To address the environmental challenges associated with excessive dependence on fossil fuel energy, various technologies for clean energy generation, storage, and utilization, such as fuel cells, electrolyzers, and batteries, have been developed.1–6 Among these technologies, anion exchange membrane fuel cells (AEMFC) are especially notable owing to their cost-effectiveness, as they do not require expensive platinum-group metal catalysts and are environmentally friendly.7,8 However, the working environments of AEMFCs, characterized by high pH, temperature, and humidity, pose challenges for organic membranes. In such conditions, hydroxides can deprotonate cations and hydrolyze the backbones, rendering the membranes less or non-conductive.9–11 Despite these challenges, significant advancements have been made in the field, including the development of ether-free poly(aryl piperidinium) (PAP) anion exchange membranes (AEMs) with high hydroxide conductivity, chemical stability, and mechanical strength,12 as well as perfluoroalkyl sidechains functionalized PAP AEMs.13 Furthermore, alkaline stable cations such as piperidiniums, imidazoliums, or ammoniums are commonly incorporated into ether-free polyethylenes or polyarylenes. These linear polymer-based AEMs offer versatility in design and thus demonstrate effective performance. However, a critical problem associated with linear polymer-based AEMs is that the ion-transporting pathways are often ill-defined. To address this issue, the phase separation technique has been used to establish effective ion-transporting pathways.14 Such a method suggests that the construction of clearly defined ion transporting pathways is a promising strategy to prepare highly conductive electrolytes or membranes for energy generation or storage devices.
Covalent organic frameworks (COFs), which are porous and crystalline polymers composed of light elements connected through covalent bonds, have recently emerged as effective ion transporting media because of their unique two-dimensional porous structures and stable skeletons resulting from covalent bonds.15,16 Since they were first reported by the Yaghi’s group in 2005,17 COFs have been extensively explored for various applications, such as energy storage,18 sensing,19 and water treatment.20 In particular, ionic COFs (iCOFs), in which ionic groups are functionalized along the walls or ionic monomers are used as the main building blocks, have garnered significant attention for energy-related applications, as the ionic moieties on the backbone can directly transport counter ions through the aligned pores. The first iCOFs, in which spiro-borate linked COFs were used to transport Li+, were reported in 2015 by the Zhang’s group.21 Since then, iCOFs have been explored for different roles in lithium-ion (Li+) batteries, including solid-state electrolytes designed for Li+ transportation22,23 and artificial solid electrolyte interphase layers for improving Li+ platting/stripping processes.24 Moreover, the unique and desirable structural properties of iCOFs are not limited to Li+ conduction.25–27 Various iCOFs have been studied for ion transport involving Na+,28,29 H+,30,33 Zn2+,31 OH–,32 34 and other ions. For example, He et al. developed a phase-transfer polymerization method to synthesize a series of quaternary ammonium COFs membranes with excellent hydroxide transportation performance.32 Moreover, Tao et al. reported a family of imidazolium-functionalized COFs and realized the transport of hydroxide anions based on a proton-exchange hopping mechanism through the one-dimensional (1D) pores of the COFs.27
Although several researchers have focused on cation (such as Li+ or H+) conductive COFs, research on anion transportation in COF materials is limited. Most of the extensively studied anion conductive structures are based on quaternary ammoniums and other N-based resonance-stabilized cations, such as imidazolium, guanidinium, and ethidium. In this study, we introduced a novel class of COFs that can efficiently transport hydroxides. In designing the hydroxide-conductive COFs, we focused on phosphoniums as cationic components, as they exhibit a high alkaline stability but have not been explored.35–37 An examination of the degradation mechanisms for cations (e.g., deprotonation, Hofmann elimination, nucleophilic ring opening, and nucleophilic addition to N-based resonance-stabilized cations) revealed that the high stability of phosphonium ions is attributable to the conjugation and electron-donating effects of the triphenyl groups. Moreover, the high steric hindrance of the methyl group prevents phosphine oxidation and shields the core phosphonium ions from hydroxide ion attacks. In fuel cell applications, it is necessary to install freestanding membranes between electrodes, and their mechanical stability and conformal contact are crucial factors that are often overlooked. Thus, we aimed to prepare freestanding phosphonium COF membranes.38 COFs synthesized using solvothermal methods typically exist as powders and require a polymer matrix to be formed into membranes. However, the addition of a non-ion conductive polymer matrix would lower the IEC, thus decreased conductivity. Thus, the fabrication of freestanding pure COF membranes is highly desirable.
In this study, we first pre-synthesized the quaternized phosphonium monomers with aldehyde functional groups and allowed them to react with p-phenylenediamine to prepare two-dimensional (2D) COFs (Fig. 1). However, using phosphonium monomers alone did not yield COFs with high crystallinity and surface areas, presumably because of the three-dimensional (3D) nature of the phosphonium monomers extending out of the plane. To address this problem, we introduced tris(4-formylphenyl)phosphine, a similar tri-fold flat symmetric monomer with a neutral charge, into the COF assembly. The optimized ratio of 6:4 for the phosphonium and phosphine monomers yielded freestanding membranes with a high OH– conductivity of 126.3 mS cm–1, and 84.6% of initial conductivity maintained for up to 2000 h in alkaline conditions at 80°C 1 M KOH. To facilitate practical utilization, we used a vapor-assisted method to fabricate freestanding films,39 which reduced grain boundaries at the meso- and macro-scale, thereby maximizing the transport properties achievable based on the designed chemical structure. Also, we conducted conventional solvothermal synthesis to prepare powder samples for comparison with the film samples.
The phosphonium COFs, OH–@P+-X-COFs, were synthesized using phosphorus-centered monomers, i.e., tris(4-formylphenyl)(methyl)phosphonium iodide and tris(4-formylphenyl)phosphine, (synthesized according to the details in SI Scheme S1,2; Fig. S1-4), reacted with p-phenylenediamine. The ratio of the charge-neutral and cationic monomers was modified to fabricate films with the desired properties. Using pure tris(4-formylphenyl)(methyl)phosphonium iodide for the condensation reactions did not yield freestanding membranes. Thus, the phosphonium ratio X was changed to 60, 40, and 20% to generate freestanding membranes (Fig. 1a). Our study focused on film samples rather than powder samples from solvothermal reactions, as they exhibited considerably improved conduction properties because of their well-organized layers and minimized grain boundaries and for the consideration of the fuel cell application. The thin film samples were prepared using a vapor-assisted planar self-assembly method (Fig. 1b). Monomers dissolved in solvents were evenly dispersed on a glass substrate (unless otherwise specified) placed inside a glass container. The mixed solvent (mesitylene:1,4-dioxane = 3:1 v:v) and catalyst (acetic acid) were added to the glass container, which was then sealed and placed in an oven at a temperature of 100°C for 24 h. The vaporized solvents containing the acids initiated the reaction of the monomers, resulting in self-assembled nanosheets that formed a thin film with a yield of 94% (Figs. 1b, c; See SI for the details). After the desired reaction times, the products were washed with acetone, ethanol, and N,N-dimethylformamide (DMF) and subjected to Soxhlet extraction with tetrahydrofuran (THF) for 24 h to remove the residual monomers and solvents trapped inside the frameworks. Notably, we obtained freestanding COFs membranes from the glass substrate by rapidly immersing the samples in a 0.5 M KOH solution (Fig. 1c). The resulting composites exhibited adequate mechanical strength to bear tension, compression, bending, and vibration and were chemically stable when exposed to organic or aqueous solvents. The pore alignment along the out-of-plane direction of the films likely promoted ion conduction with minimum resistance (Fig. 1d). For solvothermal synthesis, mesitylene:1,4-dioxane (3:1 vol.%) and 6 M aqueous acetic acid were heated at 120°C for 5 d, resulting in COF powders with a yield of 86%.
The crystalline structures of the OH–@P+-X-COFs were characterized through powder X-ray diffraction (PXRD) analyses (Figs. 2a, b). The experimental PXRD results exhibited diffraction peaks at 3.8, 5.8, 7.0, and 11.7° for the film samples, corresponding to the (100), (110), (020), and (111) reflections, respectively (Fig. 2a). Lattice modeling and Pawley refinement were performed using the Materials Studio software. The AA-stacking model optimized in terms of the geometry and crystal cell with the Forcite module in the software indicated a unit cell with dimensions of a = b = 26 Å, c = 2.6 Å, α = β = 90°, and γ = 120° in the space group of primitive rotational symmetry (P1). According to the comparison of the simulated and experimental PXRD patterns, the calculated AA-stacking model fitted the experimental profile in terms of the peak positions, with a weighted profile R-factor of 2.44% and an expected R-factor of 1.21%, suggesting the possibility of eclipsed AA-stacking of the charge-neutral OH–@P+-0-COFs (Fig. S5). The PXRD profiles of the OH–@P+-X-COFs (X = 20, 40, or 60) samples displayed nearly identical diffraction peak positions as the X = 0 samples, suggestive of similar crystalline structures. However, the peak intensity of the phosphonium COFs decreased with an increasing proportion of phosphonium monomers, X. This phenomenon likely occurred because the introduction of methyl-functionalized 3D monomers into the 2D skeleton disrupted the π–π stacking between layers and hindered the tight 2D packing of the layers. The simulated structures of the OH–@P+-X-COFs (X = 0, 20, 40, and 60) exhibited good agreement with the experimental PXRD profiles (Fig. S6).
Nitrogen adsorption–desorption isotherms of OH–@P+-X-COFs were obtained at 77 K to measure the porosity and surface area of the samples (Figs. 2c, d). The sorption data corresponded to a type I isotherm: rapid gas absorption occurred in the low-pressure range (P/P0 < 0.1), indicating microporous characteristics. The Brunauer–Emmett–Teller (BET) surface area values of the COFs, calculated from the isotherms, decreased with an increasing proportion of the ionic monomer: 624.2, 662.5, 413.2, and 392.4 m2 g–1 for X = 0, 20, 40, and 60, respectively. The pore size distributions calculated based on non-local density functional theory were 1.59, 1.55, 1.43, and 1.31 nm for X = 0, 20, 40, and 60, respectively (Figs. 2d, S7). The reduction in the BET surface area and pore size of the OH–@P+-X-COFs with the increasing proportion of the phosphonium monomer indicated that the methyl-functionalized phosphonium groups and counter ions on the backbone occupied the pore volume. Furthermore, the disturbance of the interlayer stacking by the 3D ionic monomer weakened the porosity of the frameworks. Compared with the powder samples, the thin film samples exhibited lower BET surface areas and pore sizes: 482.6, 431.1, 359.0, and 321.1 m2 g–1 and 1.52, 1.33, 1.12, and 1.08 nm for X = 0, 20, 40, and 60, respectively (Figs. 2c, S7).
The long-range ordered crystalline structures of the OH–@P+-X-COFs were further verified by scanning transmission electron microscopy images (Fig. S8). Large areas within the focus plane contained noticeable lattice patterns with different orientations. These highly ordered lattice fringes in different alignments clearly demonstrated the high crystallinity of the sample. All the measured d-spacing distance (2.6 Å) of the lattices matched well with the simulated model, providing structural evidence for the crystal structure of the material. However, because of the disruptive effect of the phosphonium monomer on the interlayer stacking, large-scale ordered stacking lattice fringes were not observed in the transmission electron microscopy images of the remaining samples (Fig. S8). Fourier transform infrared (FT-IR) spectroscopy, X-ray (XPS) photoelectron spectroscopy, and energy dispersive X-ray (EDS) spectroscopy techniques were used to study the chemical bonds, characteristic peaks, and compositions of the phosphonium COFs. The FT-IR spectra of the reactants and resulting COF samples were compared. The typical C = N stretching peaks at 1612 cm–1 confirmed the formation of imine bonds (Fig. S9). The absence of the N–H bond peaks around 3195–3409 cm–1 indicated that the p-phenylenediamine monomers were completely transformed and any excess monomers were thoroughly washed out. The XPS spectra confirmed the successful introduction of cationic phosphonium centers into the COF skeletons. New signals at binding energies of 618 and 630 ppm, corresponding to I 3d of I–, were attributable to the counter ion of the phosphonium centers (Figs. S10a, c, and d, Table S1-3). The XPS spectrum of OH–@P+-60-COFs demonstrated successful ionic functionalization and ion exchange (Fig. S10c). In particular, the signal pertaining to the iodide ion disappeared after ion exchange, indicating that I– was completely replaced by the hydroxide ion. The EDS mapping images highlighted the uniform distributions of P and I, indicating the homogeneous distribution of phosphorus cations in both the powder and thin film samples (Figs. S11–S16).
We studied the morphology growth over time of the OH–@P+-60-COF film samples (Fig. 3a). Four sets of parallel reactions were conducted in the same conditions and terminated sequentially at 6 h intervals, followed by scanning electron microscopy (SEM) imaging. The samples obtained in the initial stage (6 h) of thin film synthesis exhibited a spherical morphology similar to the powder samples. However, the particle size (diameters of 100–140 nm) was smaller than that of the powder samples. The smaller size could be explained by restricted crystal growth by the 2D substrates, owing to which the primary crystals could not continuously grow vertically but grew horizontally (12 h). As the reaction progressed, the micro-crystalline particles gradually accumulated and intertwined to form a continuous porous film (18 h). After 24 h, the surface of the formed COF thin film became dense and flat. The exfoliated freestanding OH–@P+-60-COF film samples were transferred to characterize their integrity and microscopic surface morphology (Figs. 2e, 3b–d). SEM images of the films obtained under different magnifications indicated an intact and defect-free surface morphology. Furthermore, the cross-section SEM images indicated uniform and dense internal structures (Fig. S17). The thickness of the OH–@P+-60-COFs, measured by the SEM images, was ca. 29.2 µm. By changing the number of monomers during synthesis, the thickness of the thin film sample could be reduced to a minimum of 300 nm, as indicated by the atomic force microscopy (AFM) 3D profile (Fig. 3d). Tapping mode AFM was performed to quantitatively analyze the surface flatness. A surface roughness of 1.56 nm was calculated from sample surfaces with an area of 3 µm2 (Figs. 2f, S18). In the powder form, the OH–@P+-X-COFs exhibited a spherical morphology. As the proportion of the phosphonium monomer increased, the surface roughness of the sample increased, likely because of the destruction of the orderly stacking of the layers by the phosphonium monomer (Figs. 3e–h). The thermal stability of the materials was evaluated through a thermogravimetric analysis. The materials were noted to remain stable up to 330°C without significant weight loss (Fig. S19).
Owing to their inherent 1D channels and alignment of cationic sites along the skeleton, the OH–@P+-60-COFs were expected to conduct OH– rapidly and efficiently. The OH– conductivity was evaluated by conducting electrochemical impedance spectroscopy measurements (Fig. 4). The OH–@P+-X-COF thin film samples were tested in ultrapure water (resistivity > 18.2 MΩ) after complete ion exchange and washing. The OH–@P+-60-COF thin film sample exhibited the best performance: conductivity values of 8.4 and 126.3 mS cm− 1 were recorded at 25 and 80 ºC, respectively, with an activation energy (Ea) of 0.16 eV (Figs. 4a and S20b, S21f). The X = 40 and 20 film samples exhibited conductivities of 5.2 and 55.1 mS cm− 1 at 25 ºC and 2.4 and 24.9 mS cm − 1 at 80 ºC, respectively. These values were significantly higher than those of the powder samples. The OH– conductivities for the OH–@P+-60-COF powder samples were 0.49 and 12.4 mS cm− 1 at 25 and 80 ºC, respectively, with an Ea of 0.23 (Figs. 4a and S20a, S21c). These values were three orders of magnitude lower than those of the film samples. The conductivities of the X = 40 and 20 powder samples were 0.44 and 10.3 mS cm− 1 at 25 ºC and 0.40 and 1.89 mS cm− 1 at 80 ºC, respectively. The Ea values of the three thin film samples were lower than those of the corresponding powder samples (Fig. S20a), indicating the lower energy barrier paths for anion transportation. Owing to the substrate-assisted nanosheet stacking, the freestanding COF thin film samples exhibited more closely stacked and continuous layers, as supported by the PXRD and BET surface area data. Because the powder pellet samples were prepared through cold pressing, they exhibited discontinuities, which limited their performance. Specifically, the randomness during the cold press process resulted in the microparticles stacking in irregular crystalline orientations, which hindered hydroxide conduction.
The number of cationic phosphonium functional groups inside the frameworks, which serve as hopping sites for anion conduction along the pore channel, was directly proportional to the anion conductivity. The ion exchange capacities (IECs) for the X = 20, 40, and 60 samples were 0.24, 0.61, and 1.07 mmol g–1 for the powder samples and 0.33, 0.76, and 1.17 mmol g–1 for the film samples, respectively (Fig. 4b). These values were positively correlated with the conductivity values of the samples. Notably, although the IECs of the X = 60 powder and films were similar (1.07 and 1.17 mmol g–1, respectively), their conductivity values were considerably different. This observation confirmed that the film samples with minimized grain boundaries are ideal for ion conduction. The swelling ratio and water uptake of the film samples were monitored at different temperatures. The three OH–@P+-X-COFs membranes (X = 20, 40, and 60) exhibited low swelling ratios (18.1, 16.9, and 22.2 wt.%, respectively) and low water uptake (4.8, 9.3, and 7.6 vol.%, respectively) at 80 ºC (Fig. 4c). These results were attributable to the rigid skeleton of the COFs, which provided excellent mechanical integrity, and the functionalized cationic groups and porous structure, which allowed the membranes to solvate the charge carriers, thereby enhancing the hydroxide transportation performance. Moreover, the structural and chemical stability of the COF skeletons enabled the OH–@P+-60-COF films to stably maintain high ionic conductivity in alkaline solutions at high temperatures over the long term. The initial conductivity remained over 84.6% after 2000 h of testing in 1 M KOH at 80°C (Fig. 4d). Compared with other recently reported hydroxide-conductive COFs and composite materials, the OH–@P+-60-COF film demonstrated high hydroxide conductivity, 126.3 mS cm–1, with low IEC, 1.17 mmol g–1 (Fig. 4e). The corresponding values for the COF-QA-2 samples were 211 mS cm–1 and 2.24 mmol g–1.32 These results demonstrated that the highly aligned nanopores in the proposed framework could conduct ions rapidly with minimum number of charge conduction moieties.
In summary, we designed and synthesized a novel class of cationic COFs for efficient hydroxide conduction. This study represents the first attempt at introducing phosphonium moieties into COF frameworks. Owing to their ordered and stable 1D pores, the phosphonium COFs provided ideal pathways for hydroxide transport. Furthermore, we prepared freestanding phosphonium COF films using a vapor-assisted synthesis method. The prepared films exhibited an exceptional hydroxide conductivity of 126 mS cm–1 under 80°C with a small IEC of 1.17 mmol g–1. The findings highlight the excellent potential of phosphonium ions as functional groups for AEMs. To promote the application of these COFs in fuel cells, our future work will be focused on enhancing the mechanical strength of the COFs while preserving their flexibility by exploring different synthesis methods and chemical structures.