Covalent Organic Frameworks for Energy Conversions: Current Status, Challenges, and Perspectives

Introduction

Energy conversion is driven by molecular systems that enable the input of photons and electrons/holes and the output of different energy forms.¹ The most sustainable way is to use natural resources, including carbon dioxide, oxygen, and nitrogen in the atmosphere, as well as light and water, as the input to produce electricity, fuels, and value-added chemicals. Studies on photoenergy conversion to yield electricity have established a diversity of different photovoltaic systems; some have been widely used in power plants.^2–4 In contrast, the utilization of light energy to produce chemical fuels and value-added compounds, represented by water splitting into hydrogen and oxygen, and carbon dioxide reduction into carbon monoxide and other derivatives, remains challenging goals.^5–7 On the other hand, the usage of electricity to reduce carbon dioxide is remarkable with the production of value-added chemicals, while the reduction of oxygen on an electrode is a critical reaction in fuel cells relative to green energy generation. These transformations require a systematic design of catalytic systems that would indeed promote green energy transformations, encompassing efficiency, eco-friendly, and sustainability.

The utilization of solar energy to split water into hydrogen and oxygen is an ideal plan for the production of green energy, as sunlight and water are both abundant natural resources, and hydrogen is a green chemical fuel.⁵ However, simple irradiation of water with sunlight could not split water into hydrogen and oxygen, as this reaction is accompanied by a large entropic thermodynamic penalty. Since the oil crisis in the 1970s, extensive and intensive studies have been conducted to develop catalysts that ease and promote water splitting driven by light energy.⁵ Nevertheless, a suitable catalyst that promotes this reaction efficiently is still inaccessible so far. Meanwhile, the photoreduction of carbon dioxide requires suitable catalysts so that the generation of electrons from excitons and the flow of electrons to the reduction center could be interlocked to promote the reaction. However, it is extremely difficult to merge photochemical events and catalytic cycles into one catalytic material. Distinct from photocatalysis, the electrocatalytic reduction of carbon dioxide and oxygen requires different structural criteria for designing catalysts, as the reaction occurs on an electrode and is driven by an electric field.⁶ In these cases, the injection of an electron from an electrode to the reduction center and the accumulation of carbon dioxide and oxygen on the electrode become key processes to control the reaction dynamics, and how the catalytic site promotes the reduction reaction determines the kinetics. In terms of resource availability, developing photocatalysts and electrocatalysts based on organic materials other than the traditional metal-based systems are essential for achieving sustainable transformations and energy conversions.

Covalent organic frameworks (COFs) are a unique class of polymers as they integrate organic molecules into a predictable long-range-ordered structure.^8–15 This pushes the boundaries of organic/polymer chemistry and materials science to achieve the unprecedented deterministic control of long-range structures and properties of synthetic macromolecules. COFs are formed via the polymerization of organic molecules (monomers) in which the chain propagation is controlled by a topology diagram so that each bond connection is spatially confined at either a two- or three-dimensional (2D or 3D) way or geometrically guided to produce extended, yet ordered 2D or 3D polymer architecture. The geometry and backbone of organic units predetermine the dimensionality, along with the shape and size of polygons in COFs (Figure 1). During polymerization, the 2D network crystallizes to form 2D layer frameworks driven by interlayer π–π stacking interactions; thus 2D COFs offer ordered columnar π-arrays, as well as aligned one-dimensional (1D) nanochannels (Figure 1). Therefore, the ordering structure of COFs is twofold: one is the periodic skeleton, and the other is the aligned channels. Owing to these two distinct features, COFs offer an irreplaceable platform for designing polymers to achieve a predesignable skeleton and pore. Undoubtedly, the above diverse structural features of COFs are inaccessible with traditional polymers and other porous materials, while the predesignablity of skeleton and channel opens a way to design not only structures but functions as well.

• Download figure
• Download PowerPoint

By virtue of a broad diversity of building blocks, COFs have been explored using different design strategies to develop various energy conversion systems. In this minireview, we focus on exploring COFs for diversified energy conversions, including light-driven water splitting, carbon dioxide reduction, and the electrocatalytic reduction of oxygen and carbon dioxide. We summarize the design principles of COFs for each type of conversion and outline their synthetic approaches and fundamental properties to show the structural features of these catalytic systems. We scrutinize each energy conversion system by correlating structure with function to disclose the essence, mechanistic insights, and drawbacks, intended to examine the potential of each strategy and approach. Based on these analyses, we predict key fundamental issues to be challenged, envisioned future directions worthy of the studies, and clarify perspectives from chemistry, physics, and materials science to uncover a full picture of energy conversions.

Design Strategies and Structural Features

COFs explore topology diagram as a principle to design skeletons and pores,¹⁰ and the geometric combination of monomers, that is, knot and linker determines both the primary- and high-order structures. This predesignability is twofold (Figure 2): One is constructing the polymer backbone; its dimensionality and spatial orientation pattern are predetermined. The other is the presetting the pore size, shape, and wall interface. At the primary-order structural level, the backbone offers the 2D polymer, which is an atomic sheet with an alternately connected knot and linker over the 2D plane, while the topology pattern is predetermined by the geometry of monomers. The pores in the 2D sheet have discrete size and shape and consist of specific units extruded from the knot and linker, while these parameters are predesigned by monomers. Therefore, the primary-order structure is predetermined by a polymerization reaction in which knot and linker units are covalently connected along the x and y direction to form an ordered 2D polymer. At the high-order structure level, the 2D polymer sheet stacks to create layer frameworks, achieved via crystallization, and the high-order structure is predetermined by the noncovalent interactions between the 2D layers and controlled by the total free energy for crystallization.

• Download figure
• Download PowerPoint

The layer framework configuration constitutes π-columns at the knot and linker sites, with the π-columns linked topologically to form continuous π-arrays across the material (Figure 2a). By virtue of layer stacks, the framework creates extended 1D channels, which are independent of each other, and accessible from the top and bottom layers (Figure 2b). These channels possess discrete size and shape, while the pore walls are covered with the various C–H units extruded from the knot and linker units, B, O, or N atoms from linkages, and other substituents on the knot and linker. These atoms and units are sequenced continuously along the z-direction on the channel walls and form various wall interfaces that control channels function (Figure 2b). More interestingly, the channel walls could be predesigned to install multiple functional groups such as acid, base, catalytic site, radical, hydrophobic, and hydrophilic units via pore surface engineering.^16,17

Organic units with C₁, C₂, C₃, C₄, and C₆ symmetries that bear different reactive sites (Figure 3a) have been explored as monomers for preparing 2D COFs. Figures 3b–3d summarize the major geometry combinations of monomers for designing 2D COFs to achieve different backbones and channels. Except for self-condensation reactions, COFs have been designed with the [one knot + one linker] combination, with varying geometry combinations (Figure 3b). In these cases, the polymer architecture features isotropic tiling and regular polygonal pores across the framework.

• Download figure
• Download PowerPoint

We have explored a multicomponent strategy for designing COFs to achieve anisotropic lattice tiling and irregular polygonal shapes (Figure 3c).¹⁸ This strategy enables the [one knot + two linkers], [one knot + three linkers], or [two knots + one linker] combinations, and is applied to the hexagonal and tetragonal topologies. The multicomponent strategy offers 2D lattices with anisotropic tiling and unusual pore shapes, which are inaccessible to the [one knot + one linker] strategy. Interestingly, the aligned multicomponent in the lattice triggers unique electronic interactions to exhibit distinct properties and functions, while the special pore shape allows the creation of unique nanospace for confinement and molecular separation.¹⁸ The multicomponent strategy enhances the structural complexity considerably due to the anisotropic tiling and increases the diversity of COF members owing to the introduction of more patterns for varying combinations of components.¹⁸

The double-stage strategy develops the possibility of using two different linkages to synthesize COFs, based on one C₁-symmetric unit that possesses two types of reactive sites (Figure 3d).^19,20 We have established this strategy for designing COFs to show the [two C₃ knots + one C₁ linker] and [one C₃ knot + one C₁ linker] schemes used for hexagonal COFs, the [two C₄ knots + one C₁ linker] combination for tetragonal COFs, and the [two C₂ knots + one C₁ linker] diagram for the rhombic COFs.¹⁹ Noticeably, these geometrical combinations improve the structural complexity and diversity of COFs substantially.

The topology diagram guides the growth of polymer chains into covalently linked 2D polymers, which crystallize to form crystalline porous frameworks via a one-pot reaction that enables both polymerization and crystallization (Figure 1). As monomers consist of rigid π-backbones, the above topology diagrams allow the design of different π-architectures, which are otherwise inaccessible with a supramolecular assembly of 1D polymers and single crystals of small organic compounds. The layered framework offers extended 2D topology in which the π-units stack face-to-face to maximize interactions, thereby casting a sharp contrast to single crystals, which tend to form herringbone alignment.

COFs have been designed to achieve various structures with typical topologies, as shown in Figure 4a. To design a functional COF, a basic concept is to develop an interface based on the COF structure (Figure 4b). The key fundamental issue is to disclose and understand how COFs interplay with photons, excitons, phonons, electrons, holes, ions, and molecules, as these interactions determine the property and function of COFs. How to trigger and develop unique interplays with COFs are central subjects that pave the way to explore COFs as a class of novel materials, unique in structural and functional properties that could not be replaced with other materials.

• Download figure
• Download PowerPoint

Based on the structural features of COFs, we have developed the functions of COFs from three different approaches (Figure 4c). Based on ordered π-skeleton, COFs have been developed as semiconductors, light emitters, light-harvesting antennae, energy transfer, electron transfer, charge separation, photovoltaics, spin alignment, and topology insulators. By developing 1D channels with specific interfaces, COFs have been explored for adsorption, storage, confinement, recognition, and separation. In many cases, the function is triggered by the complementary effects of both skeleton and channel, such as catalysis, sensing, mass transport, energy storage, energy conversion, and biorelated functions.

In this minireview, we focus on scrutinizing the COF systems for photocatalysis and electrocatalysis, related to energy conversion (Figure 4c, keywords in red). These catalytic properties originate from the structural feature of COFs; the π-frameworks consist of periodically aligned columnar π-arrays, while their spatial patterns are distinct from each other based on the topology. The π-arrays serve as light-harvesting antennae and offer pathways for charge carrier transport. With the diversity of building blocks, COFs have been developed into p-, n-, and ambipolar-type semiconductors.^21–32 With the phenazine^27,28 and C=C bond linkages,^29–32 fully π-conjugated COFs have been synthesized so that carrier transport could be achieved over the 2D plane and along with the perpendicular π-column directions, leading to carrier transport across the material. On the other hand, the channels provide nanospace for accommodating reactants, so that they are proximate to the reaction centers while products could be timely released from the catalytic site. These structural features offer the chemical base for exploring photocatalytic and electrocatalytic systems.

COFs for Photocatalysis

Photocatalysts require the combination of a set of different properties, including the capability of light-harvesting, photoinduced electron transfer, charge separation, and charge transport into one material. COFs offer an irreplaceable way to design photocatalysts as they merge these properties in one material by developing different π-electronic interfaces to control and connect these processes. Indeed, by using different π-units, these photochemical events could be designed topologically and managed synthetically. These distinct features render the designed COFs photocatalytic systems capable of fulfilling the energy conversion.

Light-driven hydrogen and oxygen evolutions

Water splitting into hydrogen and oxygen offers a way to produce green energy as hydrogen contains high energy density and serves as a chemical input for fuel cells. The splitting process consists of two half-reactions: one from the hydrogen evolution and on the reduction side, and the other is oxygen generation, which is on the oxidation side. The hydrogen reduction process requires two electrons, while the oxygen generation involves four electrons; these reactions do not proceed under ambient conditions without catalysts. Light-driven water splitting needs semiconductors with suitable redox potential and band gap that meet the reduction and oxidation reactions. A simplified scenario for the catalyst is that it absorbs a photon and splits exciton into electron and hole, while the resulting electron is used to reduce water into hydrogen, while the hole triggers water oxidation to generate oxygen. Therefore, a catalyst must have a high enough lowest unoccupied molecular orbital (LUMO) level (>−4.02 eV) for the reduction of water and low enough highest occupied molecular orbital (HOMO) level (<−5.25 eV) for water oxidation (Figure 5a). On the other hand, the photochemical processes in these half-reactions involve light-harvesting, exciton splitting, and charge transport at the reaction centers. These processes must be merged with suitable interfaces to enable the continuous flow of electrons and holes to the catalytic sites (Figure 5b). It has remained challenging to merge these processes and combine two catalytic functions to fabricate one organic material.

• Download figure
• Download PowerPoint

COFs are unique in that they predesign the π-electronic skeletons with knot and linker, as well as the linkage and hence, offer a well-defined molecular platform for designing HOMO and LUMO levels to tune semiconducting properties. Moreover, COFs have been developed into donor–acceptor structures,^21–32 which offer super heterojunction for splitting excitons into electron and hole, while the donor and acceptor π-columns enable the transport of holes and electrons, respectively. These distinct features are inaccessible with other polymeric architectures and molecular frameworks. Thus, COFs are highly promising for developing photocatalysts to split water driven by light.

A squaraine-linked copper porphyrin (CuP-SQ) COF (Figure 6a) is the first photocatalyst based on COFs that activate molecular oxygen to singlet oxygen under visible light.³³ Owing to the π-conjugated squaraine linkage and ordered π-columns, CuP-SQ COF exhibits greatly enhanced photocatalytic activity, compared with monomeric copper porphyrin complex. A hexagonal TFPT-COF, based on 1,3,5-tris(4-formyl-phenyl)triazine (TFPT) and 2,5-diethoxy-terephthalohydrazide building blocks (Figure 6b) with triazole knot and hydrazone linkage adopts a planar conformation and possesses a band gap of 2.8 eV.³⁴ TFPT-COF adsorbs visible light and triggers hydrogen evolution to achieve a rate of 230 µmol h⁻¹ g⁻¹ in the presence of sodium ascorbate sacrificial donor and a rate of 1.97 mmol h⁻¹ g⁻¹ for a system with triethanolamine (TEOA) sacrificial donor. These results demonstrate the possibility of COFs as a photocatalyst.

• Download figure
• Download PowerPoint

Azine-linked COFs have been synthesized by condensing hydrazine with knot units bearing aldehyde groups.³⁵ The azine linkage enables an extended π-conjugation in the network and offers dense nitrogen atoms on the channel walls as the N–N linker is short. Azine-linked Nx-COFs (Figures 6c–6f; x = 0−3) have been synthesized by integrating different Nx knots, that is, N₀ phenyl, N₁ pyridine, N₂ pyrimidine, and N₃ triazine units, respectively.^36,37 The different number (x) of the nitrogen atom in the knot induces different planarity of the resulting frameworks, as the twisted angle between the knot and linker decreases in the order of N₀ > N₁ > N₂ > N₃. The triazine N₃ knot yields a planar conformation of the 2D layer stacking to show a zero degrees (0°), demonstrating resistance to a twisted angle.

Noticeably, the photocatalytic activity increases in the order of N₀-COF < N₁-COF < N₂-COF < N₃-COF. Indeed, the N₀-COF, N₁-COF, N₂-COF, and N₃-COF exhibit a hydrogen evolution rate of 0.023, 0.090, 0.438, 1.703 mmol h⁻¹g⁻¹, respectively. This performance of N₃-COF originates from a multifold effect of the azine-linked triazine network. N₃-COF with planar conformation and extended π-conjugation possesses an appropriate band gap of 2.6–2.7 eV. Its π-structure promotes exciton migration and charge separation. The nitrogen atoms of triazine and azine units enable hydrogen-bonding interactions with the sacrificial donor TEOA, contributing to the hole quenching. Moreover, the electron-deficient triazine unit stabilizes the negative charges to promote electron transfer to the reduction center of Pt nanoparticles.³⁶

The azine linkage has been developed to synthesize the rhombic, A-TEBPY-COF, with a pyrene knot (Figure 6g). The pyrene knot and rhombic topology are expected to have extended π-conjugation over the framework.³⁸ In the presence of platinum (Pt) nanoparticles as a reduction center and 10 vol % TEOA as a sacrificial donor, A-TEBPY-COF exhibits a hydrogen evolution rate of 98 µmol h⁻¹ g⁻¹.

The β-ketoenamine linkage yields stable COFs. A series of β-ketoenamine-linked TP-BDDA COF, TP-EDDA COF, and TP-DTP COF (Figures 6h–6j) have been synthesized by condensing 1,3,5-triformylphloroglucinol (TFP) knot with 4,4'-(buta-1,3-diyne-1,4-diyl)dianiline, 4,4'-(ethyne-1,2-diyl)dianiline, and 4,4'-(ethyne-1,2-diyl)dianiline, respectively.³⁹ The TP-BDDA COF, TP-EDDA COF, and TP-DTP COF exhibit a band gap of 2.31, 2.34, and 2.42 eV, respectively. In the presence of Pt cocatalyst and TEOA sacrificial donor, the TP-BDDA COF, TP-EDDA COF, and TP-DTP COF generate hydrogen at a rate of 324 ± 10, 30 ± 5, and 20 ± 5 µmol h⁻¹ g⁻¹, respectively. The relatively high photocatalytic activity of the TP-BDDA COF originates from the extended π-conjugation in the lattice owing to the presence of a diacetylene linker.

Similarly, β-ketoenamine-linked TPCOF, AntCOF, TzCOF, and BtCOF have been designed to possess electron-donating TP and anthracene (Ant) blocks, and electron-accepting tetrazine (Tz) and benzothiadiazole (Bt) moieties as linker units, respectively (Figures 6k–6n).⁴⁰ To compare the effect of crystallinity, porosity, and stacking mode on the hydrogen evolution rate, these COFs have been synthesized under two distinct conditions; one is carried out in a mixture of mesitylene/dioxane (4/1 vol) in the presence of acetic acid (6 M) catalyst at 120 °C for 7 days to yield COF120 series, and the other is conducted in a mixture of mesitylene/dioxane (1/2 vol) in the presence of acetic acid (6 M) catalyst at 150 °C for 3 days to form a COF150 group. Among the series, the BtCOF150 achieves the highest hydrogen evolution rate of 750 ± 25 µmol h⁻¹ g⁻¹, in the presence of Pt cocatalyst as a reduction center and TEOA sacrificial donor. This result originates from the highest capabilities of light absorption and charge carrier generation owing to the aligned donor–acceptor π-arrays, as evident by solid-state absorption spectroscopy. This result suggests that photochemical processes and photocatalytic activity are highly related to the ordered donor–acceptor interface structure.

Hydrogen evolution from water involves reactions in the aqueous phase; the hydrophilicity of COFs is essential. The β-ketoenamine-linked FS-COF with fused-sulfone (FS) unit(Figure 6o) has been prepared by condensing TFP knot with dibenzo[b,d]thiophene sulfone (DBTS) linker to achieve a stable and hydrophilic framework.⁴¹ The FS-COF shows a contact angle of 23.6°, exhibits a Brunauer–Emmett–Teller (BET) surface area of 1288 m² g⁻¹, and adsorbs water to achieve a capacity of 67 wt%. Noticeably, the FS-COF has a band gap of 1.85 eV, emits at 670 nm in the solid-state, and exhibits a fluorescence lifetime of 5.56 ns. FS-COF is highly active in producing hydrogen from water in the presence of near-infrared absorbing dye WS5F cosensitizer, sodium ascorbate sacrificial donor, and Pt reduction center. The system is stable over a 50 h continuous run to show a constant hydrogen evolution rate of 16.3 mmol h⁻¹ g⁻¹, which is superior to Pd⁰/TpPa-1 (10.4 mmol h⁻¹ g⁻¹)⁴² cosensitized by Eosin Y and TpPa-COF-(CH₃)₂ (8.33 mmol h⁻¹ g⁻¹).⁴³ These results show that the wettability of COFs is an important parameter to be considered in designing COFs-based photocatalysts to achieve hydrogen evolution.

A C=C bond linked g-C₄₀N₃-COF (Figure 6p) has been synthesized via Knoevenagel polycondensation of 3,5-dicyano-2,4,6-trimethylpyridine (DCTMP) knot and 1,3,5-tris(4-formylphenyl)benzene) (TFPB) linker.⁴⁴ The band gap of g-C₄₀N₃-COF is 2.36 eV, and its fluorescence lifetime is 3.31 ns. Owing to the presence of pyridine knot and C=C bond linkage, g-C₄₀N₃-COF is expected to be an ambipolar conducting polymer that would enable the transport of both electrons and holes. The g-C₄₀N₃-COF forms a complex with Pt ion via coordination with its pyridine units, promoting electron transfer from the skeleton to the Pt sites. The g-C₄₀N₃-COF exhibits activity in separated hydrogen and oxygen evolutions: the hydrogen evolution rate is 129.8 µmol h⁻¹ in a system consisting of 10 vol % TEOA sacrificial donor and 3 wt % Pt, while the oxygen evolution rate is 2.5 µmol h⁻¹ g⁻¹ in a system containing AgNO₃ (0.01  M) as an electron acceptor and La₂O₃ (0.2 g) as a buffer under visible light. Similarly, a shorter C-chain, C=C bond linked g-C₁₈N₃-COF (Figure 6q)⁴⁵ with triazine knot and phenyl linker has been prepared with a band gap of 2.42 eV and emits at 510 nm with a fluorescence lifetime of 7.25 ns upon excitation at 365 nm. The g-C₁₈N₃-COF shows a higher hydrogen evolution rate of 292 µmol h⁻¹ g⁻¹ in an aqueous ascorbic acid solution (1 M) and 3 wt% Pt as the reduction catalyst.

Inspired by the sp²c-COFs,^14,32 condensing 4,4′,4″,4‴-(pyrene-1,3,6,8-tetrayl)tetrabenzaldehyde (Py-CHO) with 4,4′-(benzo[c][1,2,5]thiadiazole-4,7-diyl)dianiline, 4,4′-(5,6-dichlorobenzo[c][1,2,5]thiadiazole-4,7-diyl)dianiline, and 4,4′-(5,6-difluorobenzo[c][1,2,5]thiadiazole-4,7-diyl)dianiline forms Py-HTP-BT-COF, Py-ClTP-BT-COF, and Py-FTP-BT-COF, respectively (Figures 6r–6t). Py-HTP-BT-COF, Py-ClTP-BT-COF, and Py-FTP-BT-COF have similar skeletons with pyrene knot and Bt linker but varied in their H, Cl, and F substituents at the Bt ring. Interestingly, this structure variation induces different photocatalytic activities.⁴⁶ Py-HTP-BT-COF, Py-ClTP-BT-COF, and Py-FTP-BT-COF have optical band gaps of 2.25, 2.36, and 2.34 eV, as calculated from Tauc plots and possess the conduction band minimum (CBM) (vacuum level) at −2.76, −2.92, and −3.25 eV, respectively. Py-ClTP-BT-COF, Py-FTP-BT-COF, and Py-HTP-BT-COF exhibit hydrogen evolution rate of 44.00, 15.48, and 6.00 μmol h⁻¹, respectively. In the presence of Pt nanoparticles as reduction center (5 wt %), Py-ClTP-BT-COF, Py-FTP-BT-COF, and Py-HTP-BT-COF achieve enhanced hydrogen evolution rate of 177.50, 57.50, and 21.56 μmol h⁻¹, respectively. The apparent quantum yield (AQY) of Py-ClTP-BT-COF is 8.45%. Density functional theory (DFT) calculations suggest that the BT unit in the skeleton accounts for hydrogen evolution to generate H* intermediate on the surface. Compared with Py-HTP-BT-COF, the energy barrier of absorbing the H* intermediate on the Bt unit of Py-ClTP-BT-COF and Py-FTP-BT-COF was reduced by 0.83 and 0.15 eV, respectively. These results indicate that halogen substituents, especially the chlorine substituent at the BT linker site, reduce the energy barrier effectively to promote water reduction.

Nitrogen-containing groups such as imine (I), azine (Ai), and azo (Ao) serve as a catalytic site for hydrogen evolution⁴⁷ as nitrogen atom possesses lone pair to accept protons. On the other hand, theoretical calculations and experimental results suggest that benzene (BZ) rings function as catalytic sites for oxygen evolution.⁴⁸ Four C₃-symmetric BZ, triphenylamine (TA), 1,3,5-triphenylbenzene (TBZ), and 2,4,6-tripheyl-1,3,5-triazine (TST) were used to synthesize 12 COFs, that is, L–B networks (Figures 6c, 6f, and 7a–7j; L = I, Ai, and Ao; B = BZ, TA, TBZ, and TST), such that the hydrogen and oxygen evolution units are integrated as linkage and knot, respectively.⁴⁹ Theoretical calculations show that I-TST, Ai-TST, and Ao-TST, under visible light, possess electron potential U of 1.10, 1.06, and 0.75 eV, respectively, suggesting that the photocatalytic hydrogen evolution is spontaneous. In the first step of oxygen evolution, I-TST, Ai-TST, and Ao-TST are predicted to have the free-energy change of 1.83, 1.77, and 2.08 eV, respectively. I-TST and Ao-TST promote photocatalytic oxygen evolution at pH = 7, while Ai-TST requires an alkaline system with pH = 10 so that the free energy could be reduced from 1.89 to 1.72 eV to enable oxygen evolution. In the presence of the sacrificial agents as AgNO₃ and TEOA under visible light, I-TST exhibits average hydrogen and oxygen production rates of 12 and 8 μmol g⁻¹ h⁻¹, respectively. I-TA, Ai-TA, and Ao-TA possess a CBM higher than the reduction potential of H⁺/H₂ and valance band minimum (VBM) higher than the oxidation potential O₂/H₂O, suggesting that these COFs enable hydrogen production under visible light.

• Download figure
• Download PowerPoint

COFs have been hybridized with other materials to form COFs/metal–organic frameworks (MOFs),⁵⁰ COFs/metal nanoparticles,⁵¹ COFs/metal sulfides,⁵² and COFs/molecular catalysts.⁵³ Recently, COFs have been hybridized with MXenens (a group of 2D transition metal nitrides/carbides/carbonitrides) to construct photocatalysts. In this case, MXenes with tunable reactive termini have been covalently linked to COFs.⁵⁴ NTU-BDA-HTA, NTU-BDA-DHTA, and NTU-BDA-THAT (Figures 7k–7m) have been synthesized by condensing benzene-1,4-diamine (BDA) with 2-hydroxybenzene-1,3,5-tricarbaldehyde (HTA), 2,4-dihydroxybenzene-1,3,5-tricarbaldehyde (DHTA), and 4,6-trihydroxybenzene-1,3,5-tricarbaldehyde (THTA), respectively under solvothermal conditions. The NTU-BDA-HTA, NTU-BDA-DHTA, and NTU-BDA-THTA possess optical band gap of 1.81, 1.77, and 2.09 eV, respectively. In the presence of Pt reduction center and l-ascorbic acid sacrificial donor in a buffer solution (pH = 7), NTU-BDA-HTA, NTU-BDA-DHTA, and NTU-BDA-THTA under visible light, exhibit hydrogen evolution rate of 40, 800, and 1200 μmol g⁻¹ h⁻¹, respectively. Various hybrids COF/NH₂-Ti₃C₂T_x have been prepared by covalent condensation of COFs with NH₂-Ti₃C₂T_x. The resulting ATNT-2 (NTU-BDA-THAT/NH₂-Ti₃C₂T_x = 8/2) achieves a hydrogen evolution rate of 10,000 μmol g⁻¹ h⁻¹, which is 4.2 and 10.38-fold as high as those of noncovalently connected ATNT-2-HB and ATNT-2-PM simply mixed from NTU-BDA-THTA and NH₂-Ti₃C₂T_x, respectively. This result indeed demonstrates the exact role of covalent bonds between NTU-BDA-THTA and NH₂-Ti₃C₂T_x in improving the photocatalytic activity. ATNT-4 (NTU-BDA-THAT/NH₂-Ti₃C₂T_x = 8/4) achieves the maximum hydrogen production rate of 14,228.1 μmol g⁻¹ h⁻¹ and is 12.6-fold as high as that of NTU-BDA-THAT (1127.1 μmol g⁻¹ h⁻¹). ATNT-4 retains activity after six cycles over 30 h without showing noticeable changes in powder X-ray diffraction (PXRD) pattern and morphology. Interestingly, in the presence of AgNO₃ (0.05 M), ATNT-4 produces 2.3 μmol oxygen from water in 1 h. The enhanced photocatalytic activity of ATNT-4 originates from a smaller thermodynamic driving force with oxygen evolution.

Hydrogen evolution from water involves a series of photochemical events, and how to connect them is key to the reaction. This requires the design of various interfaces to enable a continuous flow of electron and hole to pump the reaction. However, it lacks a general principle for designing organic systems to interlock multiple interfaces, so that light absorption, exciton splitting, and charge transfer (CT) and transport can be merged seamlessly. As COFs enable a systematic design of the skeleton and pore, it is highly possible to engineer various interfaces into one framework. For this purpose, we sought to explore C=C bonds linked sp² carbon COFs (Figure 7n), which have a tetragonal topology so that the lattice is fully π-conjugated along the x and y directions.³² This is distinct from hexagonal COFs as their π-conjugation is blocked by the knot site.

We have synthesized sp²c-COF with a pyrene knot and phenylenevinylene linker to form a semiconducting structure with a band gap of 1.9 eV. This low band gap enables the harvesting of visible light for photoexcitation. We integrate an electron-deficient 3-ethylrhodanine (ERDN) unit to the periphery of sp²c-COF via a three-component polycondensation to form sp²c-COF_ERDN. The sp²c-COF and sp²c-COF_ERDN have the LUMO level of −3.84 and −3.81 eV, and band gap of 1.90 and 1.85 eV, respectively. The sp²c-COF_ERDN constitutes electron donor (sp² carbon skeleton) and acceptor (periphery ERDN) heterojunctions to trigger donor–acceptor push–pull effects and to induce exciton splitting (Figure 7o).⁵⁵ In sp²c-COF_ERDN, the structures are collaborative in connecting photochemical events. The π-lattice with a full π-conjugation enables the light absorbance extended to 800 nm, the π-arrays offer pathways for exciton migration, as well as the charge transport, and the heterojunctions split excitons into electron and hole. Moreover, the 2 nm-sized pores install small Pt nanoparticles, such that the distance from the sp² carbon skeleton to Pt center is shortened considerably. Therefore, the photochemical processes in sp²c-COF_ERDN are consecutively connected seamlessly. In the presence of Pt nanoparticles and TEOA sacrificial donor, sp²c-COF_ERDN is highly active to promote hydrogen evolution; the evolution rate is 2.12 mmol h⁻¹ g⁻¹ upon irradiation with light ≥420 nm (Figure 8a, red circles and line) and is 1.24 mmol h⁻¹ g⁻¹ upon irradiation with light ≥495 nm (Figure 8b, red circles and line). The AQY is 0.46% and 0.2% upon irradiation at 420 and 578 nm, respectively (Figure 8c). The sp²c-COF_ERDN is much superior to sp²c-COF (Figures 8a and 8b, blue circles and lines; rate = 1.325 mmol h⁻¹ g⁻¹ under light ≥420 nm), amorphous analogy (Figures 8a and 8b, black circles and lines; rate = 0.14 mmol h⁻¹ g⁻¹ under light ≥420 nm), and imine-linked COF shows a trace of hydrogen evolution upon irradiation (Figures 8a and 8b, black triangles and lines). Moreover, the sp²c-COF_ERDN keeps the structural integrity of crystallinity and porosity and maintains the same efficiency over a 20 h continuous run (Figure 8d).

• Download figure
• Download PowerPoint

The sp²c-COF is capable of catalyzing water oxidation to generate oxygen owing to a low-HOMO level (−5.74 eV). In the presence of Co(NO₃)₂ cocatalyst and AgNO₃ acceptor in an aqueous La₂O₃ buffer, the sp²c-COF promotes oxygen evolution to achieve a rate of 22 µmol h⁻¹ g⁻¹.⁵⁵ These results indicate that sp²c-COFs offer a platform for designing molecular systems to enable the connection of various photochemical events; this is important as it suggests a possibility of combining both reduction and oxidation into one system, which has remained challenging in organic photocatalysis.

The Bp-COF-1 (Figure 7p), upon coordination with Co(II) via the bipyridine unit, shows a HOMO level of −6.21 eV has been synthesized to promote oxygen evolution.⁵⁶ An oxygen evolution rate of 152 µmol h⁻¹ g⁻¹ is achieved in the presence of AgNO₃ in an aqueous solution. However, its crystallinity decreases upon photocatalytic reaction, thereby inhibiting its cyclability. Exploring COFs that combine low-enough HOMO level with enough stability is an important issue worthy of studies.

Light-driven carbon dioxide reduction

Carbon dioxide is a greenhouse gas yet a carbon resource; its adsorption, storage, and conversion are highly desirable relative to tackling global warming and exploring carbon resources. In this context, the photoreduction of carbon dioxide into value-added chemicals is a subject of great importance for developing new natural resources. The reduction of carbon dioxide into different chemicals requires different and multiple numbers of electrons and involves different reaction processes; in the case of reduction into carbon monoxide, the LUMO level of COFs must be higher than −3.91 eV (Figure 5a). COFs are robust to combine catalytic activity and light-harvesting capability in one material. More importantly, they are predesignable for creating suitable structures to control these properties by developing different knots, linkers, and functional sites. Two major approaches have been developed to design COFs as photocatalysts: integrating various catalytic sites of metal species to form a metal complex in the skeletons has enabled the development of diverse COFs-based photocatalysts while exploring organic sites as the catalytic center to achieve carbon dioxide reduction shows the potential of nonmetal-based systems.

Various metal complexes have been developed as catalysts for carbon dioxide reduction. Integrating these metal complexes into the skeleton of COFs enables the construction of photocatalysts. Complexes with Re(I), Ni(II), Ru(II), Co(II), and Zn(II) ions have been integrated to develop photocatalytic COFs. For example, a COF with a bipyridine linker has been used to coordinate with Re(I) ion to form Re-COF (Figure 9a).⁵⁷ As the skeleton of COF enables extended π-conjugation, Re-COF promotes the photoinduced charge separation. Upon irradiation, the photoexcited skeleton transfer electrons efficiently to the Re center for carbon reduction into carbon monoxide. Carbon monoxide evolves at a rate of 60 mmol h⁻¹ g⁻¹ with a 98% selectivity over hydrogen in the presence of TEOA sacrificial donor and under irradiation with a Xe lamp. On the other hand, the Ru-TpBpy COF with Ru(II) bipyridine linker (Figure 9b)⁵⁸ displayed a carbon dioxide evolution rate of 0.28 mmol h⁻¹ g⁻¹. Surprisingly, the Ni-TpBpy COF with Ni(II) bipyridine linker (Figure 9b) demonstrated the highest carbon monoxide generation rate of 0.966 mmol h⁻¹ g⁻¹, with 96% selectivity, using [Ru(bpy)₃]Cl₂ as a photosensitizer.⁵⁹ Nevertheless, in this system, the degradation of Ru(bpy)₃²⁺ and the leak of Ni(II) ions are two main issues to be addressed.

• Download figure
• Download PowerPoint

The coordination of Co(II) and Zn(II) via carbonyl groups produces DQTP COF-Co and DQTP COF-Zn, respectively (Figure 9c).⁶⁰ DQTP COF-Co and DQTP COF-Zn possess a LUMO level of −0.85 and −0.61 V, respectively (vs Normal Hydrogen Electrode (NHE)). The resulting DQTP COF-Co achieves carbon dioxide reduction to carbon monoxide at a rate of 1.02 mmol h⁻¹ g⁻¹ and a selectivity of 59.4% over hydrogen in acetonitrile with TEOA sacrificial donor upon irradiation with lights ≥420 nm. On the other hand, under identical conditions, the DQTP COF-Zn exhibits the conversion of carbon dioxide to formic acid with a rate of 0.1525 mmol h⁻¹ g⁻¹ and a selectivity of 90% over carbon monoxide. The mechanism for carbon dioxide reduction is a two-pathway process. The electron-rich coordination environment in DQTP COF-Co weakens and breaks the C=O bond to form carbon monoxide, and the electron deficiency coordination environment in DQTP COF-Zn enhances the C–O bond for an easy formation of formic acid.⁶⁰

Exfoliation of COFs into nanosheets offers a larger surface area and easily accessible catalytic sites than bulk COF crystallites. For example, exfoliated nanosheets of COF-367-Co (Figure 9d) with a thickness of 1.0–1.2 nm are highly active in catalyzing carbon dioxide reduction.⁶¹ Carbon monoxide is formed at a rate of 10.162 mmol h⁻¹ g⁻¹ and 78% selectivity in the presence of [Ru(bpy)₃]Cl₂ photosensitizer and ascorbic acid sacrificial donor. Notably, the rate is two orders of magnitude as high as that of bulk COF-367-Co (0.124 mmol h⁻¹ g⁻¹) under otherwise identical conditions. This result demonstrates the importance of exposing catalytic sites to carbon dioxide when designing COF systems.

Based on the fully π-conjugated sp²c-COF-2 prepared by our group,¹⁴ using a bipyridine linker instead of biphenyl produces Bpy-sp²c-COF.⁶² Re-Bpy-sp²c-COF (Figure 9e) is prepared by complexation with Re(I) ion and promotes carbon dioxide reduction to carbon monoxide with a rate of 1.04 mmol h⁻¹ g⁻¹ and an 81% selectivity. The rate is enhanced to 1.40 mmol h⁻¹ g⁻¹, and the selectivity is improved to 86% by using a dye {Ir[dF(CF₃)ppy]₂(dtbpy)}PF₆ (1.0 mmol) as a sensitizer. Distinct from the imine-linked (C=N) COFs with partial π-conjugation, the C=C bond linked Bpy-sp²c-COF is fully π-conjugated to facilitate the photochemical processes. This result again emphasizes the importance of C=C bond linkages for exploring efficient photocatalytic systems.

Removing the sacrificial electron donor from the catalytic cycle is essential to enable a green conversion. The way to such a system is to lower the HOMO level of COFs so that the generated hole can be used in water oxidation. The azine-linked ACOF-1 (Figure 7e) displays a band gap of 2.69 eV, HOMO level of −5.00 eV, and LUMO level of −2.54 eV, while the azine-linked N₃-COF (Figure 6f) exhibits a band gap of 2.57 eV, HOMO level of −4.95 eV, and LUMO level of −2.71 eV.⁶³ These band gaps and HOMO and LUMO levels are suitable for carbon dioxide reduction to carbon monoxide and water oxidation. ACOF-1 (10 mg) achieves a methanol production rate of 0.36 µmol h⁻¹ g⁻¹ in carbon dioxide saturated water (5 mL) upon irradiation with light of 420–800 nm, while N₃-COF displays a rate of 0.57 µmol h⁻¹ g⁻¹ under otherwise same conditions. Notably, the rate is much superior to that of g-C₃N₄ (0.2 µmol h⁻¹ g⁻¹). The high activity of N₃-COF originates from a multifold effect. The pores enable adsorption of carbon dioxide, the azine linkage increase wettability, and the π-conjugation distributes π-clouds to HOMO and LUMO, hence, facilitating the photoinduced charge separation.

Condensation of tetrathiafulvalene (TTF) with metalloporphyrin enables the synthesis of TTCOF-M (Figure 9f) bearing different metal ions.⁶⁴ TTCOF-Zn is suitable for carbon dioxide reduction as it possesses HOMO and LUMO levels at −4.27 and −5.76 eV, respectively. In the absence of a sacrificial electron donor, TTCOF-Zn produces carbon monoxide with a rate of 2.055 µmol h⁻¹ g⁻¹ and a nearly 100% selectivity. In this case, the irradiation of the skeleton pumps electrons from the HOMO level on the TTF unit to the LUMO level on the zinc porphyrin site, while zinc porphyrin reduces carbon dioxide, and the TTF unit is neutralized by transferring a hole to water to trigger its oxidation. This reaction cycle completes with the photoreduction of carbon dioxide and water oxidation. Noticeably, TTCOF-Zn is recyclable to retain activity after five cycles. This system showcases the importance of designing donor–acceptor COFs to enable both reduction and oxidation reactions. Studies toward donor–acceptor COFs^28,30 would offer a way to produce value-added chemicals from carbon dioxide and water driven by sunlight.

By replacing the phenyl linker of COF-366 with 2,5-dibromo phenyl produces TAPBB-COF (Figure 9g),⁶⁵ the valence band of COF-366 (Figure 9i) obtained is +0.86 V, while the O₂/H₂O redox potential is +0.82 V, leading to an electrical potential difference of only 0.04 V; this makes the oxidation reaction difficult, yielding a low carbon monoxide formation rate of 8.5 μmol g⁻¹ h⁻¹ under full-wavelength light. Halogenation of organic compounds is a standard method for adjusting energy bands.⁶⁶ Indeed, in COFs, the introduction bromine substituents to the linker enables progresses π-delocalization and promotes photocatalytic reactions.⁶⁵ Under simulated sunlight (200–1000 nm), and in the absence of metals and sacrificial agents, the carbon monoxide generation rate is 24.6 μmol g⁻¹ h⁻¹, which is nearly threefold that of COF-366 (8.5 μmol g⁻¹ h⁻¹). Under visible light (λ ≥ 430 nm), the carbon monoxide generation rate is 12.4 μmol g⁻¹ h⁻¹, also nearly threefold as high as that of COF-366 (3.9 μmol g⁻¹ h⁻¹). Under full-wavelength light, TAPBB-COF exhibits a selectivity of 95.6% to generate carbon monoxide of 295.2 μmol g⁻¹ and hydrogen of 13.6 μmol g⁻¹ over a 12 h run. The porphyrin ring, imine linkage, and bromine play roles in the activation and conversion of carbon dioxide as revealed by DFT calculations. Integrating electron-deficient linkers into the lattice to decrease the HOMO level offers a way to design photocatalytic COFs.

Condensing triazine knot and carbazole linker yields a donor–acceptor CT-COF (Figure 9h).⁶⁷ CT-COF exhibits conduction and valence bands of −0.88 and 1.16 V, respectively. Under visible light (λ ≥ 430 nm), CT-COF shows a carbon monoxide generation rate of 102.7 µmol h⁻¹ g⁻¹, which is 68.5-fold as high as that (1.5 µmol g⁻¹ h⁻¹) of g-C₃N₄. Notably, the rate is superior to metal-containing TT-COF-Zn (2.055 µmol g⁻¹ h⁻¹)⁶⁴ and hybrid COF-318-TiO₂ catalyst (69.67 µmol g⁻¹ h⁻¹).⁶⁸ The system retains a rate of 95.5 µmol h⁻¹ g⁻¹ after three cycles, while the selectivity is >98%. This performance stems from the donor–acceptor skeleton that improves the light-harvesting capability, facilitates charge transport, and decreases charge recombination, while the triazine knots enable dipole–quadruple interactions with carbon dioxide so that the electron transfer from the skeleton to carbon dioxide is facilitated.

The reduction of carbon dioxide to formic acid and formaldehyde was achieved with TFP-DM-COF (Figure 9j) as the catalyst and water as the solvent under visible light.⁶⁹ TFP-DM-COF has been synthesized by condensing TFP with 3,3′-dimethoxy-[1,1′-biphenyl]-4,4′-diamine (DM) in the presence of a p-toluene sulfonic acid catalyst. The TFP-DM-COF possesses an absorption edge at 658 nm, giving an optical band gap of 1.89 eV. At room temperature, and under white light-emitting diode (LED) light (20 W cm⁻², visible light), TFP-DM-COF (10 mg) reduces carbon dioxide into formic acid (0.027 mol) and formaldehyde (0.96 mol) in 5 h. Notably, under sunlight, the TFP-DM-COF generates formic acid (0.0195 mol) and formaldehyde (0.47 mol) in 8 h. After five cycles under white LED light, the rate of formic acid and formaldehyde formation decreases from 0.54 to 0.46 mol g⁻¹ h⁻¹ and 19.2 to 18.4 mol g⁻¹ h⁻¹, respectively.

COFs for Electrocatalysis

Electrocatalysis offers a direct approach to produce value-added chemicals in which the primary target is to reduce carbon dioxide to carbon monoxide and other chemicals. Electrocatalysis is positively related to energy conversion as electrocatalytic reduction of oxygen is a critical reaction in fuel cells, as shown in the transformation of hydrogen by oxygen, or oxidation of alcohols, with the generation of electricity at high efficiency, compared with thermal combustion. Electrocatalysis is promoted by COFs deposited on an electrode, which injects electrons into the COFs to trigger energy transformations. Therefore, electrocatalysis is highly dependent on the contact between the electrode and COFs, whereas the interface is critical to the easiness of electron injection into the COFs. The conducting property of COFs determines the electron transport to the catalytic sites, and the porosity controls the delivery of reactants and the release of products. The porous π-structure of COFs fulfills both conductivity and porosity requirements and enables the design of catalytic sites, as well as interfaces to control the reactions. Therefore, in electrocatalysis, COFs offer a platform for merging conductivity, porosity, and catalytic activity into one skeleton.

Electrocatalytic reduction of oxygen

The oxygen reduction is a reaction involving multielectron in the cathode of fuel cells, which, currently, is catalyzed primarily by Pt nanoparticles. Exploring nonmetal electrocatalyst to replace Pt requires the merging of conductivity with catalytic activity. For this purpose, various porous carbon materials have been investigated by pyrolyzing precursors to form conducting carbon structures, while gaining control over the spatial distribution of catalytic N and P heteroatom sites remains a challenge. COFs consist of well-defined 2D layer structures, which are integrated easily into various heteroatoms’ skeletons in a predesignable fashion. As a precursor of pyrolysis, COFs offer a unique platform for presetting the structure and components to combine conductivity and catalytic activity. However, as pyrolysis proceeds as a radical process at elevated temperatures, the morphology of COFs is easily destructed, so that the resulting carbon structure and the spatial distribution of the catalytic sites are difficult to control.

The integration of metal ions into COFs enables the preparation of carbons with residue metal participles. For example, using metalloporphyrin COFs as a precursor for pyrolysis generates metal-containing porous carbons in which the resulting metal nanoparticles are expected to serve as catalytic sites. The Co(II) porphyrin Co-COF (Figure 9i) has been investigated for carbonization at 900 °C to prepare graphitized nitrogen-doped carbons with distributed Co(0) nanoparticles.⁷⁰ In an aqueous solution of KOH (0.1 M), the resulting Co-COF-900 exhibits an onset potential and current density of −0.03 V and −0.16 mA cm⁻², which are comparable with those of the commercial 20% Pt/C catalyst. The rotating ring-disk electrode (RRDE) experiments reveal that the reduction reaction involves 3.86 electrons, demonstrating that Co-COF-900 promotes oxygen reduction via a four-electron process.

The C=C bond linked 2D polymer (2DPPV) has been used as a precursor for preparing porous carbon (Figure 10a).⁷¹ 2DPPV has been synthesized by condensing 1,3,5-triformyltriphenylbenzene knot and 1,4-phenylenediacetonitrile linker to exhibit a BET surface area of 472 m² g⁻¹, a pore size of 1.6 nm, and a pore volume of 0.37 cm³ g⁻¹. 2DPPV was pyrolyzed at 800 °C to form 2DPPV-800, which was further treated with ammonia at 800 °C for 15 min to yield 2DPPV-800a. This modified 2DPPV is more porous, exhibiting a BET surface area of 880 m² g⁻¹. Under an alkaline condition (0.1 M KOH), 2DPPV-800a displays a half-wave potential of 0.67 V (vs RHE) and a current density of −4.6 mA cm⁻² at 0.2 V, respectively. From RRDE, the oxygen reduction process catalyzed by 2DPPV-800a involves 3.71 electrons. These results indicate that ultrathin graphitic layers with high electrical conductivity and porosity promote electron and ion transport in an electrochemical process.

• Download figure
• Download PowerPoint

We have developed a hexagonal mesoporous TAPT-DHTA-COF (Figure 10b) by condensing 4,4′,4″-triamino-2,4,6-triphenyltriazine (TAPA) with 2,5-dihydroxyterephthaldehyde (DHTA) under solvothermal conditions.⁷² Usually, pyrolysis of COFs destroys the layer structure, as the thermal process involves complex and uncontrollable radical reaction procedure. To sustain the layer structure of COFs, we have explored a template strategy for guiding the carbonization process to produce layered carbons. We have developed phytic acid (PA) as the template and prepared PA@TAPT-DHTA-COF by mixing PA with TAPT-DHTA-COF to form hydrogen-bonding networks. Not only the channels but also space between layers are filled with PA so that the structure of TAPT-DHTA-COF is locked by PA. PA@TAPT-DHTA-COF upon hydrolysis at 1000 °C under nitrogen forms PA@TAPT-DHTA-COF@1000, which possesses pores of 0.7 and 1.6 nm and abundant catalytic edges doped with nitrogen and phosphorus heteroatoms.

The electrocatalysis is conducted in an alkali solution because alkaline-type fuel cells are more strategic planned, as they enable the use of a broad range of fuels, including alcohols, along with high efficiency, stability, and low cost, compared with proton-based fuel cells. Cyclic voltammetry measurements in an aqueous KOH solution (0.1 M) saturated with oxygen using Ag/AgCl as a reference electrode revealed a peak at −0.23 V, attributed to the reduction of oxygen by PA@TAPT-DHTA-COF₁₀₀₀. The RRDE experiments at a rotation rate of 1600 rpm in the aqueous KOH solution (0.1 M) saturated with oxygen show that TAPT-DHTA-COF₁₀₀₀ upon direct pyrolysis of TAPT-DHTA-COF without PA exhibits an onset potential of −0.12 V, a halfwave potential of −0.29 V, and a diffusion-limiting current density of 4.0 mA cm⁻², respectively (Figure 11a, green squares and curve). In contrast, PA@TAPT-DHTA-COF₁₀₀₀ achieves an onset potential of −0.02 V, a halfwave potential of −0.19 V, and a diffusion-limiting current density of 6.5 mA cm⁻², respectively (Figure 11a, blue circles and curve). PA@TAPT-DHTA-COF₁₀₀₀ is close or superior to that of the commercial Pt/C (Figure 11a, black pentagons and curve; onset potential, −0.03 V; halfwave potential, −0.16 V; diffusion-limiting current density, 6.0 mA cm⁻²). The outperformance of PA@TAPT-DHTA-COF₁₀₀₀ originates from a synergistic structural effect. First, the nanosized 2D carbon sheets with pores expose numerous edges for catalytic loading sites, while the presence of graphitic structures improves conductivity. Second, the high content of pyridinic N (N1) sites enhances the catalytic activity substantially. Third, fast-electron transport from the electrode to catalytic sites becomes possible as a result of increased conductivity.

• Download figure
• Download PowerPoint

We further treated PA@TAPT-DHTA-COF₁₀₀₀ in an ammonia atmosphere at 900 °C to prepare PA@TAPT-DHTA-COF_1000NH3. As a result, the use of NH₃ as an atmospheric gas in the second-stage pyrolysis not only increased the number of active sites by reacting with the edge parts and removing oxygen atoms from the surface but also reconstructed the porous structure to form hierarchical pores with a broad pore size distribution of 0.5–6 nm. PA@TAPT-DHTA-COF_1000NH3 displays a reduction peak at −0.18 V, which is more favorable than that (−0.23 V) of PA@TAPT-DHTA-COF₁₀₀₀, suggesting a markedly enhanced catalytic activity. The RRDE measurements of PA@TAPT-DHTA-COF₁₀₀₀ (Figure 11a, red triangles and curve) show an onset potential of 0 V, a halfwave potential of −0.11 V, and a diffusion-limiting current density of 7.2 mA cm⁻², respectively, which is much superior to the commercial Pt/C.

The superior performance of PA@TAPT-DHTA-COF_1000NH3 originates from the mesopores that act as an interconnected highway to provide quick delivery of reactant and release of the product. From the Tafel slope, PA@TAPT-DHTA-COF1000NH₃ shows a small Tafel slope of 110 mV decade⁻¹ compared with those of PA@TAPT-DHTA-COF₁₀₀₀ (146 mV decade⁻¹) and Pt/C (121 mV decade⁻¹). The above results confirm that PA@TAPT-DHTA-COF_1000NH3 is superior to PA@TAPT-DHTA-COF₁₀₀₀ and Pt/C. The electron-transfer number of PA@TAPT-DHTA-COF_1000NH3 is 3.77–3.98, with a low H₂O₂ yield (1.5–11%) between −0.2 and −1 V as evaluated using the RRDE method (Figure 11b). Moreover, linear scan voltammogram experiments revealed that the Koutechy–Levich plot offers a linear curve to reach the electron-transfer number of nearly 4, similar to the values based on the RRDE measurements. TAPT-DHTA-COF_1000NH3 catalysis is robust to methanol without showing any alteration in current density; in contrast, upon the addition of methanol to Pt/C, a dramatic decrease in current density is observed. Besides oxygen reduction, TAPT-DHTA-COF_1000NH3 serves as a catalyst to promote the oxygen evolution reaction in an aqueous KOH solution (0.1 M). TAPT-DHTA-COF_1000NH3 achieves a current density of −10 mA cm⁻² at 0.97 V, which is one order of magnitude higher (1.01 mA cm⁻²) than that of Pt/C.

Electrocatalytic reduction of carbon dioxide

Electroreduction of carbon dioxide proceeds in water and is always accompanied by water reduction, which forms hydrogen at a lower reduction potential (Figure 5a); this side reaction decreases the reaction selectivity and Faradaic efficiency, making the reaction the most challenging goal to achieve. The exploration of electrocatalysts for carbon dioxide reduction is critical, concerning the enhancement of reactivity and selectivity, and one way to prevent the side reaction is to integrate metallocomplex into the formulated electrocatalyst that is specific to the carbon dioxide reduction reaction.

Co(II) porphyrin is an electrocatalyst used typically for carbon dioxide reduction. Porphyrin COFs with different metal species and linkages have been synthesized by condensing porphyrin or metalloporphyrin knots with various linkers.^73–78 Integration of Co(II) porphyrin into the COF hexagonal skeleton enables the synthesis of Co(II) porphyrin COF-366-Co (Figure 9i).⁶¹ Depositing COF-366-Co on a glass electrode enables the preparation of COF-366-Co electrocatalyst for carbon dioxide reduction. In an aqueous potassium bicarbonate solution (0.5 M) saturated with carbon dioxide, COF-366-Co exhibits a carbon monoxide selectivity of 90%, as determined by Faradaic efficiency, with an overpotential of −0.55 V (vs reversible hydrogen electrode, RHE). The reduction reaction likely undergoes at the Co(II) porphyrin sites on the exterior surface layer of COFs, while most sites in the bulky framework are not involved in the reaction. In this case, COF-366-Co shows a current density of 5 mA cm⁻² at −0.67 V (vs RHE) and a turnover frequency of 98 h⁻¹.

To enhance the catalytic activity, COF-367-Co(1%) was prepared by diluting the Co(II) porphyrin catalytic sites with catalytically inert Cu(II) porphyrin to a 1% content. COF-367-Co(1%) displays a selectivity of 40%, an overpotential of −0.55 V, a turnover number of 9600 for each Co(II) porphyrin site, and a turnover frequency of 296,300 h⁻¹.

Using highly ordered pyrolytic graphite as a supporting base to prepare a thin film of COF-366-Co electrocatalyst improved the current density to 45 mA mg⁻¹, which is ninefold as high as that of bulk COF-366-Co, suggesting an enhanced accessibility at the catalytic sites.⁶¹ Integrating different electron donating and withdrawing groups such as methoxy and fluorine to the benzene ring unit enabled the synthesis of COF-366-(OMe)₂-Co, COF-366-(F)₄-Co, and COF-366-F-Co (Figure 10c), respectively. Among the series, the COF-366-F-Co achieved a current density of 65 mA mg⁻¹.⁶⁶

Combining the electron-rich TTF unit with Co(II) porphyrin catalytic sites yields TTF-Por(Co)-COF (Figure 9f).⁷⁹ TTF-Por(Co)-COF exhibits an overpotential of −0.7 V (vs RHE), a selectivity of 95% at −0.7 V, and a current density of 6.88 mA cm⁻² at −0.9 V in an aqueous potassium bicarbonate solution (0.5 M) saturated with carbon dioxide.

Replacing Co with Fe is an interesting test; a FeDhaTph-COF (Figure 10d) containing Fe(III) species was synthesized by condensing 5,10,15,20-tetrakis-(4-aminophenyl)-porphyrin Fe(III) chloride knot (FeTAPPCl) with 2,5-dihydroxyterephthalaldehyde (Dha) linker.⁸⁰ Depositing FeDhaTph-COF on carbon cloth yields an electrode for testing carbon dioxide reduction. In this case, the selectivity is 80%, and the turnover frequency reaches 600 and 800 h⁻¹ mol⁻¹ of electroactive Fe sites in dimethylformamide and acetonitrile, respectively.⁸¹

By condensing tetra-(4-anilyl)methane knot with a terephthalaldehyde linker, the imine-linked COF-300 was formed (Figure 10e), which upon reduction with sodium borohydride, yields an amine-linked COF-300-AR (Figure 10e).⁸² Depositing COF-300-AR on a flat silver forms electrode, stable in aqueous HCl (6 M) and NaOH (6 M) solutions. COF-300-AR on the silver electrode in an aqueous potassium bicarbonate solution (0.1 M) exhibits a current density of 0.5 and 2 mA cm⁻², and a selectivity of 53% and 80% at overpotentials of −0.70 and −0.85 V, respectively. The selectivity is superior to that of COF-300-AR on a bare silver electrode (13% and 43%). The COF-300-AR captures carbon dioxide to form carbamate through the interaction with the amine linkage, while the carbamate is reduced to carbon monoxide by electrons directly from the silver electrode.

Stability, conductivity, and catalytic activity are major issues for exploring electrocatalysts to reduce carbon dioxide. To merge these requirements of different structural aspects into one material, we have explored a stable conducting COF with built-in catalytic sites in the lattice. We designed and synthesized CoPc-PDQ-COF (Figure 10f) by condensing Co(II) 2,3,9,10,16,17,23,24-octakis(amino) phthalocyanine with 4,5,9,10-pyrenediquinone (PDQ) under solvothermal conditions.⁸³ The phenazine linkage offers a platform to combine both stability and conductivity in one framework; phenazine is a fused six-membered ring that is chemically stable and connects the tetragonal lattice into a fully π-network with extended π-conjugation along both x and y directions to improve the electrical conductivity. Moreover, CoPc-PDQ-COF is stable in cyclohexane, methanol, dimethylformamide, tetrahydrofuran, dimethyl sulfoxide, boiling water, concentrated HCl (12 M) solution, and aqueous NaOH (14 M) solution; thus, crystallinity and porosity could be retained even after 40 days.

Linear sweep voltammetry (LSV) of CoPc-PDQ-COF reveals a steep increase at −0.40 V that originates from carbon dioxide reduction. In this case, CoPc-PDQ-COF achieves a geometric cathodic current density of 49.4 mA cm⁻² at −0.66 V (vs RHE) (Figure 12a, red curve). The selectivity is 96% at −0.67 V, which is much higher than that (53%) of COF-366-Co. From the linear Tafel plot, the slope is 112 mV decade⁻¹ at 0.24–0.42 V, which suggests that the CO²⁻ anion is an intermediate produced by electron transport from COF to carbon dioxide, which is the key step in the reduction reaction. The mass-based catalytic activity is 762 mA mg⁻¹ at −0.66 V. Based on the effective Co(II) catalytic sites for the reduction, the turnover number, and turnover frequency are evaluated to be as high as 320,000 and 11,412 h⁻¹ at −0.66 V. Notably, this catalytic activity is 32-fold as high as that of monomeric Co(II) phthalocyanine (Figure 12a, blue curve), and much far superior to commercial CoPc (Figure 12a, black curve). The turnover number and frequency are higher than those of COF-367-Co(1%) (290,000 and 9396 h⁻¹) and COF-367-Co (1900 and 165 h⁻¹). The performance of CoPc-PDQ-COF is much superior to the state-of-the-art catalysts represented by perfluorinated CoPc, nanoporous Ag, Au nanowires, and Pd nanoparticles. Its current density and selectivity were retained over a 24 h continuous run (Figure 12b), sustaining constant current density and selectivity, distinct from COF-366-Co, whose turnover frequency dropped from 98 to 56 h⁻¹ after 24 h run. The stable catalytic activity originates from the fully π-conjugated and a stable skeleton that prevents the leak of the catalytic site and generates high catalytic current density (Figure 12c).

• Download figure
• Download PowerPoint

Spin-polarized DFT calculations on single-layer CoPc-PDQ-COF reveal that the reduction proceeds on the Co(II) center (Figure 12d).⁸³ The first step is converting carbon dioxide coordinated to the Co(II) phthalocyanine into COOH* by one electron, while the transformation of COOH* to carbon monoxide in the second step. The free-energy changes (ΔGs) for the two reaction steps are −0.48 and −0.06 eV at 298 K, respectively. The valence change of the metal ion from Co(II) to Co(I) occurs in the first step, driven by the electron injection, which is majorly on the Co d_z2 orbital and minorly on the carbon p_z orbital of Pc ring.

The CoPc-PDQ-COF is distinct from other electrocatalysts owing to its structural features that can be fully used in the electrocatalytic reduction of carbon dioxide. The tetragonal COF explores phenazine linkage to integrate extended π-conjugation, a robust skeleton, and active catalytic sites so that the resulting framework combines stability, conductivity, and catalytic activity in one material. The COF electrocatalyst achieves exceptional turnover frequency and selectivity. These results established a platform based on COFs for designing organic electrocatalysts to convert carbon dioxide into value-added chemicals.

Perspectives and Remarks

Environmental and energy issues are vital to the ecosystem on this planet and are related to the sustainable development of society. The challenges encountered regarding these endeavors could be mainly minimized dependent on how we explore chemistry and materials science to establish efficient catalytic systems. COFs, owing to their unique structural features, offer an irreplaceable platform for engineering organic materials to confront environmental and energy issues.

From a design and synthesis perspective, COFs’ development is in the infancy stages, as we still have critical fundamental issues to address. The thermodynamics and kinetics of the polymerization, as well as crystallization processes, remain unclear. Although certain COFs have been prepared in single crystals, most COFs are fabricated as polycrystalline materials. The formation of defects and their mechanisms are still not well understood; currently, less is known about the defects and their 3D distributions in the framework. This is a subject that connects with the preparation of high-quality crystallites of COFs. A diversity of reactions has been developed for the synthesis of COFs; however, many are very limited in terms of crystallinity, monomer scope, and reaction conditions. Exploring reactions that are widely applicable to form stable crystalline porous frameworks is still a significant and challenging venture. In this case, the crystallinity and porosity need to be considered, as materials of low crystallinity and porosity hardly present the inherent properties of a COF with the desired structural integrity and functionality. Studies on these basic issues stand a great chance to disclose the nature of COFs both in structure and property.

There is an increasing tendency of lab-scale studies on properties and functions of COFs for various applications; however, some reports have not indicated identifying these inherent characteristics of COFs owing to their low crystallinity and porosity. How the macroscopic boundary and microscopic irregularity, and even their local irregular parts affect properties and functions remain unclear and need to be assessed. For scale-up applications, the ordered structure is not sufficient. More importantly, unique property and function are essential to demonstrate that only COFs work for the purpose. In this regard, we are still far from meeting expectations of large-scale implementations.

In relation to energy conversion, COFs have shown considerable advantages in designing π-architectures via the merging of π-arrays with catalytic sites and built-in channels. Specifically, COFs’ polymerization under topology guidance enables the integration of various organic blocks into well-defined polymer backbones to achieve long-range-ordered structures, which opens an avenue to design catalysts for energy conversion. The precise alignment of each component and various linkages showcases the possibility of systematic engineering of HOMO and LUMO levels, band gaps, redox potentials, catalytic sites, and even diversity of different interfaces using one framework. This is the standpoint of COFs in exploring catalysts to promote energy conversion. Currently, we are at a stage to understand the effect of each structural parameter on the catalytic conversion process and face the challenges of how to make full use of the structural features to explore the framework so that each process is interlocked seamlessly. Toward this goal, more try-and-error and/or proof-of-concept investigations are necessary to gain broad and deep enough insights into how COFs control the processes during energy conversions.

Pertaining to photocatalytic conversions, the exploration of catalytic sites for reduction and oxidation of water into hydrogen and oxygen is a crucial issue. Although some examples of photocatalysis have shown the potential of using COFs other than Pt nanoparticles and Ir complex for these purposes, most cases focus on properties or performances without systematic identification of active sites and proven mechanisms, as well as catalytic cycles. To establish a real system of application other than replacing other materials with a COF, these fundamental issues need to be explored and addressed. In this sense, cutting-edge techniques are required to disclose the dynamics of each reaction step.

Regarding the photocatalytic reduction of carbon dioxide, noble metal-free and sacrificial electron donor-free systems are essential for implementations. For this purpose, exploration and identification of organic catalytic sites that promote carbon dioxide conversion and water oxidation are necessary, as they are two critical processes involved in the catalytic cycle. Currently, we are at a stage of using sacrificial electron donors to show the potential of COFs, which, however, cannot be deployed in real applications. Integrating organic catalytic sites into a π-framework so that the skeleton enables continuous flow of electron and hole is a direction worthy of further investigation in the future.

For the electrocatalytic reduction of oxygen, various carbon materials derived from COFs have been developed. Some of these materials have shown potential with performance, superior to Pt nanoparticles currently implemented in fuel cells. However, many issues and hurdles remain to be addressed relative to applications. Cost, stability, durability under harsh conditions, processing, efficiency, and power density must be investigated and tested in detail. Currently, less attention has been paid to these issues. Concerning this, most carbons made from COFs work under alkaline conditions. Developing COFs-based carbon materials that work under acidic conditions deserves further investigations.

Concerning the electrocatalytic reduction of carbon dioxide, COFs have shown a great potential of combining conductivity with catalytic activity to promote the transformation. The energy efficiency of the electrocatalytic reduction of carbon dioxide is unknown and needs to be evaluated. This point becomes a key factor in applications. Converting carbon dioxide into value-added chemicals such as methanol and ethylene deserves further investigations. For applications, a large-area electrode is necessary; preparing meter-scale COF electrodes without macroscopic and microscopic defects remains a challenging issue.

As scrutinized in this minireview, COFs emerged as a platform for designing organic materials with long-range structural order via direct polymerization, accompanied by in situ crystallization. This paves the way to produce a polymer with predesigned primary- and high-order structures. This attempt is possible as the polymerization system specifies covalent bonds and supramolecular interactions for the control of primary- and high-order structures, respectively. The covalent link guided by the topology diagram enables the spatial control over each step of chain propagation to form a well-defined primary-ordered structure, while the noncovalent interactions enable crystallization to create in a well-defined high-order structural manner. This structural control regime resembles those of biological polymers such as proteins and DNA in which primary- and high-order structures are determined by covalent and noncovalent bonds, respectively. Considering 100-year-long-pursued goals in synthetic macromolecules to achieve predesignable primary- and high-order structures, COFs represent a breakthrough in chemistry over such a long scientific journey. We envision that collaborations of chemistry, physics, materials science, and engineering would undoubtedly uncover a full picture of this class of amazing polymers. With the predesignable structures, the COFs’ field is up to our imagination.

Funding Information

This research was made possible due to a generous grant from the MOE Tier 1 grant (no. R-143-000-A71-114) and NUS start-up grant (no. R-143-000-A28-133).

Conflict of Interest

There is no conflict of interest to report.

Author Contribution

D.J. conceived the topic and structure. D.J. and S.T. wrote the manuscript.