and carbon cloth.¹⁹ Among them, 3D graphene has become a research hotspot in the fields of supercapacitors and CDI due to its high specific surface area, excellent conductivity, and tunable pore structure.²⁰ 3D graphene can be categorized into macroscopic and microscopic forms. Macroscopic 3D graphene typically consists of integrated layers of a few-layer graphene sheets and can reach sizes of several centimeters or larger. In contrast, microscopic 3D graphene has dimensions typically ranging from tens of nanometers to a few micrometers and appears as a powder-like material, but both possess rich porous structures.²¹ For supercapacitors, macroscopic 3D graphene is often used directly as an electrode, saving further processing steps. Consequently, this article primarily focuses on macroscopic 3D graphene. 3D graphene cannot only improve the energy storage capacity of supercapacitors and the efficiency of capacitive desalination but also their performance stability.²² However, 3D graphene still faces some challenges and shortcomings in the application process, such as high preparation costs and difficulties in designing specific pore sizes and pore structures, which require further exploration into their preparation methods, material characteristics, and actual application performance.

Therefore, as shown in Fig. 1, this review will start with the application of 3D graphene in supercapacitor energy storage and capacitive desalination. It systematically assesses their preparation methods, modification methods, structural characteristics, performance, and application prospects, aiming to provide useful references for their further development.

A supercapacitor, also known as an electrochemical capacitor or ultracapacitor, is an energy storage device that offers high capacitance, high power density, and long cycling life. Its principle is based on the charge adsorption/desorption at the electrode surfaces and the migration of ions in the electrolyte.²³ The structure and charging/discharging process of a supercapacitor are illustrated in Fig. 2. The electrolyte between the electrodes can be liquid or solid and typically contains an ion solution or an ion-conductive polymer.²⁴ Electrode materials usually possess the characteristics of a high specific surface area and electrical conductivity. The energy storage mechanisms of supercapacitors can generally be classified into two types: the electric double-layer capacitance (EDL), which is generated by the accumulation of static charges at the electrode interface, and the pseudocapacitance, resulting from fast and reversible surface redox reactions at specific potential ranges.²⁵ EDL refers to the process of charge adsorption/desorption on the electrode surfaces. Typically, positive ions are attracted to the vicinity of the negatively charged electrode and negative ions are attracted to the vicinity of the positively charged electrode when electrode is immersed in an electrolyte solution (typically an ionic solution). As ions from the electrolyte accumulate on the electrode surfaces, they form two layers of opposite charges: a layer with positive ions (cations) near the negatively charged electrode and a layer with negative ions (anions) near the positively charged electrode. These two layers of opposite charges constitute the “double layer,” thereby creating a capacitive effect for storing electrical energy. Unlike EDL, pseudocapacitance involves reversible redox reactions occurring on the surface of electrodes. These reactions typically occur on electrode materials with specific surface functional groups, such as metal oxides or conductive polymers. In pseudocapacitance, charge is stored not only through charge distribution in the electrolyte but also through charge adsorption/desorption via the oxidation–reduction reactions taking place on the electrode surface. Through these redox reactions, charge transfer occurs at the electrode surface, further enhancing the capacitive effect and providing additional energy storage capacity.²⁶ 

The CDI method is an emerging and promising technology for desalination that has made significant progress since the 21st century. Similar to the energy storage mechanism of supercapacitors, the principles of CDI for desalination also involve EDL capacitance and pseudocapacitance.²⁷ As shown in Fig. 2, an external voltage or current stimulus is applied between two electrodes, creating an electric field between them. This electric field enables the capacitive deionization and adsorption of ions as the salt solution flows through the electrode gap. The cathode adsorbs anions while the anode captures cations, effectively declining the concentration of charged ions in the salt solution. When the electrode adsorption is saturated, the circuit is short-circuited, or a reverse voltage is applied, then the absorbed ions in the electrode will desorb into the solution under the influence of the electric field, facilitating electrode regeneration. Hence, the method of voltage application²⁸ and the characteristics of electrode materials are important factors that affect the desalination performance of CDI. Generally, a higher voltage implies a more potent electric field strength, resulting in greater forces acting on the ions and consequently accelerating both the adsorption and desorption processes. The thickness of EDL is also improved, resulting in a higher desalination capacity. For matching the ion movement and residence time as well as reducing energy consumption, voltage stimulation duration and pattern often need to be optimized to suit the CDI device. Apart from voltage, other operating parameters, such as the distance between anode and cathode, solution concentration, and liquid flow rate, also play a crucial role in ion diffusion dynamics,²⁹ thus affecting the adsorption/desorption rate of the electrode. Therefore, researchers have made many innovative designs for CDI devices and developed various CDI structural types, such as flow-through CDI,^30,31 membrane CDI,³² and flow CDI.³³ Limited by the decomposition voltage of water (1.23 V),³⁴ the exploration of electrode materials has become a more promising research direction. In general, the ideal porous materials for capacitive deionization should possess the following characteristics: good chemical stability, a high specific surface area, excellent hydrophilicity, superior capacitive adsorption performance, and high voltage resistance.^35,36

To more fairly evaluate the desalination ability of electrode materials in CDI, a series of standardized parameters such as salt adsorption capacity (SAC), average slat adsorption rate (ASAR), cycle stability, charge efficiency, and energy consumption were proposed.^27,30 SAC is a parameter to describe the mass/moles of ions or the mass/moles of salt adsorbed by the electrode per unit mass or volume, reflecting the ability of the electrode material to store ions. When considering the desalination speed of the CDI apparatus, ASAR is introduced and defined as salt adsorption capacity per unit time (SAC/t). Cycle stability assesses the SAC remaining rate after a certain cycle, demonstrating the desalination stability of the electrode material. Charge efficiency illustrates the energy efficiency of electrode materials in adsorbing ions. It is determined by dividing the amount of charge needed for adsorbing ions by the quantity of charge input into the electrode.²² Energy consumption demonstrates the energy consumed per mass/mole of salt adsorbed, which can be used to compare energy consumption with other desalination technologies and provide a cost reference for practical applications.

Graphene possesses numerous advantages such as a high specific surface area, ultra-high electrical conductivity, excellent mechanical properties, and high chemical stability, making it highly promising for applications in the field of energy storage, particularly in capacitors.³⁷ Stoller³⁸ and colleagues were the first to apply graphene to supercapacitors for energy storage. However, due to π-π interactions and van der Waals forces in graphene’s inherent 2D structure, it tends to stack together, leading to a reduction in its electrochemical performance. Ibrahim et al.³⁹ prepared graphene materials with varying numbers of layers and found that the specific capacitance of eight layers of graphene was significantly lower than that of two layers. Therefore, reducing the stacking of graphene sheets can effectively improve the performance of supercapacitors, and preparing 3D graphene is an effective way to reduce the stacking of graphene. 3D graphene is typically prepared using various methods, such as chemical vapor deposition (CVD), chemical reduction, foaming methods, and template methods. These methods enable the formation of interconnected graphene structures in 3D space, creating a rich pore structure. This porous structure provides more surface area, increases active sites for electrochemical reactions, and enhances charge transfer rates.⁴⁰ 

CVD is a commonly used method for preparing 3D graphene, typically employing copper or nickel as a substrate. Carbon source gas is deposited onto the substrate under controlled conditions, and after removing the substrate, graphene is obtained.⁴¹ The key to preparing 3D graphene through CVD lies in controlling the reaction conditions to achieve high-quality graphene growth. This involves selecting suitable substrate materials, optimizing a temperature control program, controlling gas flow, and ensuring appropriate carbon source decomposition and deposition rates. By adjusting these parameters, 3D graphene materials with the desired morphology can be obtained.⁴² Zheng et al.⁴³ successfully grew a binder-free and highly conductive 3D graphene network on nickel foam using camphor as the raw material, following the method illustrated in Fig. 3(a), which exhibited a high specific capacitance of 130 F g⁻¹ as a supercapacitor electrode. Zang et al.⁴⁴ utilized CH₄ as the carbon source and synthesized a graphene fabric through the CVD method on a copper mesh, showing a high specific capacitance of 267 F g⁻¹ and excellent cycling performance with a capacitance retention rate of 100% after 1000 cycles.

In addition to CVD, the chemical reduction of oxidized graphene is also commonly used to prepare 3D graphene. While CVD can produce single-layer graphene of high quality, its high cost and complex equipment make large-scale production challenging. The chemical reduction of oxidized graphene provides an ideal solution to these issues.⁴⁵ Gao et al.⁴⁶ used L-glutathione to reduce oxidized graphene, producing a 3D graphene hydrogel with interconnected porous structures. As shown in Figs. 3(b) and 3(c), symmetric supercapacitors made from this material exhibited a specific capacitance of 157.7 F g⁻¹ at a current density of 1 A g⁻¹. Xu et al.⁴⁷ prepared graphene hydrogel as a flexible electrode for supercapacitors. The graphene hydrogel demonstrated a highly interconnected 3D network structure, offering excellent electrical conductivity and mechanical stability, making it an outstanding flexible energy storage material. Its specific capacitance reached ∼180 F g⁻¹, and after 10 000 cycles, the capacitance only decayed by 8.4%. Luan et al.⁴⁸ prepared a reduced graphene oxide (rGO) hydrogel through cross-linking with ethylenediamine, followed by hydrazine reduction. When used as an electrode for supercapacitors, it exhibited a high specific capacitance of 232 F g⁻¹ at a current density of 1 A g⁻¹. Shi et al.⁴⁹ prepared a relatively well-ordered 3D graphene through hydrothermal treatment of a hydrochloric acid-treated graphene oxide (GO) solution. The supercapacitors based on reduced graphene exhibited a high capacitance, achieving a specific capacitance of 220 F g⁻¹ at a current density of 2 A g⁻¹ and maintaining over 80% of the initial capacitance after 10 000 cycles.

In addition to (CVD) and chemical reduction methods, foaming methods are also commonly used to prepare 3D graphene due to their simplicity, controllable structure, and other advantages. Wang et al.⁵⁰ used glucose as a precursor and NH₄Cl as a foaming agent to prepare 3D graphene bubble networks. As shown in Fig. 4(c), the walls of these bubble networks are usually very thin, comprising only a few graphene layers, resulting in a large specific surface area and low electron conduction resistance. The specific capacitance reached 163 F g⁻¹ at a current density of 10 A g⁻¹. Wang et al.⁵¹ successfully prepared zinc-guided 3D porous graphene through a zinc-assisted solid-phase pyrolysis route. The graphene exhibited a high specific surface area and abundant microporous and mesoporous structures, demonstrating good specific capacitance and cycling stability as an electrode material for supercapacitors. In a neutral electrolyte, the specific capacitance was 110 F g⁻¹ at 8 A g⁻¹, with a retention rate of 93% after 10 000 cycles. Jiang et al.⁵² used sucrose as a precursor and ammonia-assisted chemical foaming to successfully prepare self-supporting 3D porous graphene. The material was composed of interconnected graphene membranes with a large surface area and suitable porosity. It showed a specific capacitance of 136.6 F g⁻¹ at 10 A g⁻¹ and retained 88% after 1000 cycles. Hao et al.⁵³ prepared 3D porous graphene using GO as a precursor and NH₄Cl as a foaming and reducing agent. This structure effectively overcame the restacking and agglomeration issues of graphene nanosheets. At a current density of 1 A g⁻¹, the specific capacitance reached as high as 231.2 F g⁻¹. Furthermore, the capacitance could retain over 99% after 8000 cycles.

Template methods are also commonly used to prepare 3D graphene with advantages such as simplicity and controllable porous structures, and the commonly used templates include nickel foam or copper foam. Shin et al.⁵⁴ utilized agarose gel as a soft template and prepared 3D wrinkled graphene through sequential steps of gelation, freeze-drying, and calcination. The specific capacitance of this material can reach 56.6 F g⁻¹. Ping et al.⁵⁵ prepared a large amount of graphene using an electrochemical exfoliation method and then utilized nickel foam as a template to create a 3D graphene foam. This 3D graphene foam electrode exhibited superior electrochemical performance, displaying a specific capacitance of ∼128 F g⁻¹ in a 6M KOH electrolyte at a current density of 1 A g⁻¹. Yu et al.⁵⁶ successfully constructed a hierarchical porous graphene foam/polyaniline nanowire array through the combination of template and printing techniques, which promotes the rapid transfer path of ions and electrons. The prepared graphene foam/polyaniline nanowire array exhibited a specific capacitance of 939 F g⁻¹ at 1 A g⁻¹ and retained 88.7% after 5000 cycles.

In comparison with traditional methods, direct laser writing (DLW) is an efficient, high-resolution, mask-free, and low-cost processing method.^57,58 Liu et al.⁵⁹ achieved the simultaneous induction and activation of 3D porous graphene by direct laser writing on polyimide films coated with KOH at room temperature. The laser-induced and activated graphene-based planar micro-supercapacitors exhibited high areal capacitance (32.00 mF cm⁻² at 0.05 mA cm⁻²), about 10 times higher than non-KOH-activated ones, and retained 95.73% capacitance after 6000 cycles. Zhang et al.⁶⁰ used amorphous carbon nanoparticles as precursors to prepare graphene electrodes dominated by mesopores and macropores by laser writing. The material exhibited a specific capacitance of 0.498 mF cm⁻² at a current density of 0.078 mA cm⁻². After 12 000 cycles, it retained 98.6% of the initial capacitance at a current density of 1.56 mA cm⁻². Xu et al.⁶¹ introduced a laser-assisted transfer printing method that significantly improved the success rate of the transfer. The prepared laser-scribed graphene (LSG) material was used to manufacture stretchable supercapacitors. At a current density of 0.02 mA cm⁻², the capacitance can reach 18 mF cm⁻². Furthermore, LSG supercapacitors exhibited excellent cycling stability, offering prospects for the development of wearable electronic devices. However, this technique also has some challenges, such as higher requirements for laser equipment, slower processing speeds, and the need for further research in material selection and optimization.

In capacitor applications, pure 3D graphene can be further modified in some aspects, such as tapping density and conductivity.⁶² Low tapping density is an inherent limitation of porous materials, which generally results in lower energy density in graphene-based materials.⁶³ To overcome these limitations, researchers have explored various methods, including composite materials. Incorporating composite materials into 3D graphene can enhance its tapping density and improve its conductivity.⁶⁴ The most common composite materials involve metal oxides or hydroxides, such as MnO₂,⁶⁵ Ni(OH)₂,⁶⁶ NiO,⁶⁷ and Co₃O₄,⁶⁸ which offer lower costs and excellent redox properties. These materials substantially enhance the capacitance of graphene.⁶⁹ Additionally, metal sulfides are often used in combination with graphene. Compared to metal oxides, metal sulfides exhibit higher conductivity and can undergo more redox reactions.⁷⁰ Conductive polymers used as doping materials can offer excellent electrochemical activity, tunability, good cycling stability, and low cost in supercapacitors.⁷¹ In addition, some novel materials such as MOFs, MXenes,⁷² carbon nanotubes, and carbon fibers are also utilized as dopant materials for graphene due to their unique structures and abundant functional groups on the surface.

Bai et al.⁷³ synthesized an in situ 3D graphene/MnO₂ foam composite using a combination of CVD and hydrothermal methods. The composite exhibited high crystallinity and low contact resistance, enhancing charge transfer efficiency. This material showed high specific capacitance (333.4 F g⁻¹ at 0.2 A g⁻¹) and outstanding cycling stability (92.2% retention after 2000 cycles). Dong et al.⁷⁴ used CVD to synthesize graphene foam on nickel foam and in situ grew Co₃O₄ nanowires on the graphene foam via a hydrothermal process. The graphene foam provided efficient conductive pathways, ensuring fast charge transfer and conduction. Benefiting from the excellent electrochemical and electrocatalytic effects of Co₃O₄ nanowires, the specific capacitance of graphene foam @ Co₃O₄ nanowires can reach 768 F g⁻¹ at a current density of 10 A g⁻¹ and increase to around 1150 F g⁻¹ after 1000 cycles. This phenomenon may result from the more complete intercalation and deintercalation of electrochemical species after some initial cycles. Cao et al.⁷⁵ utilized ethanol as the carbon source and prepared a 3D graphene network through CVD, followed by the electrochemical deposition of NiO onto the graphene. The photographs and SEM images of the prepared 3D graphene are shown in Fig. 5(c). The synergistic effect arising from the redox reaction of metal oxides combined with graphene’s high surface area/conductivity led to a specific capacitance of 754 F g⁻¹ at 1.4 A g⁻¹, with capacitance increasing to 115% after 200 cycles and no significant decrease after 2000 cycles.

In addition to single metal composites, some bimetallic oxides show even better pseudocapacitive performance due to the synergy between different metals.⁷⁶ Xia et al.⁷⁷ used CVD to prepare a novel 3D porous thin graphene foam with core–shell structured nanowires grown on its surface. The core of the nanowires was Co₃O₄, while the shell was a composite of the conductive polymer poly [3,4-ethylenedioxythiophene] (PEDOT) and MnO₂. This composite exhibited a specific capacitance of about 390 F g⁻¹ at 10 A g⁻¹ and retained 90.2% capacitance after 20 000 cycles at 5 A g⁻¹. Yu et al.⁷⁸ used CH₄ as the carbon source and synthesized 3D graphene via CVD, then coupled it with a honeycomb-like structure, CoMoO₄, through a hydrothermal reaction. This honeycomb-like structure not only provided a stable electrode structure but also offered a large surface area for ion transport and redox reactions. The specific capacitance reached 1882 F g⁻¹ at 9.28 A g⁻¹, and even after 100 000 cycles at 400 A g⁻¹, the capacitance remained at 96.36%.

In addition to metal oxide, metal sulfides are also widely applied in graphene-based supercapacitors. Compared to metal oxides, metal sulfides offer higher specific capacitance and faster ion transfer rates.⁷⁹ Naz et al.⁸⁰ synthesized highly defective molybdenum disulfide (MoS₂) nanosheets on a 3D rGO network using a one-step hydrothermal method. The high defect density of the MoS₂ nanosheets/rGO structure provided better electrochemical performance, including faster charge transfer rates and more ion diffusion channels. The material exhibited a specific capacitance of 442.0 F g⁻¹ at a current density of 1 A g⁻¹, with 84.5% capacity retention after 3000 cycles at 5 A g⁻¹. Xin et al.⁸¹ uniformly embedded the in situ synthesized Co/Zn–S polyhedra into an rGO film for hybrid supercapacitors, following the synthetic steps depicted in Fig. 5(a). The Co/Zn–S@rGO composite electrode demonstrated a capacitance of 1640 F g⁻¹ at 1 A g⁻¹. After 8000 charge-discharge cycles, the device displayed 90.3% capacity retention. The excellent performance of the composite electrode was attributed to the electrochemical activity of Co/Zn–S polyhedra and the interconnected porous network that prevents rGO stacking. The unique interlayer porous structure facilitated the entry of electrolyte ions into the Co/Zn–S surface, fully utilizing the surface of the rGO sheets while mitigating the volume changes during charge-discharge. Jiang et al.⁸² successfully designed and synthesized a 3D network porous graphene/cobalt-nickel sulfide composite electrode using a combined method of solvothermal and sulfidation processes. The composite electrode exhibited good conductivity, high-density electrochemical active sites, and cycling stability. At a current density of 0.5 A g⁻¹, it showed a high specific capacitance of 1739 F g⁻¹, and even after 10 000 cycles, the specific capacitance retained 79.6%.

Wu et al.⁸³ prepared a composite film comprising chemically converted graphene and polyaniline nanofibers by employing vacuum filtration of a mixed dispersion. The composite film had a layered structure, with polyaniline sandwiched between graphene layers, exhibiting excellent mechanical stability and high flexibility. At a current density of 3 A g⁻¹, the specific capacitance was 197 F g⁻¹. Zhao et al.⁸⁴ utilized pyrrole to prevent the self-stacking of graphene gel during formation, leading to the creation of a super-light, N-doped, 3D graphene framework. The resulting material maintained a high specific capacitance over a wide range of current densities and had a long cycle life. At 1 A g⁻¹, the specific capacitance was 484 F g⁻¹, and at 100 A g⁻¹, it was 415 F g⁻¹, with 105% capacity retention after 1000 cycles. Chen et al.⁸⁵ used a hydrothermal method to prepare N-doped graphene hydrogel, where the organic amine used as the nitrogen source also regulated the assembly process of graphene sheets. As shown in Figs. 5(d) and 5(e), the graphene hydrogel forms a complete block structure, but the internal graphene sheets are still not stacked obviously, presenting a single-layer or multi-layer structure. At a current density of 10 A g⁻¹, the specific capacitance reached 190.1 F g⁻¹. Even at a high current density of 100 A g⁻¹, the specific capacitance retained 95.2% after 4000 cycles. Hassan et al.⁸⁶ synthesized polyaniline nanospheres through microemulsion polymerization and combined them with oxidized graphene nanosheets. Subsequently, through layer-by-layer deposition and in situ chemical reduction, a graphene/polyaniline nanostructure with an interwoven network and a 3D open structure was formed. Due to the synergistic effect of graphene and polyaniline nanostructures, this material achieved a specific capacitance of ∼350 F g⁻¹ at a current density of 10 A g⁻¹. After 5000 cycles, the capacitance retention rate of the composite material reached 81%. Yan et al.⁸⁷ prepared a 3D porous poly(3,4-ethylenedioxythiophene): poly(4-styrenesulfonic acid) (PEDOT: PSS)/rGO composite sponge through self-assembly for use in stretchable microsupercapacitors. This self-assembly process integrated PEDOT: PSS and rGO, providing enhanced electrochemical performance and mechanical flexibility. The fabricated stretchable capacitors demonstrated high areal specific capacitance (45.98 F g⁻¹ at 1 A g⁻¹) and excellent electrochemical stability (88.6% capacitance retention after 5000 cycles).

Li et al.⁸⁸ synthesized COF (Covalent Organic Framework)/rGO aerogels through a hydrothermal method. COFs were grown in situ on the surface of 2D graphene sheets, forming a hierarchical porous structure of ultra-light aerogels. The aerogel exhibited a high specific capacitance of 269 F g⁻¹ at a current density of 0.5 A g⁻¹ and retained 96% of its capacitance after 5000 cycles. Kshetri Tolendra et al.⁸⁹ prepared a hierarchical ternary carbon aerogel structure composed of graphene (G), carbon nanofibers (CNFs), and carbon nanotubes (CNTs), as shown in Fig. 6. The CNTs@G-CNF electrode material showed excellent electrochemical performance, achieving a high specific capacitance of 314.2 F g⁻¹ at 1 A g⁻¹ and retaining 98% of its capacitance after 10 000 cycles. Zhang et al.⁹⁰ utilized a simple hydrothermal method to self-assemble layered GO and MXene nanosheets into columnar 3D macro-porous graphene/MXene-based hydrogels. The presence of MXene prevented graphene stacking, resulting in a material with a large specific surface area and pore volume, showcasing high energy density and excellent long-term cycling stability. At a current density of 10 A g⁻¹, the specific capacitance reached 164.8 F g⁻¹, and after continuous 10 000 cycles at 4 A g⁻¹, the basic capacitance remained unchanged. In addition, the properties of the above materials are summarized in Table I.

Graphene is an outstanding 2D carbon material with a high specific area (2620 m² g⁻¹),⁹¹ excellent mechanical strength, good chemical stability, and extraordinary electrical/thermal conductivity. It is used in energy storage,⁴² catalysis,⁹² water treatment,⁹³ etc. In the 21st century, with the urgent demand for the improvement of the desalination performance of CDI, graphene has emerged as a promising candidate due to its substantial specific surface area and exceptional electrical conductivity, both of which facilitate the rapid capture of ions to form EDL. The first study using graphene as an electrode active material for CDI was reported by Li et al.,⁹⁴ which exhibited a maximum electrosorption capacity of 1.85 mg g⁻¹ in NaCl solution. Apparently, the adsorption capacity of conventional 2D graphene is far from being suitable for industrial applications. It is constrained by issues such as agglomeration⁹⁵ and stacking,⁹⁶ which naturally trigger the amelioration of graphene. Therein, the synthesis of 3D graphene with a 3D framework structure is a very effective method. Benefiting from the more spacious internal channels of the 3D-network structure, ion transport and charge transfer resistance can be reduced,⁹⁷ but the utilization rate of ion adsorption sites will be significantly improved. Besides, owing to the continuous spatial structure and good conductivity of 3D graphene, it can be made into a monolithic electrode without binders or conductive additives that affect the inherent characteristics of the electrode material, further improving the desalination ability of the CDI system.

The size distribution and shape of pores, as well as the specific surface area of 3D graphene, have significant effects on ion transport and storage. Conventionally, micropores (<2 nm) could guarantee a remarkably large specific area for ion anchoring, while mesopores (2–50 nm) and macropores (>50 nm) are conducive to the rapid transport of ions.⁹⁸ Therefore, in order to obtain good comprehensive desalination performance, it is imperative to optimize electrode-active materials that strike a balance between ion transport and storage, even though they may be contradictory. Chang et al.⁹⁸ constructed 3D channel-structured graphene (CSG) with a hierarchical pore structure by reacting liquid potassium with carbon monoxide gas. As expected, CSG obtained a continuous channel structure [Fig. 7(a)], mesoporous structures with pore sizes centered at 3.9 and 27.8 nm [Fig. 7(b)], and a high specific area (711.9 m² g⁻¹). In desalination experiments, CGS showed a rising SAC with increasing voltage, reaching 5.70 mg g⁻¹ at 1.5 V in a 50 mg l⁻¹ NaCl solution [Fig. 7(c)]. In addition, CGS exhibited good cycle stability at 1.2 V. This good desalination performance can be ascribed to the excellent pore size distribution and fast ion transport pathways. Analogously, Liu and co-workers⁹⁹ prepared holey graphene hydrogels (r-HGH) as electrode material by using a one-step hydrothermal method [Fig. 7(d)]. Obviously, Fig. 7(e) showed graphene was etched with holes of uneven size. The r-HGH-0.6% achieved a large specific area (398.0 m² g^−.1) and a wide pore size distribution (1.80–6.43 nm) through the engraving of 0.6 wt. % H₂O₂ and sodium ascorbate, which revealed an excellent SAC (44.44 mg g⁻¹) and a rapid adsorption/desorption equilibrium (200 s) in a NaCl solution with an initial conductivity of 1600 µS cm⁻¹. In addition, Kang et al.¹⁰⁰ prepared gram-scale micro-meso-macroporous 3D graphene (MMM-3DG) by a fumed silica template/graphitization method [Fig. 7(f)]. Compared to activated carbon electrodes, MMM-3D showed a more porous structure with an average size of 20 nm [Fig. 7(g)], which is beneficial for electrolyte penetration and ion transport. Meanwhile, it showed lower energy consumption (33.17 kJ mol⁻¹ vs 47.24 kJ mol⁻¹), faster ASAR (2.79 mg g⁻¹ min⁻¹ vs 1.01 mg g⁻¹ min⁻¹), and a high SAC of 9.37 mg g⁻¹ in 100 mg l⁻¹ NaCl concentration. Its excellent performance could be attributed to the lower ion transport impedance of the micro-meso-macroporous structure and a high specific surface area (1492.8 m² g⁻¹).

Proper spatial structure and orientation of 3D graphene are key influencing factors. They can prevent the aggregation of graphene sheets, optimize the penetration of salt water into the interior, and adjust the transport path of ions to reduce adsorption/desorption resistance. Li et al.¹⁰¹ proposed an antifreeze-assisted freezing strategy to fabricate holey graphene foam (HGF) with tough a 3D hierarchical hole structure through the self-assembly of porous GO [Fig. 8(a)]. As shown in Fig. 8(b), the obtained HGF possessed a 3D structure and numerous macropores on a micrometer scale, which prevented the aggregation of graphene sheets and reduced the resistance of the electrolyte to enter the inner electrode. After being corroded by H₂O₂, HGF also exhibited a higher specific area (124 m² g⁻¹) and a more prominent pore size (2–4 nm) ratio than that of graphene foam (91 m² g⁻¹). In CDI experiments, due to the hierarchical holes in the graphene layer and its favorable mechanical and wettability characteristics, the monolithic HGF showed a high SAC of 29.6 mg g⁻¹ at 2.0 V in 572 mg L⁻¹ NaCl. Dianbudiyanto et al.⁹⁷ reported a hierarchically porous 3D architectural graphene (GO-Mw-Hyd) by treating GO via microwave irradiation followed by H₂O₂ hydrothermal treatment. In comparison to graphene, GO-Mw-Hyd obtained a higher surface area (287.45 m² g⁻¹ vs 0.03 m² g⁻¹), a more mesoporous structure (mesoporosity increased by 89.9%) [Fig. 8(c)], and less agglomeration, thereby exhibiting more efficient ion transport as well as less charge transfer resistance. In the desalination experiment, GO-Mw-Hyd showed an impressive electrosorption capacity of 21.58 mg g⁻¹ at 1.4 V and a better adsorption rate than that of graphene. Furthermore, Ma et al.¹⁰² prepared graphene hydrogel [Fig. 8(d)] and graphene aerogel [Fig. 8(e)] and compared their desalination performance under the same conditions. Due to its larger specific area, good hydrophilicity, and electrical conductivity, graphene hydrogel achieved a higher SAC (49.34 vs 45.88 mg l⁻¹) at 2.0 V in 500 mg l⁻¹ NaCl.

On the other hand, the prepared 3D graphene also showed a good removal effect for heavy metal ions. Bharath et al.¹⁰³ fabricated a 3D graphene (prGO) cathode by heating lemon juice-reduced, f-DNA-linked GO at 750 °C. The effects of f-DNA and heat treatment enabled prGO to obtain a good interconnected 3D network with numerous porous structures, which presented a high specific capacitance of 760 F g⁻¹ and small resistance. When it was combined with a Ni/MAX anode to form a CDI device to deal with a mixed solution containing 100 mg l⁻¹ Pb²⁺ at 1.4 V, prGO achieved a high SAC of Pb²⁺ (76 mg g⁻¹). This achievement is mainly attributed to the good microporous structure and improved electronegativity achieved by the use of lemon juice and f-DNA. Besides pumping into industrial waste water, prGO also exhibited a good removal rate of 89.1% for Pb²⁺, demonstrating that 3D graphene has excellent potential for heavy metal ion treatment. You et al.¹⁰⁴ prepared 3D rGO through freeze-dried GO powder and Na₂CO₃ hydrothermal reactions at 150 °C. Benefiting from a high specific surface area (662 m² g⁻¹) and continuous mesopore distribution with pore size in the 2–10 nm range, 3D rGO gained abundant adsorption sites, reducing the transport resistance of electrons and electrolyte ions. As a result, 3D rGO achieved a high adsorption of Cu²⁺ with SAC of 18.1 mg g⁻¹ in 50 mg l⁻¹ Cu(NO₃)₂ solution at 1.0 V, demonstrating that 3D graphene is a potential candidate material for heavy metal ion removal.

The transformation of 2D graphene into wrinkled 3D graphene¹⁰⁵ can increase the actual specific area and expand the distance between graphene sheets. This, in turn, facilitates the penetration of solutions and the access of ions to internal adsorption sites. Meanwhile, inserting other porous materials between the wrinkled 3D graphene can further expand the interlayer distance and reduce the number of closed spaces. For example, Khan et al.⁹⁶ reported a 3D intercalated graphene sheet-sphere nanocomposite architecture (GSSNA) that uses graphene spheres inserted in the middle of graphene sheets, which inhibits irreversible aggregation of graphene sheets and facilitates the electrolyte to enter the interior. GSSNA-12, GSSNA-21, and GSSNA-11 were obtained by adjusting the ratio of pristine graphene (GR) to graphene balls (GBs). Among them, GSSNA-11 exhibited the best desalination capacity and high SAC of 22.9 mg g⁻¹, good cyclic stability, and a high salt removal rate (90%) at 1.2 V, which can be attributed to the excellent structure. Similarly, Feng et al.²² introduced a 3D graphene-supported N-doped hierarchically porous carbon (3DNHPC) such that N-doped hollow mesoporous carbon spheres were uniformly distributed within the graphene layer by a simple template-direct method, which successfully suppressed the aggregation and restacking of graphene sheets. The incorporation of N-doped hollow mesoporous carbon spheres also increased the number of adsorption sites, hydrophilicity, and conductivity. When used as a CDI electrode, the 3DNHPC electrode exhibited a high SAC of 25.5 mg g⁻¹ as well as remarkable stability (100%) after 22 adsorption/desorption cycles in a 500 mg l⁻¹ NaCl solution at 1.2 V. In contrast, 3D graphene only presented a SAC of 10.4 mg g⁻¹ under the same experimental conditions.

The preparation of graphene into porous 3D graphene spherical shells is an effective way to suppress graphene agglomeration and reduce ion transport resistance. Zhu et al.¹⁰⁶ introduced isolated graphene hollow shells (3DGA-C) and interconnected graphene hollow shells (3DGA-OP) by annealing polystyrene wrapped with large-scale and small-scale graphene, respectively. Compared to pure graphene or 3DGA-C, 3DGA-OP displayed a higher SAC of 14.4 mg g⁻¹ at 1.2 V [Fig. 8(h)], which can be attributed to the unobstructed porous structure [Figs. 8(f) and 8(g)] and the increased specific surface, pore volume, and ion diffusivity.

Unlike the method that required a template to obtain 3D graphene, binder-free 3D graphene¹⁰⁷ was synthesized by assembling GO on the surface of Ni foams. The resulting material can be directly used as cathode and anode in the CDI system, which exhibited an ion removal capacitance of 22.3 mg g⁻¹ and 65% desalination capacity retention after 200 cycles in 500 ppm of NaCl. Li and co-workers also reported a reborn graphene called 3D graphene after Nirvana (NvG).¹⁰⁸ This material was created through sequential reassembly of pulverized porous graphene particles (PGPs) and graphene chippings (GCs) with graphene sheets [Fig. 9(a)]. Therein, PGPs endow high porosity, enabling efficient electrolyte penetration, ion adsorption, and the enhancement of mechanical properties through a particle-reinforced effect. The graphene sheets functioned as a structural framework, offering uniform dispersion of PGPs, while GCs efficiently connected the graphene sheets to further enhance mechanical properties. Hence, NvG obtained higher porosity, better electrical conductivity (1.41 times), and mechanical strength (32.4 times) as compared to 3D graphene. In the CDI experiment, NvG exhibited a high volumetric SAC of 8.02 mg cm⁻³ and outstanding cycling stability (94% retention rate) after 95 cycles in an initiating solution of 500 mg l⁻¹ NaCl. Liu et al.¹⁰⁹ reported a novel method to fabricate 3D flexible graphene frameworks (3DFGFs) with micro-/meso-/macro-pores from commercial carbon paper [Fig. 9(b)] for CDI application. Combining the modified hummers, freeze-dying, and hydrothermal reduction processes, 3DFGFs obtained excellent conductivity and abundant oxygen-containing functional groups that improved their hydrophilicity. When the external forced voltage was 1.6 V, 3DFGFs achieved an area electrosorption capacity of 581.51 mg m⁻² in the initial concentration of 500 mg l⁻¹ NaCl and exhibited superior stability (>90% retention rate) after 50 cycles under a 50 mg l⁻¹ NaCl solution.

The surface state of materials has a crucial impact on electrical conductivity and hydrophilic properties, thereby affecting ion adsorption/desorption rate in CDI. Due to the poor hydrophilicity of pure graphene, many ionic anchor sites are underutilized, and the adsorption/desorption rate would be constrained. Heteroatom doping is an effective strategy to regulate the electronic structure and then ameliorate the hydrophilic properties of the material surface.

N-doping is an effective method to improve the conductivity, wettability, and specific capacitance performance of carbon materials¹¹⁰ without altering their morphological advantages. Zhang et al.¹¹¹ synthesized a freestanding N-doped graphene membrane electrode (F-N-GPM) by assembling the graphene nanosphere and heating it under an NH₃ atmosphere [Fig. 10(b)]. Compared to F-GPM, the doping of nitrogen reduced the resistance of the electrolyte to the inner pore surface and improved the wettability of the prepared electrode material, which was confirmed by the smaller contact angle [Fig. 10(c)], small internal resistance [Fig. 10(d)], and large specific capacitance [Fig. 10(e)]. Therefore, F-N-GPM displayed a higher SAC at various concentrations [Fig. 10(f)] and a higher SAC of 21.8 mg g⁻¹ at 1.8 V in 100 mg l⁻¹ NaCl in comparison to F-GPM. Noonan et al.¹¹² developed a hierarchical porous N-doped spray-dried graphene (N-SDG) through annealing with urea in nitrogen flow. The incorporation of abundant nitrogen atoms effectively improved the hydrophilicity and electronic conductivity of N-SDG. N-SDG showed a high SAC of 19.6 mg g⁻¹_, surpassing that of SDG (14.3 mg g⁻¹). Furthermore, N-SDG retained 100% of its SAC after 50 charge/discharge cycles, soaking in 500 mg l⁻¹ NaCl at 1.4 V. In another study, N-doped graphene aerogel, fabricated by hydrothermal reduction of GO with ammonia solution, was further tuned to nitrogen and phosphorus-doped 3D graphene with more mesopores and a larger specific surface area by H₃PO₄ activation [Fig. 10(a)].¹¹³ Nitrogen and phosphorus-doped 3D graphene also exhibited the best specific capacitance (177.19 F g⁻¹) and SAC (19.57 mg g⁻¹) at 1.6 V, as compared to 3D graphene and N-doped 3D graphene. Mamaril et al.¹¹⁴ prepared nitrogen and fluorine co-doped 3D rGO (3D NFrGO) by hydrothermal treatment of a mixed solution of GO, urea, and HF at 180 °C. In desalination experiments, 3D NFrGO made an excellent removal efficiency of Cu²⁺ with SAC of 52.4 mg g⁻¹ in 100 mg l⁻¹ Cu²⁺ solution at 1.2 V. This can be attributed to the increased conductivity and the formation of vacancy defects by the doping of nitrogen and fluorine, the large specific area (182.7 m² g⁻¹), and the suitable bimodal pore size distribution at 2.1 and 5.6 nm in the 3D framework.

In addition, surface functionalization can tune the charge distribution and wetting properties of graphene surfaces. For example, Liu et al.¹¹⁵ grafted sulfonic groups and amine functional groups on 3D graphene by using an aryl diazonium salt solution and 3-aminopropyltriethoxysilane, respectively. These modifications effectively inhibited the aggregation of graphene sheets, increased the wettability of the electrode, and minimized the co-ion expulsion effect due to the ion-selective function. In desalination experiments, the CDI apparatus composed of an amino-modified graphene (3DNGR) anode and sulfonic acid-modified graphene (3DSGR) cathode possessed more than a two-fold improvement in the desalination efficiency compared with the unmodified electrodes. In addition, it exhibited a higher SAC of 13.72 mg g⁻¹ at 1.4 V and an initial conductivity of 1008 µS cm⁻¹. Analogously, El-Deen¹¹⁶ functionalized microporous activated graphene with carboxymethyl cellulose and quaternary ammonium cellulose to obtain a negatively charged (COO²⁻) cathode (C-3DAPGr) and a positively charged (NR₄⁺) anode (Q-3DAPGr), respectively. These materials were then assembled together or combined with uncoated microporous activated graphene to form asymmetric CDI. Compared with symmetric CDI with an uncoated electrode, asymmetric CDI exhibited significantly improved electrosorption performance due to the lower co-ion expulsion. Particularly, the QC-3DPGr cell not only obtained the best performance, achieving a SAC of 18.43 mg g⁻¹ at 1.4 V, but also achieved a killing rate of 98.55% against E. coli. In another study, El-Deen et al.¹¹⁷ modified activated graphene with a charged polystyrene sulfonate as the cathode, which achieved exceptional hydrophilicity and a high specific area (1401.53 m² g⁻¹). Therefore, modified electrodes showed exceptional SAC of 26.33 mg g⁻¹, high charge efficiency (88%), and good recyclability at 1.4 V. Liao et al.¹¹⁸ synthesized phosphate-functionalized graphene hydrogel (HGP) by creating porous structures through H₂O₂, then modifying holy graphene nanosheets with phytic acid. Compared to rGO, the HGP obtained a larger specific area (186.8 m² g⁻¹ vs 70.9 m² g⁻¹) and better wettability, which enhanced its specific capacitance. Meanwhile, HGP exhibited a SAC of uranium (Ⅵ), reaching 545.7 mg g⁻¹ at 1.2 V, as well as excellent selectivity to uranium (Ⅵ) in mixed solutions containing other metal ions.

The desalination ability of pure 3D dimensional graphene is still limited, and optimizing its internal structure usually involves a complex and difficult process. Meanwhile, the fabrication of 3D graphene is too expensive for large-scale manufacturing, which limits its application as a standalone electrode material in CDI. To tackle this issue, researchers pay more attention to combining carbon-based or non-carbon-based materials with 3D graphene to synthesize multifunctional composites. These composites yield effects that go beyond simple joint effort. Therein, 3D graphene can be employed as a scaffold to provide an excellent conductive network, accommodate nanoparticles with more ion adsorption sites, and evenly disperse the constituent materials.

In terms of carbon-based 3D graphene composite materials, carbon materials with a large specific surface area and rich pore size are generally considered to be excellent candidates. For example, Xu et al.¹¹⁹ prepared novel hierarchical hybrids (3DGF-MCS) by decorating microporous carbon spheres on 3D graphene frameworks, endowing the porous structure with micro-/meso-/macropores, a high specific surface (676.9 m² g⁻¹), and good electrical conductivity. When used as a CDI electrode, 3DGF-MCS showed good specific capacitance (288.8 F g⁻¹), a high SAC of 19.8 mg g⁻¹, and excellent cyclic stability (96% retention rate) after 30 cycles in 500 mg l⁻¹ NaCl. In another study, Liu et al.¹²⁰ reported an N-doped activated porous carbon decorated by 3D interconnected graphene (NAPC/G). This material amalgamates the advantages of the ultra-high specific surface area (2993.5 m² g⁻¹), good hydrophilic properties, sufficient micro-mesoporous structure, and excellent electrical conductivity of 3D graphene. In desalination experiments, NAPC/G demonstrated a high SAC (38.5 mg g⁻¹), fast ASAR (6.6 mg g⁻¹ min⁻¹), and stable cycle performance (retention rate reached 93.5% after 50 cycles). Similarly, Mi et al.¹²¹ synthesized a hierarchical composite of N-doped carbon spheres and holey graphene hydrogels (N-HMCS/HGH) by a one-pot hydrothermal process [Fig. 11(a)], where N-doped spheres provided abundant hierarchical pores, good hydrophilicity, and holey graphene endowed continuous electron pathways and many in-plane pores to facilitate solution and ion access to the inner surface. As shown in Fig. 11(b), N-HCMSs were efficiently connected by HGH and uniformly distributed in N-HMCS/HGH. In CDI experiments, the N-HMCS/HGH composite exhibited the largest change in conductivity, a significant SAC of 17.8 mg l⁻¹ in a feeding solution of 500 mg l⁻¹ NaCl, and excellent cycle stability after 35 cycles.

Cao et al.¹²² composited 3D graphene hydrogel with single-walled carbon nanotubes (SWCNTs) or multi-walled carbon nanotubes (MWCNTs) by a one-step water bath method to prepare SWCNTS/rGO and MWCNTS/rGO, respectively. The intercalation of SWCNTs or MWCNTs successively prevented the aggregation of graphene sheets and provided more contact points and a conductive network to increase the conductivity of composites, which were consistent with the TEM images [Figs. 11(d) and 12(e)]. However, compared with MWCNTS/rGO, SWCNTS/rGO obtained a larger specific surface area (308.37 m² g⁻¹) and lower charge transfer resistance, thus achieving a higher SAC (48.73 mg g⁻¹ vs 39.53 mg g⁻¹) at 2 V in 300 mg l⁻¹ NaCl.

Among non-carbon-based component materials, some metal oxides, such as MnO₂,¹²³ TiO₂,¹²⁴ Fe₃O₄,¹²⁵ MoS₂,¹²⁶ and layered double hydroxide (LDH),¹²⁷ are pseudocapacitive materials with high specific capacitance that are considered to be promising ion-adsorbing materials and have been utilized in batteries, supercapacitors, and water treatment. However, these materials suffer from poor electric conductivity and easy agglomeration, which hampers their applications, especially in the CDI field. Therefore, graphene with good electrical conductivity was introduced to prepare composites. For example, Gao et al.¹²⁸ found MoS₂/rGO composites gained a desalination capacity of 34.2 mg g⁻¹ in 300 mg l⁻¹ NaCl at 1.4 V, which is higher than that of pure MoS₂ (20.5 mg g⁻¹) and rGO (11.26 mg g⁻¹), demonstrating good synergy. Ren et al.¹²⁷ also observed that, as a result of incorporating graphene, MgAl-Ox/G obtained a higher SAC (13 mg g⁻¹) than MgAl-Ox, a kind of LDH, which only exhibited a low SAC of 5 mg l⁻¹ in 500 mg l⁻¹ NaCl.

As graphene with a 3D framework, 3D graphene has also been successfully used to solve the above-mentioned drawbacks. For instance, MnO₂ is a promising material with ultra-high theoretical specific capacitance (1232 F g⁻¹), but poor electric conductivity (∼105 S cm⁻¹)¹²⁹ hinders its direct application in CDI. Zhou et al.¹³⁰ loaded MnO₂ on 3D graphene to prepare nanocomposite (PGA/MnO₂) by steam activation and MnO₂ deposition [Fig. 11(d)]. PGA/MnO₂ exhibited a high surface area of 490 m² g⁻¹ and loaded MnO₂ with a small size (5–10 nm), which limited the agglomeration of MnO₂ and improved its utilization. In desalination tests, PGA/MnO₂ exhibited a SAC of 15.3 mg g⁻¹ at 1.2 V in a 500 mg l⁻¹ NaCl solution that was higher than PGA (∼8 mg g⁻¹) and MnO₂ (11.9 mg g⁻¹).¹³¹ This enhancement can be ascribed to the extra ion adsorption capacity of MnO₂. In addition, PGA/MnO₂ presented high SAC toward Pb²⁺ (247.3 mg g⁻¹), Cu²⁺ (83.5 mg g⁻¹), and UO₂²⁺ (172.5 mg g⁻¹), demonstrating outstanding prospects for metal ion removal. Gu et al.¹³² simultaneously loaded polypyrrole (PPy) and MnO₂ onto 3D graphene by hydrothermal processes to fabricate graphene-polypyrrole-Mn (rGO-PPy-Mn) composites. Thanks to the synergistic effect of the three materials, rGO-PPy-Mn composites obtained a well-defined 3D conductive skeleton structure, improved capacitance, and smaller internal resistance as compared to pure rGO or PPy. When tested in a 1000 µS cm⁻¹ NaCl solution at 2.0 V, the rGO-PPy-Mn electrode displayed a SAC of 18.4 mg g⁻¹ higher than that of rGO (4.8 mg g⁻¹) and PPy (9.2 mg g⁻¹). In another study, Hosseinzadeh et al.⁹⁵ functionalized 3D graphene with MnO₂ and NiO nanoparticles with opposite charges as cathode and anode, respectively. Due to dissimilar isoelectric points, the decoration of MnO₂ and NiO nanoparticles not only increased the potential difference of electrodes but also declined ion transport resistance, which dramatically enhanced the electrosorption capacity, reaching 21.01 mg g⁻¹ in a 1000 mg l⁻¹ NaCl solution.

TiO₂ is another low-cost, highly hydrophilic, and easily anchored metal oxide, but it is also hindered by its poor electric conductivity¹³³ and small surface area. In order to address this problem, Yin and co-workers³⁶ synthesized graphene aerogel/TiO₂ (GA/TiO₂) by hydrothermal treatment and freeze-drying. Compared to graphene (9.9 mg g⁻¹) or activated carbon (1.2 mg g⁻¹), GA/TiO₂ displayed a higher SAC of 15.1 mg g⁻¹ and a faster desalination rate because of the lower mass transport resistance of ions inside the electrode at 1.2 V in 500 mg l⁻¹ NaCl. Xu et al.¹³⁴ prepared an rGO/titanium dioxide (rGO/TiO₂) hydrogel with different TiO₂ mass loadings by hydrothermal treatment. When the mass ratio of TiO₂ to rGO is 20%, the obtained composite (rGO/20%TiO₂) achieved the best specific capacitance (325.8 F g⁻¹) and NH₄⁺ removal (8.52 mg g⁻¹) at 2.0 V in 1.0 mmol l⁻¹ NH₄Cl due to the enhanced structural and electrochemical properties.

On the other hand, some metal oxides with special functions attached to 3D graphene can further expand the application potential of CDI. For instance, Hu et al.¹³⁵ prepared λ-MnO₂/rGO composites by modifying 3D graphene with a lithium-ion sieve (λ-MnO₂) through an in situ redox reaction. As expected, consisting of the as-prepared material as the cathode and activated carbon as the anode, the CDI cell exhibited more significant Li⁺ adsorption quantity compared to other cations [Fig. 12(a)] and good Li⁺ selectivity [Fig. 12(b)] in simulated salt lake brine, as well as a high SAC (601 µmol g⁻¹) of Li⁺ at 0.7 V.

In addition to metal oxides, some new materials have also been developed to combine with 3D graphene to obtain better desalination performance. Vafakhah et al.¹³⁷ proposed a graphene aerogel decorated with Prussian blue nanocubes (PB/rGA) as a binder-free anode by hydrothermal method. Prominently, apart from the electric double-layer mechanism, PB introduced the ion intercalation principle to adsorb Na⁺. Comprising a PB/rGA anode and a rGA cathode, the CDI cell showed a pretty high SAC of 130 mg g⁻¹ in a fixed concentration of 2500 ppm NaCl and maintained an excellent desalination capacity over 100 cycles, which was superior to most of the above pure 3D graphene. In addition, Wang and co-workers¹³⁶ synthesized a lanthanum-doped hydroxyapatite/3D-rGO composite (LaHAP/3D-rGO) by ultrasonic treatment [Fig. 12(c)]. Therein, 3D-rGO provided a good framework to uniformly disperse LaHAP and developed the electrochemical properties of LaHAP. When coupled with an activated carbon anode, LaHAP/3D-rGO exhibited a better adsorption F⁻ adsorption capacity of 1132 µmol g⁻¹ in comparison to 3D-rGO (472 µmol g⁻¹) at 1.6 V in a 3 mmol NaF solution and excellent F⁻ selective adsorption in F⁻-containing dianion solutions [Fig. 12(d)].

3D porous graphene materials have attracted significant attention and research interest in recent years due to their advantages, such as their high specific surface area, excellent conductivity, and rich porous structure. This review summarizes the recent progress in 3D porous carbon materials for both supercapacitor energy storage and capacitive desalination. It particularly focuses on the preparation methods of 3D graphene, subsequent modification approaches, overall electrochemical performance, and capacitive desalination efficiency.

Despite the numerous advantages of 3D graphene, there are still challenges and technical difficulties in its practical applications. First, the insufficient capacitance and poor cycling stability of graphene are important issues that hinder its potential use as a high-energy density capacitor. To address these problems, the optimization of graphene’s structure to increase its specific surface area and porosity and the introduction of stable conductive polymers or metal sulfides as dopant materials emerge as effective solutions for enhancing capacitance.

Second, in terms of SAC (Table II), salt adsorption rate, and cycle stability, 3D graphene still needs more progress to stand out among other materials in CDI applications. Furthermore, modulating 3D graphene structures and rationally designing 3D graphene composites to endow more adsorption sites, enhance the hydrophilicity of the surface, and reduce ion transport resistance are still the focus of continued research. In particular, greater efforts should be dedicated to the integration of 3D graphene with functional materials, aiming to accomplish selective ion removal, ion recovery, and the isolation of toxic ions. Meanwhile, the fabricated 3D graphene, 3D graphene composites, and the corresponding preparation methods can also be utilized or expanded to other fields, such as batteries,¹³⁸ catalysis,^92,139 and photothermal evaporation.¹⁴⁰ 

Notably, for a fairer comparison of the energy storage and electrosorption properties of the reported 3D graphene, consistent evaluation criteria should be maintained, such as applied voltage, current density, solution type and concentration, and cycle times. Then, researchers can better understand the drawbacks and strengths of various 3D graphene materials, facilitating the selection of the best material for research and future applications.

Fourth, in practical applications, cost and large-scale production are also crucial issues to be considered. Producing high-quality 3D graphene networks typically involves high-temperature treatments and multi-step chemical processes. These processes can consume a significant amount of energy and resources, leading to environmental impacts and high costs. Additionally, some graphene production methods involve the use of hazardous chemicals, such as strong acids or bases, which have serious security risks. Therefore, researchers need to make efforts to find more environmentally friendly synthesis methods.

In conclusion, although 3D graphene holds great potential in the fields of supercapacitor energy storage and capacitive desalination, addressing the aforementioned challenges requires successive research and improvement. With the continuous advancement of science and technology, it is believed that the performance of 3D graphene can be further optimized, and its broad application prospects in energy storage and desalination can be expanded.

This work was supported by the National Natural Science Foundation of China (Grant Nos. 52272039, U23B2075, and 51972168), the Key Research and Development Program in Jiangsu Province (Grant No. BE2023085), the Natural Science Foundation of Jiangsu Province of China (Grant No. BK20231406), and the High-Performance Computing Center (HPCC) of Nanjing University.

The authors have no conflicts to disclose.

Hongda Zhu: Formal analysis (equal); Investigation (lead); Methodology (lead); Writing – original draft (lead). Dingfei Deng: Formal analysis (supporting); Investigation (supporting). Chiwei Xu: Formal analysis (supporting); Investigation (supporting). Xuebin Wang: Funding acquisition (equal); Supervision (equal); Writing – review & editing (equal). Xiangfen Jiang: Supervision (equal); Writing – review & editing (equal).

The data that support the findings of this study are available within the article.