Nanoarchitectured MOF-derived porous carbons: Road to future carbon materials Open Access

Carbon materials have been in the hands of the mankind from as early as ancient Egyptian and Sumerian eras (∼B.C. 3000) when the people of that time recorded the use of porous charcoals as a remedial tool to treat indigestion.¹ Ever since the carbon materials set their foot into the history of mankind, possibly even before the first record, they were never left unused and the ways of applying their novel properties have gradually evolved to various forms with more in-depth understanding of the structure-to-property relationships and more sophisticated synthetic methods. Indeed, the advance of materials science and nanotechnology has not only paved the deeper understanding of carbon microstructure even at the atomic scale but also devised tailored synthetic methods to control carbon nanoarchitectures at the nano-scale, therefore, expanding the pool of carbon materials available for various applications with historic examples such as fullerene, carbon nanotube (CNT), graphene, ordered mesoporous carbons, etc. Though the discovery of new materials is not always the prerequisite for the new science (as demonstrated by the ground-breaking discovery of the Li-ion intercalation into the interlayer space of graphite²), the great passion for “new materials” keeps growing among materials scientists who willingly explore persisting problems on how to selectively achieve the target advantages without the accompanying disadvantages in materials (e.g., “crystallinity vs porosity,” “surface area vs diffusion efficiency,” “activity vs selectivity,” “kinetics vs stability,” etc.). Likewise, “new carbon materials” are also given tremendous thoughts and efforts beyond the scope of measure, resulting in an unprecedented increase in the number of scientific reports related to carbon materials.

Since 2011, metal-organic frameworks (MOFs) have been widely used as carbon precursors to easily obtain microporous carbon materials through direct-carbonization.^3–5 Despite the facile synthetic methods and promising properties of MOF-derived porous carbons, they present few daunting disadvantages such as a low level of graphitization, low diffusion efficiency, and large electrochemically inactive surface area attributed to the narrow pore size distribution in the range of micropores (pore diameter below 2 nm). To address such problems, novel synthetic methods and approaches have been actively investigated under the thematic concept of “Nanoarchitectonics.” It aims to achieve the arrangement of nano-sized structural units comprised of a specific combination of atoms and/or molecules in an intended configuration to bring about novel or enhanced structure-to-property relationship of nanomaterials. There have been meaningful approaches for nanoarchitecturing involving the manipulation of structural units in atomic and/or molecular levels through physical interactions, chemical reactions, applied fields, or self-assembly. Particularly, nanoarchitecturing of MOF-derived porous carbons can be achieved by a range of measures, for example, microstructural engineering for highly graphitic or defective sites, local atomic environment engineering to tune the surface charge or electronic distribution, pore size engineering to larger and/or mixed nanopores, morphological engineering to unique tunable nanostructures (e.g., hollow, layered, low-dimensional, etc.), and superstructural engineering through the self-assembly of individual MOF single crystals to induce specifically packed morphologies.^6,7 Interestingly, most nanoarchitectured morphologies of MOF crystals can be maintained even after carbonization, whereas the atomic environment of MOFs changes to the carbon microstructure with defect sites and heteroatom/metal dopants. Therefore, the nanoarchitectured MOF-derived porous carbons hold enormous potential to achieve the target carbon materials for various applications, especially in the field of energy storage/conversion and environmental remedy. To this end, this review aims to highlight nanoarchitectured MOF-derived porous carbons as the forefront materials toward future carbons because of their clear advantages specified as follows: (1) MOFs are highly porous with large surface area and high pore volume; (2) coordination chemistry and crystalline nature of MOFs allow diverse synthetic pathways to effectively alter the nanoarchitectures by various means ranging from template-free to template-based methods; (3) MOFs can be easily converted to their porous carbon derivatives through the simple pyrolysis at high annealing temperatures (i.e., carbonization); and (4) the atomic compositions of the resulting MOF-derived porous carbons can be easily varied by in situ or post-synthetic doping of heteroatoms and metals, thus altering the local atomic environment to favor the target applications.

The first part of this review presents the previous studies on the microstructural evolution of traditional carbons in the temperature-controlled pyrolysis process to allow better scope of knowledge on the synthesis of carbon materials. We also extend to the recent in situ observation of the annealing process for zeolitic imidazolate framework (ZIF) type MOFs and describe how the presence of certain metals in MOFs influences the development of microstructure in MOF-derived porous carbons. The second part of this review rationalizes the need for nanoarchitecturing in porous carbon materials in the context of electrochemistry, specifically, in consideration of the electrical double layer (EDL) formation on the surface of nanopores. To effectively identify the point of improvement or come up with entirely novel synthetic methods for nanoarchitecturing, it is important to grasp what types of synthetic methods and approaches have been reported so far. In this sense, the two main methodological categories of template-free and template-based methods are outlined with relevant example materials resulting from each method.

The focus of this review is then narrowed down to a specific theme of nanoarchitectured MOF-derived porous carbons by introducing various synthetic methods and pathways to tailor control the nanoarchitectures. The synthetic methods toward nanoarchitectured MOF-derived porous carbons are categorized into template-free and template-based methods based on the use of template materials in the synthetic pathways. Depending on the intactness of the original MOF crystals in the process of nanoarchitecturing prior to the carbonization, the template-free methods are further divided into sub-classes of bottom–up (MOF crystals remain intact prior to carbonization) and top–down approaches (MOF crystals are etched or dissolved prior to carbonization). Various nanoarchitectured MOF-derived porous carbons obtained by the bottom–up and top–down approaches of the template-free methods (e.g., MOF-on-MOFs, multimetallic MOFs, polymer-on-MOFs, self-assembled MOFs, and recrystallized MOFs) are specifically presented and discussed. The template-based methods, subclassified to hard-template and soft-template approaches, are presented with examples including hard-templated MOFs, soft-templated MOFs, and sacrificial MOFs. With extensive scope of understanding for the tailored control at different stages (e.g., atomic, nano-, and macroscopic stages), we envisage that nanoarchitectured MOF-derived porous carbons can achieve potential “future carbon” and present future directions to address few challenges faced in the field (Fig. 1).

The understanding of microstructures of traditional carbon materials from the previous studies holds profound implications for the synthesis and characterization of MOF-derived porous carbons of new nanoarchitectures. As it is the case for the conventional carbon black, where the annealing conditions significantly influence the evolution of the carbon microstructure, the parameters employed for the carbonization of MOFs profoundly shape the unique morphologies and properties in the resulting MOF-derived porous carbons.

As the annealing conditions for the production of carbon materials became more sensitive and controllable, the research on the microstructure of carbon materials has begun to prevail. For instance, the advance of x-ray crystallography ignited an enormous research interest to reveal the structure of the carbon materials based on their distinct x-ray diffraction (XRD) patterns. As one of the early XRD studies on carbon materials, the structure of carbon black was investigated and proposed by Warren in 1934. Based on his Fourier integral analysis of XRD patterns, he claimed that the carbon black is the structural mixture of graphitic carbons (GCs) (ranging from a single layer to several layers thick) and amorphous carbons.⁸ Few years later, Warren reported a substantial sharpening of the diffraction pattern of carbon black after thermal treatment at 2000 °C.⁹ Typically, the XRD pattern tends to sharpen toward that of the graphite, indicating an increase in the level of graphitization upon heating at such high temperatures [Fig. 2(a-i)]. Based on this observation, it was suggested that the “turbostratic” structure of the unheated carbon black consisting of randomly oriented stacks of graphite layers gains the orderedness to form a continuous graphitic structure as the stacks of graphite layers undergo thermal movement and rotation at high annealing temperatures [Fig. 2(a-ii)].

With the help of transmission electron microscope (TEM) analysis, it was possible to better visualize the internal structure and carbon layers of carbon materials, hence providing more scientific evidence to verify previously proposed models of carbon microstructure from the XRD analysis. In 1968, Heidenreich and colleagues reported the first high-resolution TEM (HRTEM) images of lattice fringes of graphitized carbon black and proposed that the carbon black may adopt the structural model composed of concentrically arranged turbostratic graphitic domains.¹⁰ They also claimed that the appearance of defects by dislocations or bending of graphitic layers in the carbon black is rather occasional under the TEM microscope. However, more recent TEM observations have revealed that the graphitic layers of the carbon black involve abundant defects, making the carbon black being regarded as disordered or amorphous carbon with short- and medium-range graphitic domains. A number of studies implementing both XRD and TEM analyses to investigate the effect of pyrolysis temperature on the level of graphitization were subsequently reported. In 1989, Rouzaud and Oberlin reported their work on the TEM analysis of the carbon obtained at various temperatures from unheated to 2900 °C and traced the evolution of graphitic layers.¹¹ Based on the TEM observations, they claimed that the structural evolution of parallel graphitic layers occurs at four main stages depending on the temperature [Fig. 2(b)]. In Stage 1, the individual basic structural units (BSUs) of less than 1 nm in diameter are randomly scattered in the fragment, and non-carbon elements are thermally removed from the sample. In the following stage, the BSUs arrange themselves into short to medium range graphitic layers that are intermittently present in the carbon matrix. As the temperature enters Stage 3, the structural defects from the previous two stages are mostly removed to form continuous graphitic layers that are arranged parallel to each other. Above 2000 °C in Stage 4, the graphitization is further enhanced as distortions are annealed and defects and heteroatoms are removed.

In 1951, Franklin investigated the effect of even higher temperatures up to 3000 °C with various types of carbon materials.¹² Her study was first to classify different carbon samples as either graphitizing or non-graphitizing carbons according to the change in their XRD patterns. From the study, only certain types of carbon samples demonstrated an increase in the level of graphitization upon high-temperature heat treatment. She claimed that the gradual displacement of graphitic carbon crystallites is the major factor for graphitization. Based on the claim, she proposed that graphitic and non-graphitic carbons can be differentiated by the level of cross-linking between graphitic carbon crystallites. Specifically, the displacement graphitic carbon crystallites to form a continuous graphitic layer is more favored when the crystallites are less cross-linked to each other. In contrast, when the crystallites are highly cross-linked to each other, the displacement of the crystallites is impeded, hence forming non-graphitic carbons [Fig. 2(c)]. Franklin also suggested that the more cross-links between the crystallites potentially lead to the greater hardness and the higher porosity of the resulting carbon materials.

Later studies on non-graphitizing carbons proposed that they are mainly composed of sp²-hybridized carbon atoms with several types of defects causing the formation of curvature in the form of fullerene-like or nanotube-like microstructures.^13,14 The proposed microstructure model of glass-like carbon at various temperatures was generated by feeding the experimental diffraction data in both reciprocal and real spaces to the atomic pair distribution function [Fig. 2(d)].¹⁵ With high correlation between the experimental and simulated data, the microstructure models of the glassy carbon at 800, 980, 1500, 2000, and 2500 °C (GCx, where x is the pyrolysis temperature) were used to investigate the change in defects and graphitic layers. From the model, the generation of large structural units is observed as small building blocks coalesce at higher pyrolysis temperatures. For low-temperature glassy carbons (GC800 and GC980), their microstructure models exhibit high disorderedness with folded and cross-linked planar sp²-hybridized carbon layers in a short range order [Figs. 2(d-i) and 2(d-ii)]. The short range order microstructure of GC800 and GC980 inevitably involves various defects, especially at the boundaries linking the neighboring domains. The presence of such defects induces the formation of curvature and irregular interlayer spaces in the carbon microstructure while allowing some of the planar sp²-hybridized carbon layers to come closer to each other or even merge together. GC1500 gains the longer range order microstructure with substantially reduced number of defects, which leads to the formation of graphene-like planes and fullerene-like curved fragments shaping more defined and elongated pore structure [Fig. 2(d-iii)]. The microstructure models of high-temperature glassy carbons (GC2000 and GC2500) demonstrate even longer range of order with significantly less defect sites in their microstructure [Figs. 2(d-iv) and 2(d-v)].

In carbon microstructure, various types of defects can be introduced to effectively alter the charge and electron distribution of the carbon surface. Indeed, several types of defects are visualized under the aberration-corrected TEM (AC-TEM).¹⁶ For instance, Stone–Wales defects exhibit a 90° rotation of a single carbon–carbon bond in the sp²-hybridized carbon network relative to the midpoint of the bond, giving rise to two 5-membered carbon rings and two 7-membered carbon rings (5775) [Fig. 2(e-i)].¹⁶ Alternatively, one missing carbon atom in the sp²-hybridized carbon network can generate monovacancy defect consisting of a 5-membered carbon ring and a 9-membered carbon ring (5–9) [Fig. 2(e-ii)]. When two carbon atoms are missing, several divacancy defects of different configurations can be formed. For instance, when two carbon atoms are removed from the carbon network consisting of ten 6-membered carbon rings, defect reconstruction to two 5-membered carbon rings and one 8-membered carbon ring (5–8–5) takes place [Fig. 2(e-iii)]. As the divacancy defect with 5–8–5 configuration induces the local curvature, it can undergo further defect reconstruction to more stable 555–777 configuration consisting of three 5-membered carbon rings and three 7-membered carbon rings [Fig. 2(e-iv)]. Alternative divacancy defect of 5555–6–7777 configuration with four 5-membered carbon rings and four 7-membered carbon rings can be formed by rotating a single carbon–carbon bond in the 555–777 configuration by 30° [Fig. 2(e-v)]. Larger and more complex configurations of multivacancy defects are also possible with more than two missing carbon atoms. Any type of defect can serve to alter the local atomic property of carbon materials, thus potentially achieving preferred adsorption sites for different chemicals and tuned electronic, magnetic, mechanical, and thermal properties for a range of different applications.^17–20 Along with defects, adatoms and substitutional impurities can be introduced to the carbon microstructure, therefore, changing the local atomic environment.

As each type of MOF consists of its unique combination of metal and organic linker, it undergoes distinct thermal decomposition and carbonization processes under specific annealing condition. The insight gained from the previous studies on traditional carbon materials offers more in-depth understanding and diverse ways of engineering the carbon microstructure of MOF-derived porous carbons. Based on the knowledge of the evolution of different carbon microstructures depending on the type of carbon precursors and annealing conditions, researchers can advance the design of MOF-derived porous carbons with tailored nanoarchitectures and unique properties for a wide range of applications, including energy storage, catalysis, and environmental remediation. ZIF-8 is a type of MOF widely used to obtain porous carbons via direct-carbonization and the understanding of the carbon microstructure of ZIF-8-derived nanoporous carbon (ZIF-8-C) is crucial to correlate the structure-to-activity relationship for certain applications, especially energy-related applications (e.g., secondary batteries, supercapacitors, and electrocatalysts). In 2015, Gadipelli and Guo investigated the effect of carbonization temperature on the development of microstructure of ZIF-8-C.²¹ Based on XRD analysis, a broad diffraction peak at 2θ = ∼25° appears at 600 °C, indicating the initial formation of amorphous carbon with short-range graphitic layers [Fig. 3(a-i)]. A slight shift of the diffraction peak and continued increase in its intensity at higher carbonization temperatures from 700 to 1000 °C demonstrate the maturing carbon microstructure of ZIF-8-C acquiring a greater degree of graphitization. However, the diffraction peak at 2θ = ∼25° remains broad even at 1000 °C, revealing the non-graphitizing nature of ZIF-8-C. The degree of graphitization of carbon materials can also be investigated by Raman spectroscopy. For carbon materials, there are two characteristic Raman bands at about 1355 cm⁻¹ (D) and 1580 cm⁻¹ (G) that correspond to sp² carbon rings adjacent to an edge or a defect and the planar configuration of sp²-hybrdized carbon, respectively.²² Therefore, the intensity ratio of D and G bands (I_D/I_G) presents the degree of graphitization. In the case of ZIF-8-C, D and G bands tend to become more clearly resolved and their I_D/I_G values decrease at higher carbonization temperatures [Fig. 3(a-ii)]. This indicates that there is an increase in the degree of graphitization when ZIF-8 is subjected to greater temperatures although most of its carbon microstructure remains largely amorphous based on the XRD spectra at corresponding temperatures. Young et al. further investigated ex situ TEM observation of ZIF-8-C obtained at 700, 800, 900, and 1000 °C to understand the development of local microstructure at elevated carbonization temperatures.²³ Although the TEM images do not show a significant structural change of ZIF-8-C obtained at carbonization temperatures between 700 and 1000 °C, notable changes in their carbon microstructure can be observed from HRTEM images [Fig. 3(b-i)]. Specifically, ZIF-8-C tends to develop long-range graphitic layers around the mesopores at higher carbonization temperatures [Fig. 3(b-ii)].

In more recent studies, the real-time development of carbon microstructure was observed by a time-resolved in situ TEM analysis.^24–27 The carbonization of ZIF-8 can be achieved in TEM by in situ heating by microelectromechanical systems (MEMS) heater integrated on the chip.²⁸ Based on the in situ observation of the thermal annealing process of ZIF-8, there is no significant change in the overall morphology at 900 °C and selected area electron diffraction (SAED) pattern reveals its amorphous carbon microstructure [Fig. 4(a)].²⁹ From the HRTEM images, there is no evident development of mesopores in the carbon matrix, hence lacking the graphitic layers [Fig. 4(b)]. In the carbonization process, nitrogen (N) and zinc (Zn) atoms are gradually removed from the carbon microstructure before the temperature reaches 900 °C. However, when the annealing temperature was held at 900 °C for 2 h, a sudden dramatic decrease in the content of N and Zn is observed due to accelerated thermal decomposition of N and thermal evaporation of Zn [Fig. 4(c)]. It is known that Zn passes through the carbon matrix to create additional micropores during its thermal evaporation, therefore, contributing to a large specific surface area of ZIF-8-C.

In situ TEM studies of MOF-derived porous carbons were also carried out with ZIF-8 incorporating other metal species to observe the thermal behavior of metal species encapsulated in the crystal structure of ZIF-8 during the carbonization process.^29–32 In 2018, Wei et al. reported in situ TEM observation of ZIF-8 encapsulating palladium (Pd) nanoparticles with thermal annealing.³¹ Based on high-angle annular dark-field scanning TEM (HAADF-STEM) images, as the carbonization proceeds to 900 °C, Pd nanoparticles become larger to their crystalline intermediate I (after heating for 0.5 h), which then transform into amorphous intermediate II (after heating for 1.5 h). After 3 h of carbonization at 900 °C, the amorphous Pd nanoparticles are fully digested by ZIF-8-C support and become atomically dispersed [Figs. 5(a-i) and 5(a-ii)]. They pointed out that the presence of Pd nanoparticles within the ZIF-8-C support is important to induce thermal motion and down-sizing of Pd nanoparticles as the surface Pd atoms form coordination bonds with the N-defects of ZIF-8-C. On the contrary, the atomization of Pd nanoparticles is not feasible when they are on the surface of ZIF-8-C as their thermal motion is restricted to the edge of ZIF-8-C. Despite the presence of Pd nanoparticles in ZIF-8, the resulting ZIF-8-C doped with single Pd atoms is highly amorphous as evidenced by XRD spectra [Fig. 5(a-iii)].

When cobalt (Co) is introduced during the synthesis of ZIF-8, bimetallic Zn-Co ZIF can be effectively obtained. For Zn-Co ZIF synthesized by Zhang et al., they defined the thermal annealing process of Zn-Co ZIF by four stages based on in situ TEM observation [Fig. 5(b-i)].³⁰ In Stage 1 (up to 500 °C), tiny Co clusters are formed, and they tend to disperse atomically in Stage 2 (500–800 °C). As temperature increases to 850 °C, dispersed cobalt species aggregate to form larger nanoparticles, marking Stage 3. It is also noteworthy that there is a significant shrinkage in the overall structure and appearance of clearer porous structure in Stage 3. Finally, in Stage 4, Co nanoparticles undergo sublimation at 1000 °C, eventually dispersing Co atoms as single atom dopants in the carbon matrix after 520 s [Fig. 5(b-ii)]. The corresponding SAED pattern shows the diffraction rings for graphitic carbons, indicating that it involves graphitic carbons without crystalline Co species [Fig. 5(b-iii)]. In contrast to ZIF-8-C, the porous carbon derived from bimetallic Zn–Co ZIF (Zn–Co ZIF-C) develops graphitic layers due to the intermediate formation of Co nanoparticles, which can facilitate catalytic graphitization like iron (Fe) and nickel (Ni).^33,34 From the time-revolved observation of Co, it was noted that the molten-state of Co nanoparticle at 850 °C, having an irregular morphology and amorphous structure, is mobile and behaves like a liquid droplet to migrate randomly on the carbon support. During their migration, molten Co nanoparticles exhibit three behaviors: (1) etching the carbon support to generate larger nanopores; (2) catalyzing the formation of graphitic layers; and (3) anchoring Co single atoms onto the carbon support [Fig. 5(b-iv)].

In the case of pure Co-based ZIF (ZIF-67) derived carbon (ZIF-67-C), however, severely aggregated Co nanoparticles are formed throughout the ZIF-67-C support after carbonization at 900 °C.³⁵ Initially, small Co clusters (∼2.3 nm in diameter) start to appear at 600 °C, which then grow into larger nanoparticles as annealing temperature increases [Fig. 5(c-i)]. The development of crystalline Co nanoparticles in ZIF-67-C is further evident from the corresponding SAED patterns. In addition, a closer in situ TEM observation reveals that the catalytic graphitization is induced around the Co nanoparticle in ZIF-67-C [Fig. 5(c-ii)]. When only small Co clusters are present in the carbon matrix at 600 °C, they do not seem to induce graphitization and the carbon matrix largely remains amorphous. At 700 °C, a larger Co nanoparticle forms and short-range graphitic layers with an interlayer distance of 0.34 nm appear. The fragmented graphitic layers become more connected, and the number of graphitic layers increases to fully encapsulate the Co nanoparticle at elevated annealing temperatures up to 900 °C. The development of graphitic carbon layers in ZIF-67-C catalyzed by Co nanoparticles is also confirmed by XRD analysis, showing that the initially weak peak of graphite (002) at low annealing temperature becomes sharper and more intense from 800 °C [Fig. 5(c-iii)].

To this end, it is clear that the process of carbonization is extremely dynamic involving extensive bond breaking and formation in the carbon microstructure during heating and cooling stages. The microstructural components ranging from the defects, length of graphitic layers, degree of disorderedness, and hardness are important factors to consider for gaining deeper understanding of the structure-to-property relationship of carbon materials. Therefore, the study on the development of carbon microstructure of MOF-derived porous carbon is essential, and facile synthetic methods to control the distribution of specific microstructural component need to be devised to achieve the desired properties required for various applications.

In the case of porous carbon materials being used as electrode materials, the most basic form of electrochemical interaction occurring at the interface between the carbon surface and the electrolyte solution is electrical double layer (EDL) formation. Typically, the EDL is formed by the electrostatic force between the polarized carbon surface and counter ions in the applied electric field to induce electrosorption of ions on the surface. As ions electrostatically adsorb onto the polarized carbon surface, charge continues to accumulate and gets stored in the porous carbon materials, therefore, serving as one of the major charge storage mechanisms in energy-storage applications (Fig. 6).³⁶ The formation of EDL on the carbon surface is therefore non-faradaic process, which does not involve any charge transfer at the electrode–electrolyte interface.

To explain the distribution of charges at the interface between the charged electrode and electrolyte solution, a variety of models have been proposed as described in Fig. 7. The first model for EDL formation proposed was the Helmholtz model, which describes the arrangement of counter ions along the Helmholtz plane near the electrode surface [Fig. 7(a)]. However, there are a few limitations of the Helmholtz model as follows: (1) it assumes a finite distance between the electrode surface and the Helmholtz plane, which is not the case in the real system; (2) the effect of electrolyte ions and other factors in the system existing further away from the electrode surface is neglected; and (3) chemical adsorption of ions on the surface is not considered. The Gouy–Chapman model was next proposed to complement some of the misconceptions presented in the Helmholtz model by describing that the counter ions are present not only near the electrode surface but also along the diffuse layer further away in the system following the Boltzmann distribution, therefore compensating for the remaining potential difference [Fig. 7(b)]. Although this model covers the effect of distribution of ions in the diffuse layer, it does not consider the effect of the size of ions because ions are expressed as point charges. The consideration of the ions as point charges causes the effective separation distance between the electrode surface and the ions to decrease to zero at high polarization, which is not realistic. The following Stern model combined the previous two models and divided the portion of electrolyte solution into three parts consisting of a compact layer (or Stern layer), diffuse layer, and bulk layer [Fig. 7(c)]. In the Stern model, the thickness of the compact layer is generally determined by the radius of ions, which mostly compensate the potential of the electrode. This renders the contribution of the charged ions in the diffuse layer and bulk layer sufficiently low. In the following Grahame model, it was proposed that the chemical adsorption of ionic or uncharged species can occur directly on the surface of electrode [Fig. 7(d)]. Consequently, the presence of specifically adsorbed ions on the electrode resulted in further dividing the EDL into four regions. Typically, the inner Helmholtz plane lies along the specifically adsorbed species, and the outer Helmholtz plane is formed where the electrochemical adsorption of the solvated ions occurs near the surface of electrode. Diffuse layer and bulk layer are in the region beyond the outer Helmholtz plane as in the Stern model. The Bockris-Devanathan-Müller (BDM) model further elaborated the effect of the solvent on the formation of EDL [Fig. 7(e)]. In this model, the inner Helmholtz plane consists of specifically adsorbed ions and solvent molecules, and the outer Helmholtz plane is located where the electrochemical adsorption of solvated ions occurs.

Based on the previous models describing the formation of EDL, the surface area of electrode materials is generally treated as one of the most important factors in achieving a greater level of EDL formation.³⁷ To this end, nanoporous carbon materials are considered as the most preferred type of EDL-forming materials.^5,23,38 This leads to a simple but often false analogy that nanoporous carbon materials of higher specific surface area would have a higher level of EDL formation. Young et al. clearly demonstrated the observation of higher specific capacitance (i.e., greater level of EDL) from ZIF-8-C with a specific surface area of 1591 m² g⁻¹ as compared to the typical activated carbon with a much higher specific surface area of 2370 m² g⁻¹.²³ This counter proves that the idea of having the greater surface area is not always advantageous to achieve the greater level of EDL formation. In addition, despite ZIF-8-C achieving a higher level of EDL as compared to the activated carbon, its capacity to form EDL rapidly deteriorates when the change in current or voltage occurs to a much greater extent. Such a discrepancy in the electrochemical properties of porous carbon materials is mainly attributed to their three-dimensional porous structure, making the formation of EDL more complicated in comparison to the electrode with the planar surface as described in the previous models [Figs. 7(a)–7(e)].^36,38

For the porous carbon materials, the effect of pore size plays a significant role in the EDL formation due to the three-dimensional nature of their porous structure.^38–40 Depending on the size of nanopores, the surface area of porous carbon materials can greatly vary in a way that the smaller the size of nanopores, the larger the surface area of the porous carbon material. Microporous carbon materials, in particular, present extremely high surface area above 1000 m² g⁻¹. Interestingly, however, some microporous carbon materials demonstrate the opposite trend of the inverse relationship between the capacitance and the surface area.⁴⁰ In addition, when higher current is applied to the microporous carbon electrode, a complete utilization of surface for the EDL formation is not always ensured because the current transition is more rapid than the ionic motions in the electrolyte.^41–43 This results in even lower level of EDL formation at higher current densities. Therefore, more appropriate and comprehensive consideration of the EDL formation in the three-dimensional porous structures is necessary to better understand the relationship between the pore size and capacitance (i.e., electrochemically accessible surface area). When the pore is modeled in the form of cylinder, the equation for the capacitance of a cylindrical pore model is derived as below (Fig. 8).

From “Eq. (2),” the capacitance of cylindrical pore is linearly related to the area of cylinder but inversely related to the size of nanopore. However, as the size of nanopore is continually reduced, it may become equal to or smaller than the thickness of the EDL. This is when a phenomenon called EDL overlapping appears prominent and causes the capacitance to decrease despite the size of nanopore becoming even smaller (i.e., greater surface area) [Figs. 9(a-i) and 9(a-ii)].³⁶ In the presence of the EDL overlapping, the amount of charge that can accumulate on the surface of the pore substantially diminishes, therefore, making the smaller nanopore adversely contribute to the capacitance [Fig. 9(b-i)]. Apart from the EDL overlapping, highly restricted ionic motions in the inner surface of the nanoporous carbon materials also present the limitation. The restricted ionic motions result in the increased level of ionic inaccessibility to the internal pores, therefore, worsening the electrochemical utilization of the surface area.⁴³ 

To ensure the efficient generation of the EDLC in the porous structure, the effect of the EDL overlapping must be minimized while maximizing the electrochemical accessibility of the surface area. The EDL overlapping can be effectively relieved by successfully controlling the pore size to be larger than the thickness of the EDL [Fig. 9(b-ii)].⁴² Maximizing the electrochemical accessibility of the surface area can also be achieved by incorporating the nano-confined space in the porous carbon materials. Unlike porous carbons without nano-confined space, which experience a significant impediment in ionic diffusion, porous carbons with nano-confined space facilitate more rapid ionic motions by shortening the ionic diffusion pathway with the nano-confined space serving as a separate reservoir for the electrolyte [Figs. 9(c) and 9(d)].⁴² 

As porous carbon materials can be thermally derived from porous materials involving a high carbon content, the research on the synthesis of novel porous materials and their successful conversion to porous carbon materials has progressed rapidly. Conventional porous materials are synthesized by template-based or template-free methods, and the resulting porous materials can be subjected to high-temperature pyrolysis to obtain the corresponding porous carbon materials.

Most template-free methods are established on the basis of self-assembly process, which occurs spontaneously under specific conditions, therefore, ensuring uniform distribution of highly homogeneous nanopores [Fig. 10(a)].^44,45 This makes the methods highly simple and convenient because the need to prepare and remove templates is effectively omitted. The simplicity of the template-free method, in turn, leads to high cost-effectiveness, energy efficiency, and environmental friendliness in fabricating high-quality porous carbon precursors. However, the template-free method also presents a number of limitations: (1) it is a highly dynamic process, and a high level of control over the reaction is rather difficult to achieve, (2) the aggregation of the self-assembled porous materials is often inevitable, (3) the nanopore size of the self-assembled porous materials is generally limited to the micropore range, and (4) it offers the low level of control over the size, dimension, and structure of the nanopore.⁴⁶ It is also often a limited approach, especially when considering the feasibility of the self-assembled materials [e.g., metal-organic frameworks (MOFs), covalent organic frameworks (COFs), zeolites, etc.] for the direct-carbonization.^47–49 For instance, certain metal species in the self-assembled materials are likely to undergo thermal diffusion and aggregation via Ostwald ripening to form metal nanoparticles during the direct-carbonization process. The formation of metal nanoparticles in the self-assembled materials can cause the original morphology and porosity to deform.^50,51 As different methods to synthesize porous materials as carbon precursors present their distinct advantages and disadvantages, it is necessary to carefully examine the available synthetic methods prior to the material designing and synthesis.

In 1995, the synthesis of highly porous, and crystalline self-assembled materials, known as MOFs, was reported by Yaghi et al., and there has been a growing interest in this new class of porous materials because of their unique properties.^52,53 The time-resolved observation for the self-assembly of ZIF-8 was reported by Liu et al. in 2021.⁵⁴ With in situ liquid-phase TEM observation during the nucleation of ZIF-8 nanocubes, three different stages of ZIF-8 nucleation are suggested: (1) phase separation between solute-rich and solute-poor regions in the reaction solution at 15 s; (2) gradual condensation of the solute-rich regions at 31 s; and (3) transformation to ZIF-8 cubes at 62 s [Fig. 10(b)].⁵⁴ Particularly, MOFs are formed by highly ordered coordination bonding between inorganic and organic units at specific bond angles, orientations, and lengths, thus consisting of periodically arranged unit cells serving as nanopores. As a result, MOFs often present extremely large surface area and ultrahigh porosity. On the contrary to their obvious structural advantages, MOFs generally suffer from poor physical and chemical stabilities and insufficient electrical conductivity to be considered useful in a range of applications, especially in energy storage or conversion devices. Since such limitations of MOFs can be well complemented by typical properties of carbon materials (e.g., outstanding physical and chemical stabilities and great electrical conductivity), intensive research efforts were dedicated to integrating the properties of MOFs and carbons. Consequently, the concept of a direct-carbonization of MOFs was introduced to overcome the complexity and environmental burdens in synthesizing novel nanoporous carbons.^3–5 

As an early example, Hu et al. reported the feasibility of direct-carbonization method to obtain porous carbons from a type of MOF known as aluminium (Al)--based porous coordination polymer (Al-PCP) [Fig. 10(c-i)].⁴ In a typical synthesis, Al-PCP was simply subjected to high-temperature pyrolysis at temperatures ranging from 500 to 800 °C under a nitrogen (N₂) atmosphere and Al species in the carbons were subsequently washed with hydrogen fluoride (HF) solution to obtain the corresponding PCP-x (where x indicates the pyrolysis temperature) [Figs. 10(c-ii) and 10(c-iii)]. In their study, PCP-800 was identified as the optimized sample with an extremely high surface area. In terms of carbon microstructure, PCP-800 was identified as highly amorphous carbon based on the SAED pattern [inset of Fig. 10(c-ii)]. In the following year, Torad et al. also reported the use of ZIF-8 for the production of nanoporous carbon by the direct-carbonization as described in Fig. 10(d-i).⁵⁵ ZIF-8 is a type of MOF formed through the formation of coordination bonding between Zn ion (Zn²⁺) and 2-methylimidazole (2-meim) [Fig. 10(d-ii)]. As a carbon precursor, ZIF-8 presents particular advantages over other MOFs that: (1) it can be easily synthesized under ambient conditions, (2) presents good thermal stability, (3) involves Zn that evaporates during the direct-carbonization due to low boiling point of (907 °C), and (4) results in N-doped porous carbon due to the presence of N in 2-meim. Upon direct-carbonization, C and N atoms of 2-meim in ZIF-8 become thermally unstable, and bond rearrangement takes place to form the porous carbon matrix with N heteroatoms, while Zn evaporates to leave additional nanopores behind in the carbon matrix. It is interesting to note that the nanopores of ZIF-8 are well-maintained although a certain level of particle shrinkage and thermal demolishment of nanopores are possible at high pyrolysis temperatures above 1000 °C [Figs. 10(d-iii) and 10(d-iv)].⁵⁵ The resulting ZIF-8-C therefore has a high specific surface area of ∼1000 m² g⁻¹ and typical attributes of most carbon materials such as electrical conductivity and structural strength. Apart from ZIF-8-C, other MOF-derived carbon materials have been subsequently reported to extend the scope of applications to environmental remedy, energy storage and conversion, sensing, biomedical applications, etc.

As the engineering of porous materials became more sophisticated and controllable with the new discoveries, there had been meaningful attempts to synthesize novel porous carbon materials by implementing template materials, including silica, aluminosilicate, colloidal particles and crystals, and self-assembled polymers.^56–62 The template materials that are rigid and well-defined in terms of shape are termed as “hard-templates,” and their size and morphology directly reflect the resulting nanoarchitecture of porous materials and corresponding carbon derivatives.⁶³ Typical steps of hard-template methods include template synthesis, filling precursors to the hard templates, polymerization or crystallization of precursors, and finally, carbonization and template removal to obtain porous carbon materials [Fig. 11(a)]. Although the use of hard templates can ensure the synthesis of highly ordered and well-defined porous materials, it presents clear limitations such as the use of harsh and toxic chemicals to synthesize and remove hard templates, less adjustable pore size and structures, and low yield caused by inevitable precursor nucleation outside of porous structures.⁶³ 

An alternative class of template materials termed as “soft-templates” has also been implemented for the synthesis of porous carbon materials. The soft-templates are supramolecular polymeric aggregates or micelles derived from amphiphilic surfactants or block copolymers.⁶³ Typically, the amphiphilic nature of surfactants or block copolymers drives the formation of micelles via a self-assembly occurring under suitable conditions in which hydrophobic heads interact with each other to form a hydrophobic core. In contrast, their hydrophilic tails interact with the surrounding polar molecules such as water molecules and project outward to stabilize the micelles.⁶⁴ Typical steps of the soft-template method involve micellization and precursor binding process, polymerization or crystallization of precursors, and removal of the template to obtain porous materials [Fig. 11(b)]. Due to the polymeric nature, the soft templates or the micelles do not have as well-defined and rigid structures as the hard-templates. However, the softness of the micellar pore-directing agents can be advantageous in different ways by offering an easy removal of templates with commonly used organic solvents and a significantly higher level of freedom in controlling the pore size of resulting materials upon varying the micellization conditions to alter the size of hydrophobic core. Examples of porous carbon materials synthesized by the soft-template methods were generally derived from a variety of polymers.^65–71 

In 1980s, the first use of silica gel as the form of hard templates to synthesize porous glassy carbon was reported.^56,72 After impregnating the silica gel with polymer precursors, the composite material was subjected to thermal annealing and dissolution of silica gel to finally obtain the porous glassy carbon with mesopores and a high specific surface area up to 600 m² g⁻¹. However, the formation of highly ordered mesopores was not possible due to the irregularly ordered silica gel causing the disordered porous structure. In 1997, Kyotani et al. demonstrated the use of zeolites as the hard templates to synthesize porous carbons.⁷³ Few years later in 2000, the replica of the fine crystal structure of zeolites was successfully achieved by implementing the two-step method (precursor impregnation and subsequent chemical vapor deposition with 2.0% propylene gas in N₂).⁷⁴ The resulting porous carbon produced a sharp XRD peak at 2θ = 6.26, indicating the presence of long-range pore ordering with a periodicity of micropores with a pore diameter of 1.41 nm. To achieve the high orderliness of the mesopores in the carbon materials, more sophisticated form of hard templates was synthesized and used. For instance, Ryoo et al.⁷⁵ and Lee et al.⁷⁶ in 1999 successfully demonstrated the potential use of highly ordered mesoporous silica (SiO₂) as hard templates to synthesize ordered mesoporous carbons, referred to as CMK-1 and SNU-1, respectively [Figs. 12(a-i) and 12(a-ii)]. In a typical synthesis, the ordered mesoporous silica was first prepared by soft-template methods and subsequently impregnated with carbon precursors. Next, the ordered mesoporous silica impregnated with carbon precursors was pyrolyzed at 800–900 °C, and the silica template was removed to finally obtain the ordered mesoporous carbons. Later, Ryoo's group reported CMK-3 and CMK-5 synthesized using another type of hard-template typically known as SBA-15.^77,78 SBA-15 consists of randomly interconnected cylindrical channels of diameter between 6 and 15 nm and carbon materials synthesized using SBA-15 produce interconnected mesoporous carbon. When carbon precursor is allowed to fully impregnate the mesopores of SBA-15, CMK-3 with interconnected carbon rods is synthesized [Figs. 12(a-iii) and 12(a-iv)],⁷⁷ but a partial impregnation on the pore walls of SBA-15 results in CMK-5 with interconnected carbon tubes [Figs. 12(a-v) and 12(a-vi)].⁷⁸ Numerous porous carbons were derived from different carbon precursors (e.g., furfuryl alcohol, acetonitrile, ethylene, phenol-formaldehyde resins, resorcinol-formaldehyde resins, and other aromatic precursors).^77–79 Alternative to porous silica gel, silica particles were also used to produce various porous carbons.^61,80–82 Specifically, Li and Jaroniec adopted colloidal imprinting approach using mesophase pitch as a fluid-type carbon precursor that softens to allow penetration of colloidal silica particles at 260 °C [Fig. 12(b-i)].⁶¹ The resulting colloidal silica-pitch composite was then carbonized at 900 °C and treated with 3M NaOH to obtain highly disordered mesoporous carbon [Fig. 12(b-ii)]. Monodispersed colloidal silica can self-assemble to ordered silica upon slow evaporation of solvent. The ordered silica template can then be impregnated with carbon precursor to produce ordered mesoporous carbon upon carbonization and template removal [Figs. 12(c-i) and 12(c-ii)].⁸³ In another case, monodispersed polystyrene (PS) beads were implemented to replace silica to obtain ordered hard-template that can be easily removed during carbonization [Fig. 12(c-iii)].⁸⁴ In 2001, Lee et al. used mesocellular aluminosilicate foams as hard-templates to obtain mesocellular carbon foam with a large mesopore of ∼27 nm in diameter [Figs. 12(d-i) and 12(d-ii)].⁸⁵ Anodic aluminum oxide (AAO) film with densely populated aluminum oxide mesochannels was also used as a type of hard-template by Kyotani et al. to synthesize high aspect ratio carbon tubes [Fig. 12(e)].⁸⁶ 

Contrary to the hard-template, the use of soft-template allows more facile synthesis of porous materials without the need to use harsh chemicals to remove them while also ensuring versatile tuning of their sizes and morphologies. This is because the self-assembly of amphiphilic surfactants to soft-templates (i.e., micelles) is induced by the thermodynamically preferred weak interaction between hydrophobic domains and their reduced contact with polar molecules in the system rather than strong covalent bonds.⁸⁷ Therefore, the self-assembled micelles are more susceptible to disruption upon the changed polarity of the solvent or thermal decomposition. In dilute conditions, the amphiphilic surfactants tend to self-assemble to different morphologies depending on their packing behaviors with typical examples of sphere, cylinder, vesicle, and lamella [Fig. 13(a)].

In addition to the packing behavior of surfactants, the volume fraction of the hydrophilic component (f_A) of the surfactant can serve as an alternative parameter that predicts the self-assembled morphologies [Fig. 13(a)].

Generally, the self-assembled morphology takes inverse sphere when f_A < 0.25 and gradually transits to lamellar ( f_A $≈$ 0.50), vesicular (0.35 < f_A < 0.45), cylindrical (0.40 < f_A < 0.50), and spherical (0.45 < f_A) morphologies.^90,91 The effect of f_A becomes more significant when surfactants are present at high concentration. When the concentration of surfactant is increased by various means (e.g., solvent evaporation), more condensed forms of self-assembly take place to give rise to lyotropic liquid crystals (LLCs) with high periodicity in their morphological arrangement. As the value of f_A increases, LLCs implement morphologies such as closely packed spheres with a face-centered (fcc) or body-centered (bcc) lattice, hexagonally packed cylinders, bicontinuous gyroids, lamellae, inverse hexagonally packed cylinders, and inverse closely packed spheres (inverse face-centered cubic or body-centered cubic), therefore, offering a much greater level of morphological control for the soft-template and the resulting porous materials.^87,92 Apart from the two predictive parameters of p and f_A, other factors such as the type of surfactants (e.g., molecular weight, number of blocks, ionic/nonionic, etc.), additives, temperature, pressure, light, magnetic field, and polar/non-polar solvent ratio play significant roles in determining the final self-assembled morphologies of the surfactants.^93–98 

In 2004, Liang et al. made the first report of the successful synthesis of a highly ordered mesochanneled carbon film by implementing a type of diblock copolymer, poly(styrene)-b-poly(4-vinylpyridine) (PS-b-P4VP), as the source of soft-template [Fig. 13(b-i)].⁶⁵ The interaction between PS-b-P4VP and resorcinol is mainly established by hydrogen bonding between resorcinol and P4VP block. They claim that this hydrogen bonding induces the swelling of P4VP block, thus increasing the f_A value of PS-b-P4VP in the solution, resulting in the formation of cylindrical micelle. Subsequently, the solution of cylindrical micelle composite of PS-b-P4VP and resorcinol was spin-coated on the silicon substrate to undergo supramolecular vertical assembly prior to the polymerization to PS-b-P4VP/phenolic resin. The resulting PS-b-P4VP/phenolic resin film was subsequently subjected to direct-carbonization at 800 °C to get highly ordered mesochanneled carbon film. During the thermal annealing process, the cylindrical micelle consisting of PS-b-P4VP was removed by thermal decomposition. Based on the cross-sectional SEM image, the mesochannels derived from the cylindrical micelles demonstrate continuous opening from the bottom to the top of the carbon film [Fig. 13(b-ii)]. From the top view, the highly uniform diameter of mesochannels (∼33.7 nm) is clearly observed [Fig. 13(b-iii)]. This indicates that the formation of highly uniform mesopores is still possible using soft-templates that are much less rigid as compared to hard-templates. Alternatively, triblock copolymers consisting of polyethylene oxide (PEO) and polypropylene oxide (PPO) chains (PEO_x–PPO_y–PEO_x) are widely used for the synthesis of mesoporous carbon materials. For example, Tanaka et al. adopted similar spin-coating method but used a type of triblock copolymer, Pluronic F127 (PEO₁₀₀–PPO₆₅–PEO₁₀₀, M_w = 12 600), to form cylindrical micelles and synthesize highly ordered mesoporous phenolic resin and its carbonaceous form (COU-1) on silicone substrate [Fig. 13(c-i)].⁶⁶ It is noteworthy that they also involved triethyl orthoacetate (EOA) as a carbon co-precursor to improve the stabilization of the highly periodic mesoporous structure. Indeed, the SEM images of COU-1 demonstrate a highly periodic structure with long-range ordering, thus yielding a typical hexagonal arrangement of mesopores (∼5.9 nm in diameter) [Figs. 13(c-ii) and 13(c-iii)]. Later, Meng et al. demonstrated that the use of different types of PEO_x–PPO_y–PEO_x type of triblock copolymers can lead to the synthesis of a family of highly ordered mesoporous phenolic resin, including 2D hexagonal, (p6m), 3D cubic (Im $3¯$m), 3D bicontinuous (Ia $3¯$d), and lamellar mesostructures.⁹⁹ To obtain several types of ordered mesoporous phenolic resin, they implemented the solvent evaporation induced self-assembly (EISA) method to effectively induce a gradual increase in the concentration of triblock copolymer, thus inducing the formation of lyotropic liquid crystals and polymerization on glass dish. The study describes the effect of phenol/template ratio and hydrophilic–hydrophobic balance of the template on the formation of specific mesostructures. They claimed that the increase in the phenol/template ratio leads to the swelling of hydrophilic PEO blocks and gives more curvature to the resulting template. The increased hydrophilic volume and curvature of the template, in turn, results in the phase transformation from lamellar to bicontinuous cubic, hexagonal, and body-centered cubic phases. To this end, a type of triblock copolymer P123 (PEO₂₀–PPO₇₀–PEO₂₀, M_w = 5800) can give rise to mesoporous polymer FDU-14 with bicontinuous cubic mesostructure with Ia $3¯$d symmetry when phenol/P123 molar ratio is within a narrow range of 1.0:0.018 to 1.0:0.019 [Figs. 13(d-i) and 13(d-ii)].⁹⁹ Alternatively, at higher P123 concentration (i.e., phenol/P123 molar ratio = 1.0:0.022), lamellar mesostructures appear, whereas hexagonal mesostructures appear at lower P123 concentration in the range of 1.0:0.007 to 1.0:0.016. Likewise, with F127, either hexagonal mesostructure (at higher F127 concentration) or body-centered cubic mesostructure with Im $3¯$m symmetry (at lower F127 concentration) can be formed in FDU-15. Since F127 has much longer hydrophilic blocks and higher hydrophilic–hydrophobic balance than P123, increased phenol/F127 molar ratio can generate more curvature of the template, therefore, giving hexagonal and body-centered cubic mesostructures but not lamellar and bicontinuous cubic mesostructures. When F108 with even longer hydrophilic blocks and higher hydrophilic–hydrophobic balance is used to synthesize FDU-16, only body-centered cubic mesostructure with Im $3¯$m symmetry appears even from low F108 concentration [Figs. 13(d-iii) and 13(d-iv)].

More recently, an alternative type of polymer known as polydopamine (PDA) has been widely used as the source of carbon precursor to obtain a variety of well-defined porous structures based on the templated methods involving hard templates or soft-templates or mixed-templates.^71,100–103 The use of PDA as carbon precursor is particularly beneficial because of the following characteristics: (1) mild polymerization conditions (at ambient temperature, in air and weakly alkaline pH of ∼8.5); (2) highly controllable polymerization by varying factors such as polymerization time, pH, oxygen concentration in solution, etc. (3) exceptional adhesive property to form surface chemical interactions with almost any materials; (4) intrinsic content of O and N for heteroatom-doped carbon; (5) high chelating property to metal cations for metal-doped/loaded carbon; and (6) catechol and amine groups of dopamine (DA) can tune the volume of hydrophilic blocks of soft-templates, leading to altered micelles. As an early example, Liu and Dai et al. adopted the hard-template approach and successfully coated PDA on SiO₂ to obtain a core-shell structure of SiO₂@PDA.¹⁰¹ The core-shell structured material was then pyrolyzed at 800 °C followed by SiO₂ removal by HF to obtain a hollow microporous carbon [Fig. 14(a-i)].¹⁰¹ The HAADF-STEM images of the resulting hollow microporous carbon clearly demonstrate the hollow cavity with an ultrathin carbon shell of 4 nm [Figs. 14(a-ii) and 14(a-iii)]. With appropriate choice of micelle derived from a specific block copolymer, mesoporous PDA could also be synthesized. Tang et al. demonstrated the use of polystyrene-block-poly(ethylene oxide) (PS-b-PEO) that forms spherical micelle to synthesize mesoporous PDA particles with a highly uniform particle size of ∼200 nm [Fig. 14(b-i)].⁷¹ From SEM images, the incorporation of micelle to the polymer is well observed in comparison to polymer without micelle [Figs. 14(b-ii) and 14(b-iii)]. The as-synthesized mesoporous PDA was subsequently converted to mesoporous carbon by the pyrolysis process at 800 °C. The resulting mesoporous carbon possesses an average mesopore size of 16 nm [Fig. 14(b-iv)]. In this work, the control over the size of mesopores was also demonstrated by implementing alternative types of PS-b-PEO to form micelles. In the follow-up study, a dual-templating method involving both hard-template and soft-template was attempted to endow both ordered mesopores and hollowness in the PDA-derived carbon particles [Fig. 14(c-i)].¹⁰³ In a typical synthesis, SiO₂ was first synthesized to serve as a hard-template and micellization was then initiated in the presence of DA monomers. To the mixture of the dual-templates and monomers, the polymerization was induced to successfully obtain SiO₂@mesoporous PDA core-shell structure [Figs. 14(c-i)–,14(c-iii)]. Subsequent pyrolysis and etching of SiO₂ eventually gave rise to hollow mesoporous carbon [Fig. 14(c-iv)].

The template-free and/or template-based synthetic methods therefore provide the foundation for the nanoarchitecturing of porous carbon materials. By investigating more elaborate and finely controlled synthetic approaches originating from them can effectively lead to novel morphologies and nanoarchitectures of MOF-derived porous carbons. Therefore, their importance and usefulness in nanoarchitecturing of MOF-derived porous carbons are covered in Sec. V of this review.

Template-free nanoarchitecturing of MOFs can be mainly categorized as bottom–up or top–down approaches. Specifically, the bottom–up approaches involve the modification of the initial building blocks of MOFs, altered crystallization tendencies, and inducing discrepancy in thermal stability between MOFs and their modified surfaces to alter the pristine nanoarchitecture of MOFs in desired ways. Examples of the bottom–up approaches include ligand engineering, surfactant-assisted crystal engineering, compositional engineering, and surface engineering. On the other hand, the top–down approaches exploit the already-grown MOFs for the further modifications by various means such as the cation-exchange of metals of MOFs, selective etching of specific elements or crystal facets of MOFs and partial dissolution and crystallization of MOFs to alter the nanoarchitectures. Nanoarchitecturing of MOFs and their porous carbons by various template-free methods has clear advantages of facile modification processes, potentially less harming to environment, more suitable for practical applications as it is often more appropriate for the large-scale materials synthesis. However, it also accompanies a few downsides such as the limited control over the size of nanopores, difficulty achieving periodically ordered nanopores, and the restricted use of certain methods for specific MOF candidates having the target elements or crystal facets.

The template-free nanoarchitecturing of MOFs via bottom–up approaches can be achieved through various means. Among them, the epitaxial growth of MOF on MOF has been widely adopted to integrate the properties and/or compositions of multiple MOFs in a single heterostructure.^104–109 For instance, Zn-based ZIF-8 and Co-based ZIF-67 were first selected for the epitaxial growth to obtain a typical core-shell structure because they share similar crystal lattice structures and parameters (ZIF-8: a = b = c = 16.9910 Å and ZIF-67: a = b = c= 16.9589 Å) as determined by the single crystal x-ray diffraction studies.^110,111 In 2015, Tang et al. first demonstrated the core-shell structure of ZIF-8 core and ZIF-67 shell by first synthesizing the uniformly sized ZIF-8 seeds and subsequently inducing the epitaxial growth of ZIF-67 on ZIF-8 seeds (ZIF-8@ZIF-67).¹⁰⁷ The resulting ZIF-8@ZIF-67 was subjected to the direct-carbonization to obtain porous carbon materials with the varied microstructure between the core and the shell derived from ZIF-8 and ZIF-67, respectively. As compared to ZIF-8-C that exhibits high amorphousness and abundant N heteroatoms in its carbon microstructure, ZIF-67-derived nanoporous carbons possess significantly more graphitic microstructure because Co species in ZIF-67 act as thermal catalysts to induce the graphitization of carbon during the thermal annealing process. Consequently, ZIF-8@ZIF-67 derived nanoporous carbons present N-doped amorphous carbon (NC) in the core and graphitic carbon (GC) in the shell, eventually giving rise to NC@GC [Fig. 15(a-i)].¹⁰⁷ The core-shell structure of ZIF-8@ZIF-67 is well observed from the electron dispersive x-ray spectroscopy (EDS) elemental mapping showing Zn (green) in the core and Co (red) in the shell of each particle [Fig. 15(a-ii)]. As ZIF-67 grows epitaxially from the surface of ZIF-8, the overall morphology of ZIF-8@ZIF-67 follows the original rhombic dodecahedral (RD) shape of ZIF-8, and ZIF-67 shell of thickness about 170 nm grew on ZIF-8 of diameter ∼4 μm as observed from the SEM image [Fig. 15(a-iii)]. After the direct-carbonization and etching of Co species, the pristine RD shape is well-maintained along with the development of some carbon nanotubes on the carbon surface [Fig. 15(a-iv)]. The randomly oriented short stacks of graphitic layers can be easily found in the carbon microstructure from the HRTEM image [Fig. 15(a-v)]. Through the synthesis of the core-shell structure, the advantages of ZIF-8-C (high surface area and abundant N) and ZIF-67-C (high graphitization and good stability) are effectively integrated in a single carbon particle of NC@GC.

The concept of epitaxial growth-induced core-shell structure of MOF was also adopted by Chen et al. to synthesize the multishell ZIF-8@ZIF-67 by the repeated epitaxial growth of ZIF-8 on ZIF-67 or ZIF-67 on ZIF-8 up to 8^th layers [denoted as 8L-ZIFs(67) in Fig. 15(b)].¹¹² After the direct-carbonization, ZIF-8 layers in the multishell structure disintegrate as Zn atoms evaporate, creating the hollow cavity between the Co and N co-doped carbon layers derived from ZIF-67 (Co@NC) [Fig. 15(b-i)]. As a result, when ZIF-67 is used as the initial seed, the yolk-shell structure with hollow cavities between the layers is obtained [Figs. 15(b-ii)–15(b-iv)] but when ZIF-8 is used instead as the initial seed, the hollow structure with hollow cavities between the layers is obtained [Fig. 15(b-v)]. Their work also demonstrated that the successful alloying of Co with additional metal precursors (Cu, Fe, Pd, Cu, and Cu + Pd) in the resulting layered hollow carbons by introducing them during the synthesis of ZIF-8 seed [Fig. 15(b-v)].

It is well known that the existing metal-ligand coordination bonds in MOFs are broken, while new chemical bonds between carbon atoms and heteroatoms or metals are formed during the carbonization process. In addition to the formation of bond with carbon atoms, metal species in most MOFs undergo a thermodynamically spontaneous process called Ostwald ripening to form larger metal nanoparticles that are energetically more favored to lower the surface energy. The thermal migration of metal atoms or clusters during Ostwald ripening, in turn, creates additional nanopores in the resulting MOF-derived porous carbons. The formed metal nanoparticles can be chemically etched to form additional nanopores that are generally in the range of mesopores. When Ostwald ripening is intelligently applied and controlled in the synthesis of MOF-derived porous carbons, it can offer alternative ways of nanoarchitecturing in the bottom–up manner. Indeed, bimetallic ZIFs involving homogeneously distributed Zn and Co in a single ZIF crystal have been widely studied to expand the pristine micropore of ZIF-8-C to mesopore range in Zn–Co ZIF-derived porous carbon by exploiting Ostwald ripening of Co species.^113–115 During the carbonization, the Co nanoparticles act as thermal catalysts to induce the formation of graphitic carbons, and the resulting Zn–Co ZIF-derived porous carbons are generally doped with electrochemically useful Co atoms in their carbon matrix.

In addition to Co, other metal species (e.g., Fe, Cu, Mn, Ni, etc.) have also been introduced to synthesize bimetallic or multi-metallic MOFs and their carbon derivatives with additional mesopores.^116–124 The host–guest interaction between MOFs and metal species is generally considered an effective way to introduce additional metal species to the porous network of MOFs. For instance, FeCl₃ was introduced as guests to its host bimetallic Zn–Co ZIF to obtain Zn–Co ZIF encapsulating FeCl₃ (ZnCoFe-ZIF). The direct-carbonization of ZnCoFe ZIF gives rise to the typical hollow structure with abundant mesopores in the carbon co-doped with Co, Fe, and N [(Fe,Co)/N-C] [Figs. 16(a-i) and 16(a-ii)].¹²² The hollow structure mainly arises due to the Fe species that form during the carbonization process as they accelerate the decomposition of metal–imidazolate–metal linkages while catalyzing the graphitization of the carbon. The resulting (Fe,Co)/N-C demonstrates highly dispersed single Co and Fe atoms doped in the carbon microstructure [Figs. 16(a-iii) and 16(a-iv)]. Similarly, Jiang et al. reported the encapsulation of Fe(II) phthalocyanine (FePC) molecules in the nanopores of ZIF-8. The resulting FePC encapsulating ZIF-8 (FePC-x@ZIF-8, where x indicates the mass of FePC added during the synthesis of ZIF-8) was subsequently carbonized at 900 °C to obtain Fe single atom (SA) and Fe₂O₃ loaded N-doped porous carbon (Fe SAs- & Fe₂O₃-N/C-x) [Fig. 16(b-i)].¹¹⁸ As the loading of FePC increased from 8 to 20 mg, more Fe₂O₃ were present in the carbon matrix, which were then removed to leave mesopores behind according to the TEM images of the respective carbons [Figs. 16(b-ii)–16(b-iv)].¹¹⁸ Their N₂ adsorption/desorption isotherms indicate the co-existence of micro- and mesopores with the sharp N₂ uptake at relative pressure (P/P₀) lower than 0.05 and a mild hysteresis at P/P₀ at around 0.45 [Fig. 16(v)]. The surface area calculated by the Brunauer–Emmett–Teller (BET) method (S_BET) and the total volume of mesopores up to 30 nm (V_meso) tend to increase from ZIF-8-C to Fe SAs-N/C-x as micropores and mesopores are formed more abundantly after carbonization and Fe₂O₃ removal [Figs. 16(b-v) and 16(b-vi)].

Unlike the direct-carbonization of bi- or multi-metallic MOF yielding metallic nanoparticles and enlarged nanopores in meso- and macropore range, the surface polymeric coating of MOF with surfactant was proven effective to keep the internal microporous structure while mitigating the agglomeration of metal atoms during the carbonization [Fig. 17(a-i)].¹²⁵ He et al. successfully demonstrated the surface coating of bimetallic Co-ZIF-8 with nonionic triblock copolymer F127 (Co-ZIF-8@F127) [Figs. 17(a-ii) and 17(a-iii)]. Upon carbonization, the carbon shell becomes partially graphitized due to the graphitization of F127, while the ZIF-8-derived carbon core is largely amorphous but porous [Fig. 17(a-iv)]. The resulting core-shell structured carbon consists of homogeneously distributed with Co and N in the carbon matrix, indicating the metal and heteroatom doping [Fig. 17(a-v)]. More importantly, a significant confinement effect is generated by the strong cohesive interface interaction between Co-ZIF-8 and F127 surfactant, therefore, resisting the thermal collapse of the micropores and the agglomerating behavior of Co single atoms. Evidently, well-dispersed isolated Co atomic sites are clearly observed at the edge sites and in the carbon matrix of Co–N–C@F127 [Fig. 17(a-vi)].

The surface coating method can be further elaborated to endow the hollow structure to MOF-derived porous carbons. Among a variety of nanoarchitectures, the hollow structure is of a high importance due to the unique advantages including its contribution as a reservoir to accommodate reactants to shorten the diffusion pathways and allow more readily supply of reactants to the active sites to increase the reaction kinetics. Consequently, there have been many studies reporting the hollow structure in MOF-derived porous carbons achieved in bottom–up manner, with the most common example being the surface coating of MOFs with organic or inorganic materials prior to the direct-carbonization.^126–128 For instance, the surface coating of MOF crystals with PDA causes the outward-shrinkage during pyrolysis, thus effectively creating the hollow structure.^129,130 Jiang et al. reported the synthesis of hollow microporous carbon doped with Co, Fe, and N by applying an ultrathin coating of PDA on the surface of ZIF-8@ZIF-67 core-shell structure impregnated with Fe²⁺ [Fig. 17(b-i)].¹³¹ The three different concepts of nanoarchitecturing (core-shell MOF, bimetallic MOF, and surface coating of MOF) are successfully combined to obtain MOF-derived porous carbons with the hollow carbon structure doped with highly dispersed metal atoms (Co and Fe) and heteroatoms (N) upon the carbonization [Fig. 17(b-ii)]. Furthermore, alternative heteroatoms can be introduced by coating MOFs with polymers involving the desired elements in the molecular structure. For example, melamine molecules that have a high amount of N (i.e., an atomic C/N ratio of 1/2) were used to coat ZIF-8 to allow more N heteroatoms to be doped to the carbon matrix upon the carbonization.¹¹⁷ Apart from the higher level of N heteroatom-doping to the carbon matrix, the decomposition of melamine generates corrosive gas molecules such as NH_3, which effectively attack and break ligand-metal bonds in ZIF-8 to form the hollow structure in the resulting carbon. A more complex type of polymer known as poly(cyclotriphospazene-co-4,4′-sulfonyldiphenol) (PZS) was also applied on the surface of ZIF-8 along with Fe²⁺ ions [Fig. 17(c-i)].¹³² As PZS involves phosphorus (P) and sulfur (S) in its molecular structure, three types of heteroatoms (N from 2-meim, and S and P from PZS) are identified in the resulting ZIF-8/Fe@PZS [Fig. 17(c-ii)]. The highly cross-linked PZS layer on ZIF-8 causes the creation of hollow central cavity in the resulting ZIF-8/Fe@PZS-derived N, P, S co-doped hollow carbon (Fe-SAs/NPS-HC) obtained at 900 °C [Fig. 17(c-iii)]. To understand the evolutionary trajectory of hollow structure, the intermediates at 400, 500, and 600 °C were obtained and observed by TEM. At 400 °C, a partial surface etching leading to the concave morphology with rough and shrinking surfaces of ZIF-8/Fe@PZS was observed [Fig. 17(c-iv)]. The formation of hollow cavity becomes more evident from 500 °C, and the edges of ZIF-8 are gradually etched at higher temperatures [Figs. 17(c-v) and 17(c-vi)]. Based on the TEM observation of the intermediates, the mechanism for the formation of the hollow structure is proposed to follow the Kirkendall effect. Specifically, the smaller ionic radius of Zn²⁺ ion than S²⁻ causes the faster outward diffusion of Zn²⁺ than the inward diffusion of S²⁻, hence the unbalanced interdiffusion between the two ions at the interface of ZIF-8/Fe@PZS and the emergence of Kirkendall voids.

Departing from a single MOF nanoarchitecture, a more sophisticated nanoarchitecture can be done by utilizing the intermolecular interactions between the individual particles. For instance, self-assembly has been utilized to pack MOF particles into mesocrystals (i.e., crystal-of-crystals).¹³³ In the case of ZIF-8, different crystal morphologies lead to different packing structures. For instance, ZIF-8 usually adopts three main crystal morphologies, a cubic, a RD, or a truncated RD (TRD) crystal structure.¹³⁴ A simple way of controlling the crystal growth of ZIF-8 is to introduce cationic surfactants such as cetyltrimethylammonium bromide (CTAB),¹³⁵ cetyltrimethylammonium chloride (CTAC),¹³⁶ and trimethylstearylammonium chloride (STAC).¹³⁷ These surfactants tend to adhere to specific facets of ZIF-8 and act as growth inhibitors or capping agents to induce diverse nanoarchitecture.¹³⁸ For instance, CTAB had been used to selectively attach to the (100) facets to create a TRD nanostructure.¹³⁹ Taking advantage of this, different facets of the shapable ZIF-8 can be utilized to interact intermolecularly with another particle, hence creating a diverse well-ordered three-dimensional superstructure through self-assembly.

Using the TRD ZIF-8 particles synthesized through capping with CTAB, Avci et al. reported that these particles can be packed into a millimeter-sized 3D superstructure by slow evaporation under mild heating [Fig. 18(a-i)].¹⁴⁰ As a result, at 65 °C, these TRD ZIF-8 particles [Fig. 18(a-ii)] are arranged epitaxially into the dense rhombohedral packing (as in rhombohedral Bravais lattice), with lattice parameters of 188–239 nm upon increasing the particle size, and lattice angles of 61° regardless of the particle size [Fig. 18(a-iii)]. Furthermore, it showed a high packing fraction of 0.86. This packing was adopted to maximize the density of the TRD particles, as proven by Floppy–Box Monte Carlo simulations under increasing pressure. To describe the packing, one TRD particle has 12 adjacent particles, with 6 aligned through the hexagonal (110) facets, and the remaining 6 adhere to the neighboring square (100) facets. Manipulating the evaporation temperature influences the packing behavior and the overall packed morphologies of MOF crystals. At room temperature, the slower evaporation forms thick surrounding walls around the droplet and create a non-homogeneous monolith thickness. Inversely, evaporation at 100 °C causes quasi-amorphous packing, losing its highly defined crystallinity at 65 °C. Further study has been done on other MOFs such as UiO-66 and RD ZIF-8, and both show similar self-assembly properties but with a different packing morphologies. RD ZIF-8 crystallized into an fcc regular crystal, while UiO-66 adopted a hexagonal lattice.

Unlike the epitaxial packing of ZIF-8, Lyu et al. reported a low-dimensional assembly of ZIF-8 including 1D straight chains, alternating or bundled chains, 2D films of hexagonal, square, centered rectangular, and snowflake-like architectures, and quasi-3D supercrystals in a certain direction [Fig. 18(b-i)].¹⁴¹ In this case, they make use of depletion interaction, an entropic attractive interaction between two larger colloids usually formed from polymers, micelles, and nanoparticles to direct self-assembly between two ZIF-8 particles. To address the difficulties of the low surface charges on MOF particles, the authors added small-molecule ionic amphiphiles such as CTAC and sodium dodecyl sulfate (SDS) to direct the self-assembly using depletion interaction. CTAC was employed to first adhere to the surface of MOF to create a protective coating around the crystal, making it a micellar particle that can then exert a depletion force on other micellar particles. This then allows the self-assembly of the nanoparticles into a supra-framework structure. At the same time, micelles help to stabilize the colloids while maintaining the structure of the crystal. In the case of polyhedral ZIF-8, the self-assembly superstructures were observed differently in a hydroxyl-treated glass slide (smooth) and [poly(NIPAM-co-AAC)] treated glass slide (rough) surface. On a rough surface, RD ZIF-8 particles tended to aggregate into big chunks of colloidal crystal in which through depletion force, they arranged into an fcc facet via a full face-to-face overlap. This is because the rough surface was designed to prevent interactions with the nanoparticles, allowing them to interact among themselves. For smooth surface, a chain-like orientation is observed from the SEM image [Fig. 18(b-ii)]. In this case, the particles first interact with the hydroxyl-treated substrate by sticking its rhombic (110) facet to the colloid-behaving substrate via depletion interaction. This left the particles to only have two (100) facets, which behave like a sticking pad to interact with each other to form the preferred face-to-face binding. This leads to the formation of a chain-like orientation as confirmed in SEM image [Fig. 18(b-iii)]. This study has also been extended to other MOFs such as MIL-88A, UiO-66, and MIL-96, and even in 2D films.

Apart from that, ice templating can be used to nanoarchitecture MOF by acting as a sacrificial template, which can be eliminated after being subjected to vacuum sublimation.^142,143 Song et al. reported an ice-templated self-assembly to direct the growth of monolayers and bilayers of MOF nanoparticles without any external agent such as a surfactant.¹⁴⁴ In their study, TR ZIF-8 particles in water dispersion were quickly frozen in liquid N₂ prior to the self-assembly to a highly porous monolayer (TR-Z8-1) by a cryogenic process [Fig. 18(c-i)].¹⁴⁴ For the ice templating, TR ZIF-8 nanoparticles are held together by the weak van der Waals force to form TR-Z8-1. The simplest interaction comes from the perfect alignment of the two (100) facets of two particles with the other (100) and (110) facets left exposed as in SEM images [Figs. 18(c-ii) and 18(c-iii)]. Meanwhile, two ZIF-8 particles can also be aligned through hexagonal (110) facets. A mixed (100) and (110) facet interaction can also be observed in a single particle, with the angle between two connected facets being 54.8° [Fig. 18(c-iv)]. Despite certain defects, malfunction, and vacancies can be observed, the successful and stable face-to-face interaction leads to the formation of a monolayer when the concentration of ZIF-8 solution is low (<10 mg ml⁻¹). Upon increasing the particle concentration, the monolayers tend to aggregate into a bilayer. Direct-carbonization successfully converts the MOF nanoparticle monolayer into a hollow carbon monolayer (TR-Z8-1-C) [Fig. 18(c-v)]. However, as there is a rigid and ordered interface between the ZIF-8 nanoparticles, outward contraction occurs upon pyrolysis, leading to the formation of an inner cavity among the carbon monolayer [Fig. 18(c-vi)]. In the case of TR ZIF-8, the four hexagonal (110) facets cause an outward contraction forcing the formation of the inner cavity pore in the mesopore range as observed from the TEM image [Fig. 18(c-vii)].

In contrast to the bottom–up approaches that aim to modify MOFs in a constructive manner, top–down approaches advocate a destructive manner for the nanoarchitecturing of MOFs through various means including chemical etching and partial dissolution–recrystallization methods. As MOFs involve metal cations coordinated to organic linkers in their frameworks, alternative metal cations can be introduced to trigger the cation exchange within the framework of MOFs [Fig. 19(a-i)].¹⁴⁵ For instance, a regular hexahedral ZIF-8 (RH ZIF-8) synthesized with cetyltrimethylammonium bromide (CTAB) surfactant acting as a capping agent [Fig. 19(a-ii)]. To RH-ZIF-8, gold ions (Au³⁺) were added to allow cation exchange reactions during which non-equilibrium inter-diffusion occurs [Fig. 19(a-iii)]. The prolonged cation exchange between Zn²⁺ and Au³⁺, in turn, results in more rapid outward diffusion of the Zn²⁺ than inward diffusion of Au³⁺ across the shell [Fig. 19(a-iv)]. The disproportionate diffusion rate of Zn²⁺ and Au³⁺ causes the formation of Kirkendall voids and hollow nanocage (RH Au/Zn-MOF nanocage) [Fig. 19(a-v)]. After the completion of the cation exchange, both Zn and Au atoms are homogeneously distributed in the shell of hollow nanocage as demonstrated by STEM-EDS images [Fig. 19(a-vi)].

Other than the cation-exchange method, the chemical etching of specific planes can be selectively achieved by exploiting the three major factors: (1) the varied densities of metal-organic linker linkages between the crystal facets of MOFs and (2) acid-base reaction between the organic linkers of MOFs and weak organic acids; and (3) chelating behavior of weak organic acids to coordinate with metal ions. Hu et al. demonstrated the etching of various MOFs, including ZIF-8, ZIF-67, and MIL-68, to form hollow nanoparticles or microrods and yolk-shell nanoparticles.¹⁴⁶ Specifically, they utilized phenolic or polyphenolic acids [i.e., gallic acid (GA) or tannic acid (TA)] to induce surface functionalization and selective etching of the target MOFs. Since both GA and TA are rich in phenol groups, they can form a thin coating on the surface of MOFs, causing the change in zeta potential of MOFs before and after the addition of TA or GA. Moreover, clear changes in pH were also observed from the initial pH of TA at ∼3.5 increasing to 6 upon mixing with ZIF-8 followed by the subsequent gradual increase to 8 after 5 min. The trend of change in pH values indicates that the liberated protons from TA are consumed by ZIF-8 to create the hollow structure. Similarly, Avci et al. demonstrated the use of xylenol orange (XO) to induce even more elaborate and controllable etching of ZIF-8.¹⁴⁷ As XO acts as a weak acid, it leads to the protonation of imidazolate linker of ZIF-8. Subsequently, the protonation of imidazolate linker results in the break of coordination bonds and partial etching of ZIF-8 to liberate Zn²⁺. Interestingly, the etching of ZIF-8 in the presence of XO occurred along the selective facets of ZIF-8. They proposed that selective etching occurs along the preferred facets of ZIF-8 where metal-linker bonds are present and exposed. Typically, (211) facets of ZIF-8 are highly enriched with metal-ligand bonds, therefore exposing more sites for etching while (111) facets do not contain any of these linkages [Fig. 19(b-i)].^147,148 Therefore, the preferred etching along the (211) facets of ZIF-8 results in the formation of an open-mouthed nanoarchitecture with abundant micropores and clear macropores (M-ZIF-8) [Figs. 19(b-ii) and 19(b-iii)]. The same nanoarchitecture was reproduced and pyrolyzed to obtain the corresponding N-doped micro-/macroporous carbon materials (mM-NC) [Figs. 19(b-iv) and 19(b-v)].¹⁴⁸ 

Apart from the etching of MOFs by weak organic acids, Lewis acid can also be implemented to induce a selective etching as demonstrated by Zhang et al. [Fig. 19(c-i)].¹⁴⁹ In their study, ZIF-67 nanoplate (NP) was first synthesized with the help of CTAB and subsequently subjected it to the acid etching reaction by introducing the guest metal salt as a Lewis acid [Fig. 19(c-ii)]. As FeCl₂·4H₂O undergoes hydrolysis to create the acidic environment when it is added in to ethanol, they induced etching of ZIF-67 NP in FeCl₂ ethanol solution (Lewis acid) and obtained hollow Fe-ZIF-67 NP with unique cross-channels [Fig. 19(c-iii)]. The resulting Fe-ZIF-67 NP was transformed to Fe-Co₃O₄ nanosheets with highly dispersed Fe and Co atoms by solvothermal treatment [Figs. 19(c-iv) and 19(c-v)]. The structural evolution of ZIF-67 NP to the hollow Fe-ZIF-67 NP was investigated in FeCl₂ ethanol solutions of varied concentrations. Based on the TEM observations, the etching process begins selectively at the center of four sides in 0.4 mg ml⁻¹ FeCl₂ ethanol solution, and further etching occurs to form the obvious cross-channels passing through the center of the nanoplate at higher concentrations [Fig. 19(c-vi)]. The method of selective etching of ZIF-67 NP to hollow NP by Lewis acid was described as etching-coordination approach because the guest metal salt (FeCl₂) not only etches certain areas of MOFs and the metal source gets incorporated into the etched MOF by forming coordination bonds.

In parallel with the chemical etching methods to modify the nanoarchitectures of MOFs and their carbons, alternative synthetic methods based on the dissolution–recrystallization behaviors of MOFs have also been investigated to synthesize anisotropic nanostructure. This approach attempts to manipulate the slow kinetics of MOF crystallization and produce MOF single crystals by inducing amorphous or crystalline MOFs to dissolution and recrystallization.^150–154 Zou et al. reported an amorphous MOF-mediated recrystallization approach by which they synthesized single-crystal Co-MOF nanotubes from amorphous nanoparticles [Fig. 20(a-i)].¹⁵⁵ Unlike direct crystallization from metal salts, which gives rise to rod-like structures in micrometer scale (3–30 μm in length and 1–5 μm in diameter), this approach first synthesized the amorphous MOF-74 nanoparticles from metal salts [Fig. 20(a-ii)] and allowed it to recrystallize under hydrothermal treatment to form a long single-crystal MOF nanotube of size 66.8 ± 13.8 nm in diameter [Fig. 20(a-iii)]. Initially, amorphous Co-MOF-74 nanoparticles were dissolved slightly in water and the dissolved organic ligands and metal ions then crystallized slowly to form a single crystal due to the limited amount of precursors. Further carbonization at 700 °C retains the 1D structure of the CNT embedded with the Co particles to yield NCo@CNT-NF. Interestingly, adding dicyandiamide (DCDA) during carbonization forms an N-doped 3D hierarchical dendrite (carbon nanofiber trunks and carbon nanotube branches) growing from the surface of the CNT if the pyrolysis temperature is higher than 700 °C [Fig. 20(a-iv)]. Also, TEM revealed that Co nanoparticles were trapped inside the nanotubes [Fig. 20(a-v)].

In the other case, the dissolution–recrystallization strategy can be used to induce defects to increase the surface area of carbon. For instance, Bi et al. reported a metal dopant-free approach to create defect-rich carbon nanorods (d-CNRs) through dissolution–recrystallization strategy.¹⁵⁶ As in the above-discussed work, the Zn-MOF-74 nanoparticles undergo dissolution–recrystallization to form Zn-MOF-74 nanorods, before being carbonized into carbon nanorods [Fig. 20(b-i)].¹⁵⁶ They first synthesized MOF nanorods through the hydrothermal reaction of Zn acetate and 2,5-dihydroxyterephthalic acid (DHTA) in methanol to form Zn-MOF-74 nanoparticles [Fig. 20(b-ii)]. Meanwhile, dissolving Zn-MOF-74 nanoparticles in water slowly supplies ligands and metal ions for recrystallization to Zn-MOF-74-NR with an average diameter of 200 nm (±50 nm) [Fig. 20(b-iii)]. After recrystallization, these crystals usually have a better homogeneity and smooth surface. To study the defect-rich property, temperature-dependent pyrolysis was investigated at 400, 500, 800, 900, 1100, and 1200 °C. Upon increasing the temperature, d-CNRs tend to shrink but maintain their 1D rod-like structure up until 1200 °C at which they collapse into pieces. Pyrolysis at 1100 °C effectively yields defect-rich rods in the radial direction as observed from the SEM and TEM images [Figs. 20(b-iv) and 20(b-v)]. As the pyrolysis of Zn-MOF-74 NR obtained by direct crystallization does not lead to the formation of significant defects, the defect-rich property of d-CNR is attributable to the uneven concentration distribution of ligands and Zn precursors during the slow recrystallization process, which then leads to uneven pyrolysis and cracks. The presence of these cracks then results in a high specific surface area of 2459 m²/g with a high mesopore ratio.

Templated-based methods usually involve the direct growth of MOF crystals on the surface of templates.^157–165 Typically, hard templates are well-defined solid-state materials generally having their pores and surfaces infiltrated with metal or polymer precursors for the crystallization or polymerization, respectively.¹⁶⁶ Examples of the hard-template include metals and metal oxides,^167,168 polymers,¹⁶⁹ carbon allotropes such as graphene^170–172 and carbon nanotube,^173,174 and they need to be removed by harsh/toxic chemicals (strong acid or base) in the last step of the synthesis, therefore, making the overall synthetic method more tedious although the well-defined pores are achievable.^175,176

Hong et al. reported an interesting nanoarchitectured crystalline macroporous ZIF-67-derived carbon through controllable hard templating on a 3D-ordered superpolymer spheres template made of polystyrene (PS) [Fig. 21(a-i)].¹⁷⁷ In their study, single crystalline ordered macro−microporous ZIF-67 (SOM ZIF-67) was first obtained by impregnation of self-assembled monolithic PS sphere (diameter = 190 nm) template voids with Co²⁺ precursors and 2-methyimidazole in a mixed ammonia–methanol co-solvent. This novel co-solvent strategy was adopted because ammonia increases the nucleation speed of the precursors, while methanol stabilizes Co²⁺ precursors and 2-meim to adjust the balance between nucleation and growth of the crystal to achieve a homogeneous nucleation.¹⁷⁸ The as-prepared SOM ZIF-67 was then subjected to a single-step selenization–carbonization process to obtain 3D-ordered macroporous cobalt diselenide@carbon (3DOM CoSe₂@C) that largely retains the original tetrahedral macropore-arrangement of SOM ZIF-67 [Fig. 21(a-ii)]. However, from the magnified SEM images, the size of macropore in 3DOM CoSe₂@C decreased to ∼140 nm due to structural shrinkage of the overall porous structure during carbonization [Fig. 21(a-iii)]. After the crystal growth of single crystalline ZIF-67, the hard template of PS was removed in dimethylformamide (DMF) and mixed with Se powder during carbonization to dope Se into the carbon network, forming 3DOM CoSe₂@C [Fig. 21(a-iv)]. The resultant carbon has highly ordered interconnected macropores as identified in TEM images alongside its well-defined tetrakaidecahedron structure inherited from its SOM ZIF-67 precursor [Fig. 21(a-v)]. The highly interconnected macroporous structure allows large channels to facilitate a significantly improved diffusion of macromolecules throughout the large surface area. From the HRTEM and elemental mapping images, the homogeneous distribution of CoSe₂ nanoparticles [diameter smaller than 5 nm and lattice fringes of 0.26 nm for (210) plane] is observed across the carbon matrix [Figs. 21(a-vi) and 21(a-vii)].

In another instance of the hard-template method, Pan et al. reported the crystallization of ZIF-8 in the hard-template of polymer cubosomes (PCs) [Fig. 21(b-i)].¹⁷⁹ In their study, PC was specifically formed by the self-assembly of PS₂₂₄-b-PEO₄₅ copolymer as the hydrophobic polystyrene block forms the frame of PC, while the hydrophilic PEO block constructs the surrounding coronae. When p value reaches approximately 1, bicontinuous structured PCs with the double primitive (DP) morphology and mesopores (∼25 nm) in a square lattice configuration are generated [Fig. 21(b-ii)]. Upon mixing with 2-meim, the continuous open channels of PCs effectively allow impregnation of the organic linker through intermolecular noncovalent interactions such as hydrogen bonding and electrostatic interactions. The subsequent addition of Zn²⁺ precursor gives rise to the formation of ZIF-8 crystal in the mesochannels of PC, forming a ZIF-8 microsphere (ZIF-8@PC) [Fig. 21(b-iii)]. Carbonization effectively converts ZIF-8@PC to mesoporous carbon cubosomes with a single primitive (SP) topology (SP-NCs), achieving a high specific surface area of 825 m² g⁻¹ [Fig. 21(b-iv)].

In contrast to hard-templating methods, soft-template approaches involve the use of surfactants and their self-assemblies to soft-templates, known as micelles. Specific types of micelles and precursors of certain MOFs are capable of co-assembly mostly by weak noncovalent bonds such as hydrogen bonding,¹⁸⁰ hydrophobic–hydrophobic interactions,¹⁸¹ and ionic bonds.¹⁸² Therefore, the added MOF precursors can accumulate around the micelles and then undergo crystallization to eventually form mesoporous MOFs. One of the most significant advantages of the soft template approach is the tunability of pore size on the materials.^183–185 The most commonly reported method is to introduce hydrophobic swelling agents such as trimethylbenzene (TMB)^186,187 and triisopropylbenzene (TIPB), which interact with the hydrophobic part of the micelle to increase the diameter. For instance, Yamada et al. reported that the addition of TIPB induces the micellization of CTAB to a larger size than that of TMB for the synthesis of silica nanoparticles with similar morphology and nanoarchitecture.¹⁸⁸ 

Among recent studies, Li et al. reported the usage of Pluronic type triblock copolymer (PEO₂₀–PPO₇₀–PEO₂₀, P123) as the source of soft-templates to induce large-pore with size larger than 10 nm on Ce-based UiO-66 (Ce-UiO-66) [Fig. 22(a-i)].¹⁸⁹ In their study, Hofmeister salting-in ion of ClO₄⁻ was used to mediate the MOF growing moiety (terephthalic acid and Ce precursor) to the micelle through solubilization in a mild condition. Despite the weak interaction between the surfactants and growing MOFs and the associated expulsion of growing MOFs from the surface of micelles, this study successfully demonstrated the synthesis of mesoporous structure and well-crystallized microporous pore walls in Ce-UiO-66 [Figs. 22(a-ii)–22(a-iv)]. As it is common that amphiphilic micelles have different self-assembly behaviors at different temperatures, the increased temperature at 35 °C causes P123 to form cylindrical micelles, resulting in a hexagonal arrangement of uniform mesopores in Ce-UiO-66. Further raising the micellization temperature to 45 °C generates the multilamellar vesicle (MLV) structure, leading to the increased pore size of 18.8 nm as P123 tends to achieve larger packing parameters for the formation of lamellar micelles. Alternatively, a face-centered cubic structure was also obtained with F127 that has a higher PEO/PPO ratio than P123.

Other than single-micelle systems, recent studies have implemented secondary soft-templates to achieve dual-micelle systems and extended the freedom of control over more complex MOF nanoarchitectures. For instance, Xu et al. reported a hollow triple-shell mesoporous UiO-66-NH₂ particles (3S-mesoUiO-66-NH₂ HoMSs) with dual-templates derived from the self-assembly of F127 and octadecyl dimethyl betaine (ODMB) [Fig. 22(b-i)].¹⁹⁰ In their study, an inhomogeneous UiO-66-NH₂ particle was synthesized with a less stable mesopore derived from ODMB micelle and a more stable mesoporous shell derived from F127. At the beginning of the reaction, the ionic surfactant of ODMB and F127 formed a co-assembly with MOF precursors, which then crystallize to form a core-shell nanoarchitecture with ODMB-directed MOF core in the center and F127 assembly directed MOF wall growing around the center. This occurs because they form different interactions with the metal precursors. Specifically, carboxyl-terminus of ODMB molecules have a stronger interaction with Zr-oxo clusters through coordination bonds allowing them to form the disordered worm-like MOF core. On the contrary, F127 mainly interacts with Zr-oxo clusters electrostatically at its hydrophilic blocks, causing a delayed formation of MOF wall with radially oriented mesochannels growing epitaxially on the ODMB-directed core. Etching with acetic acid preferentially eliminates the defect-rich ODMB-directed MOF, resulting in a highly crystalline and accessible mesochannels in 3S-mesoUiO-66-NH₂ HoMSs. From the SEM image, a round-edge polyhedral shape with an average diameter of ∼400 nm is confirmed [Fig. 22(b-ii)] and abundant mesopores are observed in high magnification image [Fig. 22(b-iii)]. From the STEM images, Zr, C, and N are evenly distributed in 3S-mesoUiO-66-NH₂ HoMSs [Fig. 22(b-iv)]. Also, the TEM image demonstrates a multi-layer hollow mesostructured MOF [Figs. 22(b-v) and 22(b-vi)]. The pore size of 3S-mesoUiO-66-NH₂ HoMSs is observed to be 7 nm [Fig. 22(b-vii)].

Alternatively, Li et al. reported the dual-template system consisting of P123 and F127 to achieve the controllable pore size of mesopores from large mesopore (∼24 nm), small mesopore (∼7 nm), and no mesopore (i.e., microporous) in UiO-66-NH₂ [Fig. 22(c-i)].¹⁹¹ The subsequent direct-carbonization at the annealing temperatures of 800 °C effectively gives rise to mesoporous UiO-66-NH₂-derived carbon. Interestingly, the mesopores of different sizes induce varied surface curvatures of the carbons. From the SEM and TEM images, the surface pore diameters of ∼25 nm and ∼5 nm are observed from mesoporous carbons derived from mesoporous UiO-66-NH₂ (L/NC and S/NC, respectively) [Figs. 22(c-ii) and 22(c-iii)]. The pore size distribution for different pore sizes is also supported by N₂ adsorption/desorption isotherms [Fig. 22(c-iv)]. In this study, they proposed that small mesopores induce a high curvature while large mesopores induce a low curvature, both on the concave surface while the convex surface corresponds to the carbon nanoparticles with no mesopore. Also, the curvature activity simulated by Fe-N₄ was predicted with first principles calculation embedded in graphene by calculating their formation energy (E_f). The results show that the Fe-N₄ embedded in large pores (small concave curvature, lc-Fe-N₄) are more stable and thermodynamically easier to form (E_f = 0.37 eV). The calculation also indicates that the surface curvature potentially has a significant impact on charge density re-distribution of active metal sites and can manipulate the charge of the metal center (Fe^δ+). This concludes that nanoarchitecturing on the pore using micelle-directed soft templating has a significant influence on the catalytic activity.

Alternatively, MOF crystals can be adopted as the sacrificial hard-template to obtain highly unique and tunable nanoarchitectured porous carbons. For instance, inorganic materials such as SiO₂ and graphene oxide (GO) were used to coat the surface of MOF crystals to induce novel nanoarchitectures upon carbonization.^{176,192–195} Specifically, the surface charge difference between GO and ZIF-8 was exploited to induce the wrapping of ZIF-8 with GO.¹⁹⁶ The hybridized structure of GO-wrapped ZIF-8 was subsequently carbonized to obtain a unique 3D porous carbon framework (PCF) with clear hollow cavities and abundant micropores [Fig. 23(a-i)].¹⁹⁶ The resulting PCF is highly interconnected graphene nanosheets with the macropores homogenously dispersed in the interlayer space as confirmed by SEM [Fig. 23(a-ii)] and STEM images [Fig. 23(a-iii)]. In addition, the carbonization generates reduced GO (rGO) in PCF, achieving a higher level of graphitization as compared to the pristine ZIF-8-C. This is further verified by the lower I_D/I_G value obtained from the Raman spectrum of PCF as compared to that of ZIF-8-C [Fig. 23(a-iv)]. Wan et al. successfully applied mesoporous SiO₂ (mSiO₂) coating on the surface of ZIF-8 prior to the carbonization.¹⁹⁷ With mSiO₂ coating, the edge frame of the dodecahedron could be maintained while the planar faces collapsed. Such difference in the rate of decomposition between the edge and planar faces caused the anisotropic thermal shrinkage of ZIF-8, hence creating the concave morphology with enlarged micropores and mesopores in the carbon matrix. After the carbonization, the mSiO₂ coating was etched off. Alternatively, the mSiO₂ coating was also applied to bimetallic Zn–Co ZIF to induce both structural control and in situ metal doping with Co atoms [Fig. 23(b-i)].¹⁷⁶ The Zn–Co ZIF adopts a dodecahedral shape with an average size of 120 nm [Fig. 23(b-ii)]. Even after the coating, individual ZIF particles with the layer of mSiO₂ (∼20 nm) remain highly dispersed as shown in the TEM image [Fig. 23(b-iii)]. After carbonization at varying temperatures from 700 to 1000 °C, the inner ZIF particles become highly porous carbon with nanoporous system consisting of micro-, meso-, and macropores, while the mSiO₂ shell remains intact [Fig. 23(b-iv)]. The removal of mSiO₂ shell by HF eventually gives rise to Co,N co-doped porous carbon nanoframework (Co, N-CNF) [Fig. 23(b-v)].

Furthermore, soft-templates can be introduced to create ordered mesostructured polymeric coating on MOFs. Kim et al. successfully applied mesostructured PDA coating on ZIF-8 by implementing a specific soft-template consisting of Pluronic F127 and 1,3,5-trimethylbenzene (TMB). They also traced the effect of F127 and TMB on forming the mesostructured PDA coating.^185,198 With F127 alone without TMB, spherical PDA nanoparticles with the size less than 50 nm are formed, which then coated the surface of ZIF-8 to form spherical PDA coated ZIF-8 (ZIF-8@sPDA) [Figs. 23(c-i) and 23(c-ii)]. The spherical PDA nanoparticles are formed as F127 acts as the size-limiting agent without TMB.¹⁹⁸ Further pyrolysis leads to the formation of a hollow sphere coated carbon (HSC) as in the TEM image [Fig. 23(c-iii)]. With TMB that interacts with hydrophobic polypropylene oxide (PPO) chains of F127 to expand and stabilize the F127/TMB micelles, the mesostructured PDA coating was smoothly coated on ZIF-8 to form mesostructured PDA coated ZIF-8 (ZIF-8@mPDA) [Figs. 23(c-iv) and 23(c-v)].^185,198 The pyrolysis of ZIF-8@mPDA, in turn, produced different hollow nanoarchitectures in the resulting carbon materials to form hierarchically porous double-shelled carbon (HPDC) [Fig. 23(c-vi)]. Particularly, HPDC possesses clear mesopores in the carbon shell and highly microporous carbon yolk and hollow cavity between the shell and yolk, therefore, demonstrating a wide range of pore size distribution encompassing micropore, mesopore, and macropore [Fig. 23(c-vi)].

In conclusion, this review highlights the importance of nanoarchitecturing for MOF-derived porous carbons to come up with the future carbon materials that selectively achieve the target advantages without the accompanying disadvantages in materials (e.g., crystallinity vs porosity, surface area vs diffusion efficiency, activity vs selectivity, kinetics vs stability, etc.). Prior to the discussion on nanoarchitectured MOF-derived porous carbons, previous studies on the evolution of carbon microstructure in the temperature-controlled pyrolysis process are introduced to allow better understanding of the carbon materials. It was subsequently followed by the rationale for the necessity of nanoarchitecturing in porous carbon materials from the perspective of electrochemical interaction of carbon materials, specifically with EDL formation. While developing the rationale for nanoarchitecturing, the role of each class of nanopores (micro-, meso-, and macropores) is also summarized and presented to offer guidance to future research related to porous carbon materials. In light of synthetic methods and approaches to nanoarchitecture porous carbon materials, two main methodological categories of template-free and template-based methods are briefly outlined with relevant example materials resulting from each method.

Upon the buildup of the knowledge and understanding of the “microstructural evolution of carbon materials,” “rationale for nanoarchitecturing,” and “synthetic methods for nanoarchitecturing of porous carbons,” various nanoarchitectured MOF-derived porous carbons are presented and discussed based on template free (bottom–up and top–down approaches) and template-based (hard- and soft-templated approaches) methods. This review therefore offers the strategies to achieve potential “future carbon materials” through effective nanoarchitecturing by visualizing the atomic, nanoscopic, microscopic, and macroscopic levels of nanoarchitecturing for MOF-derived porous carbons (Fig. 24).

To this end, we offer some suggestions and directions for the research on future carbons from the viewpoint of current challenges in nanoarchitecturing of MOF-derived porous carbons.

1. Nanoarchitectured high-entropy MOF-derived carbon

   Recently, high-entropy materials have been actively explored for a variety of compounds such as oxides, metals, glasses, and carbides in different morphologies. High-entropy materials typically consist of five or more elements in a stable single-phase crystal or glass-like structure with the increase of configurational entropy, which is known to achieve the highest value when all elements are made of equimolar fractions. With unique structural characteristics, adjustable compositions, and excellent physicochemical properties, high-entropy MOFs have gained a significant research interest though they remain thus far the least explored materials of high-entropy compositions. Therefore, it is not surprising to mention that nanoarchitecturing of high-entropy MOF-derived porous carbons is even less investigated, requiring active research to successfully introduce novel high-entropy metal compositions in different forms (e.g., single atoms, nanoclusters, nanocrystals, etc.) to nanoarchitectured MOF-derived porous carbons.. We propose that this can be achieved by following chemistry: (1) Coordination chemistry of high-entropy metal compositions in nanoarchitectured MOFs and direct-carbonization; (2) host-guest chemistry of high-entropy metal compositions in nanoarchitectured MOFs and direct-carbonization; (3) incipient wetness impregnation or solvent-free planetary ball-milling of high-entropy metal compositions to nanoarchitectured MOFs and direct-carbonization; and (4) direct loading of high-entropy metallic crystals or glass to nanoarchitectured MOF-derived carbon.

2. Nanoarchitectured anisotropic mesoporous MOF-derived carbon

   Reports on mesoporous MOF crystals synthesized using soft-templates have been increasing, but further studies on the tuning of nanoarchitectures and direct-carbonization of various mesoporous MOFs are significantly lagging behind. This is especially the case for the synthesis of anisotropic (e.g., 1D or 2D) mesoporous MOFs because it requires tailored crystal engineering (e.g., using capping agent to restrict the crystal growth from specific facets, using alternative organic linkers to induce anisotropic crystal growth, using different solvents to allow modified crystal structures, etc.) while ensuring the formation of soft-templates that MOF crystals favorably incorporate into their crystal structures. Since anisotropic materials are highly attractive with unique physical, mechanical, optical, and electronic properties, there is a substantial room for the research interest to synthesize anisotropic mesoporous MOFs and their carbonaceous forms via direct-carbonization.

3. Nanoarchitectured MOF mesocrystal-derived carbon

   Mesocrystals (also known as crystal-of-crystals) are metastable intermediates formed as single crystals self-assemble in a periodic arrangement. To successfully induce the formation of MOF mesocrystals via self-assembly, highly uniformity in exposed facet, morphology, and size of single MOF crystals should be ensured. Since mesocrystals often exhibit coupling and amplification of properties of single crystals as well as collective and emergent properties arising from the anisotropic arrangement of single crystals, further nanoarchitecturing of MOF mesocrystals and their carbons can achieve novel or even further enhanced properties.

This work was supported by the JST-ERATO Yamauchi Materials Space-Tectonics Project (JPMJER2003), the ARC Laureate Fellowship (FL230100095), and the UQ-Yonsei International Research Project. This work used the Queensland node of the NCRIS-enabled Australian National Fabrication Facility (ANFF). Y. Bando and S. M. Alshehri appreciate the Distinguished Scientist Fellowship Program (DSFP) at King Saud University, Kingdom of Saudi Arabia for the financial support.

The authors have no conflicts to disclose.

Minjun Kim: Conceptualization (lead); Investigation (lead); Methodology (lead); Project administration (lead); Visualization (lead); Writing – original draft (lead); Writing – review & editing (lead). Kwang Keat Leong: Writing – original draft (supporting); Writing – review & editing (supporting). Nasim Amiralian: Writing – review & editing (supporting). Yoshio Bando: Writing – review & editing (supporting). Tansir Ahamad: Writing – review & editing (supporting). Saad M. Alshehri: Writing – review & editing (supporting). Yusuke Yamauchi: Conceptualization (lead); Funding acquisition (lead); Writing – review & editing (lead).

Data sharing is not applicable to this article as no new data were created or analyzed in this study.