Redox active covalent organic framework-based conductive nano ﬁ bers for ﬂ exible energy storage device

Covalent organic frameworks (COFs) constitute a family of crystalline porous polymers that are being studied for electrochemical energy storage. However, their low electrical conductivity and poor pro-cessability have largely limited their electrochemical performances and practical applications. Here, we develop an interfacial synthesis method to grow few-layered 2D redox-active COFs (DAAQ-TFP COF) on the surface of carboxylated carbon nanotubes (c-CNTs) in order to fabricate core-shell c-CNT@COF nano ﬁ bers, for which the thickness and the morphology of the COF nanolayers can be ﬁ nely controlled. When using the c-CNT@COFs as electrode material, the tailored nanostructure with high electrical conductivity allows ef ﬁ cient electron transfer, while the few-layered structure of the COF promotes fast electrolyte ion diffusion in the near-surface region, which results in an ef ﬁ cient utilization of the redox active sites in COF. More signi ﬁ cantly, c-CNT@COFs with nano ﬁ brous structure show good processability and can be assembled into freestanding and ﬂ exible nanopapers with the assistance of Cladophora cellulose. Given the good electrochemical performance and excellent ﬂ exibility, the nanopaper electrodes are assembled into ﬂ exible hybrid capacitors, showing high areal capacitance and extremely long life-time. This study provides a new pathway for the development of next generation sustainable and ﬂ exible energy storage devices based on COFs


Introduction
Covalent organic frameworks (COFs) are a new family of crystalline porous organic polymers constructed by covalently linking organic monomers in a periodic manner [1e8].Owing to their unique properties of long-range ordered nanopores and frameworks, high specific surface areas, and good physicochemical stability, COFs have been extensively studied for applications in gas storage and separation [9,10], catalysis [11e13], drug delivery [14], and sensing [15,16].As well, COFs are being investigated as electrode materials for supercapacitors [17,18], lithium/potassium-ion batteries [19,20], and zinc-air batteries [21].Their high surface areas facilitate adsorption of electrolyte ions on the electrode surfaces that could maximize double layer capacitances, while the ordered pore channels enable fast transport of electrolyte ions and thus increase the rate capability of the electrodes.More significantly, the synthetic diversity of COFs allows introducing redox-active moieties into the frameworks to achieve high theoretical capacities and high energy densities [22e27].
However, there are still significant challenges in the progress towards practical applications of COFs in electrochemical energy storage.Firstly, the intrinsically low electrical conductivity of traditional COFs hinders charge transfer in the frameworks and thus limits their electrochemical performances.For example, Dichtel et al. synthesized a b-ketoenamine-linked COF via polycondensation of 1,3,5-triformylphloroglucinol (TFP) and redoxactive 2,6-diaminoanthraquinone (DAAQ) monomers [22].Although DAAQ-TFP COF showed a higher capacity than the corresponding non-redox-active COF, only 2.5% of the redox-active sites were accessed in DAAQ-TFP COF probably because of the random orientation of the COF particles and the phase separation between the COF particles and the conductive additives in the electrode.Secondly, the 2D COF layers tend to pack closely due to their strong p-p interactions.The short distance between the layers (~3e4 Å) [11,22,26e28] results in the difficulty of electrolyte infiltration into the internal skeletons and thus leads to a low utilization of the active sites.Recent studies suggested that mechanical exfoliation of pristine 2D COFs into few-layer nanosheets could effectively accelerate the ion diffusion and facilitate the utilization of redox sites for lithium storage [24].However, the generalization of this method and the stability of COFs under the mechanical conditions are still unclear.Thirdly, the as-synthesized COFs are usually in the form of insoluble and infusible powders preventing them from being solution-or melt-processed in the way conventional polymers can be handled.Significant efforts have, however, been devoted to developing freestanding COF films by, for instance, self-assembly, interfacial polymerization, templating, and hybridization approaches [28e34].The resultant films usually show weak mechanical strength and poor flexibility.Obviously, the difficulty in processing COF crystals significantly limits their applications in electrochemical energy storage, especially for the uses in flexible devices.In this context, it would be very desirable to develop new strategies to design highly conductive, accessible and processable COFs or COF-based nanocomposites for the design of highperformance and flexible energy storage devices.
Herein, we report the controlled growth of few layers of redox active DAAQ-TFP COF on the surface of carboxylated multi-walled carbon nanotubes (c-CNTs) by interfacial synthesis.The obtained c-CNT@COF nanofibers showed a typical tube-type core-shell nanostructure where the morphology and thickness of the COF nanolayers can be finely controlled.The strong p-p interactions between the two components, COF nanolayers and conductive CNT backbones, not only stabilize their hybrid nanostructures, but also promote the electron transfer throughout the entire nanofibers.The nanolayer structure of the COF significantly increases the electrolyte ion diffusion and charge/discharge rate, especially at a high current density, leading to an efficient utilization of the redox active sites during the electrochemical processes.More importantly, the nanofibrous structure of the c-CNT@COFs greatly favors the materials processing allowing the formation of freestanding nanopaper with the assistance of cellulose nanofibers (CNFs) via a bottom-up interweaving approach.The hybrid capacitor assembled by the flexible c-CNT@COF/CNT/CNF and CNT/CNF nanopaper electrodes demonstrates excellent flexibility and foldability, high areal capacitance, good rate capability, and high cycle stability.

Synthesis of c-CNT@COF nanocomposites
c-CNT (Sigma-Aldrich; 100, 200, 300 mg for c-CNT@COF-1, -2, -3, respectively), DAAQ (Sigma-Aldrich; 0.9 mmol, 234 mg), PTSA (Sigma-Aldrich; 5.8 mmol, 1.0 g) were mixed in 1.0 mL water.The mixture was grinded for 30 min.Subsequently, TFP (Shanghai Tensus Bio-tech; 0.6 mmol, 126 mg) was added and the mixture was grinded for another 30 min.The obtained homogeneous slurry was knife cast on a glass plate and heated in an oven.The temperature was maintained at 60, 90 and 120 C for 24 h, respectively.The obtained solid was treated by Soxhlet extraction in methanol for 24 h to remove the PTSA and unreacted reagents.The purified nanocomposites were dried in an over at 75 C for 24 h.2.2.Fabrication of c-CNT@COF-3/CNT/CNF nanopaper c-CNT@COF-3 nanofiber (40 mg), multi-walled CNT (Sigma-Aldrich; 40 mg) and Cladophora cellulose (FMC Biopolymer, U.S.A; 16 mg) were dispersed in water (30 mL) and the mixture was dispersed by a probe sonicator.The homogeneous suspension was filtered on a Durapore® PVDF membrane filter (pore size: 0.45 mm, diameter: 9 cm) and thoroughly washed by deionized water.The obtained nanopaper was dried at 75 C for 12h.

Fabrication of the flexible hybrid capacitor
A piece of c-CNT@COF-3/CNT/CNF nanopaper (1 cm Â 2 cm) and a piece of CNT/CNF nanopaper (1 cm Â 2 cm) was used as the negative electrode and positive electrode, respectively.Two pieces of graphite paper were used as current collectors.A piece of filter paper was used as the separator.The separator was sandwiched between two working electrodes.The device was heat-sealed in a coffee-bag arrangement.

Result and discussion
The pure DAAQ-TFP COF was synthesized according to a previously reported method [28].The c-CNT@COF samples were prepared by stepwise reactions of c-CNT with the organic linkers via an interfacial synthesis method.As shown in Scheme 1, c-CNT, DAAQ, and p-toluene sulfonic acid (PTSA) were mixed in water.Thereafter, TFP was added and the mixture was grinded to form a homogeneous slurry, which was then heated at 60e120 C for 72 h in an oven.It should be noted that the use of c-CNT plays a key role for the synthesis of the nanocomposites, during which the carboxylic groups on c-CNTs reacted with amine groups on DAAQ to form amide linkages serving as anchors to connect c-CNTs and COF nanolayers for the construction of the chemically integrated c-CNT@COF nanofibers.PTSA was used as a solid-acid catalyst to promote the amide formation at the interfaces and the Schiff-base reactions of the monomers DAAQ and TFP for the COF synthesis.By varying the molar ratio of c-CNT and the monomers, three nanocomposites, named c-CNT@COF-1, -2, and -3, respectively, with controlled compositions and morphology were obtained.Thermogravimetric analysis indicated that the COF content of the nanocomposites was 76.6, 48.6 and 41.7 wt %, respectively (Fig. S1).
The crystalline structures of the pure COF and the c-CNT@COFs were confirmed by powder X-ray diffraction (XRD) analysis.As shown in Fig. 1a, all samples display a well-defined peak at 2q z 3.6 , corresponding to the 100 reflection planes of the COF.A weak and broad peak at 2q z 26.6 is observed for the pure COF and can be attributed to the pep stacking of the 001 planes.These results match well with the simulated pattern of the pure COF and previous reports [22].It should be noted that the 001 reflection peak of COF is overlapped with the 002 reflection peak of c-CNT at 26.6 for the c-CNT@COF samples.Hence, the intensity of this peak is observed to increase with the increasing c-CNT content.The formation of b-ketoenamine linkages in the pure COF was confirmed by solid-state 13 C nuclear magnetic resonance (NMR) spectroscopy (Fig. S2).The chemical shift at ~145 ppm can be assigned to enamine carbon (¼CNHÀ) while the signals observed at ~180 ppm correspond to carbonyl carbons (ÀC¼O) of b-ketoenamine and anthraquinone species.The chemical composition of the c-CNT@COFs was identified by infrared (IR) spectroscopy.As compared with the IR spectra of the starting materials, the emergence of the new peaks at ~1250 cm À1 observed in the c-CNT@COF-3 is attributed to the characteristic of CeN stretching confirming the formation of b-ketoenamine linkages.The disappearance of the NeH stretchings of DAAQ at 3427.4 and 3334.8 cm À1 indicates the complete consumption of the monomer (Fig. S3).Noteworthy, the position of the C¼O stretching band of c-CNT@COF-3 (1666.5 cm À1 ) located between those of pure COF (1672.2cm À1 ) and c-CNT (1660.7 cm À1 ), which can probably be attributed to the formation of amide (ÀC(¼O)NÀ) species at the COF/CNT interface of c-CNT@COF-3 [35].To gain insight into the bonding states at the interface, we performed X-ray photoelectron spectroscopy (XPS) studies as shown in Fig. 1b and S4.The high-resolution N1s XPS spectrum of pure COF can be deconvoluted into two peaks at 398.2 and 399.5 eV, which correspond to the N atoms in ArÀNH 2 and ArÀNHeC groups, respectively.Upon complexing COF with c-CNT, the bonding energy of N1s shifted towards higher energies.The spectrum of c-CNT@COF-3 can be deconvoluted into three peaks at 398.2, 399.5 and 400.5 eV.The latter new peak with a higher bonding energy can be assigned to the N atoms in amide linkages (ÀC(¼O)NÀ) at the interface formed by the reaction of the  carboxylic acid on CNT with the amine groups on DAAQ [36], indicating the change of the bonding mode of N atoms in c-CNT@COF.These results strongly support that the COF is chemically integrated on c-CNT.The porosity of the pure COF and the c-CNT@COFs was probed by N 2 sorption measurements at 77 K.All isotherms show steep uptake at low relative pressures (P/P 0 < 0.05), which is characteristic for microporous materials (Fig. 1c).Pore size distribution analysis from the adsorption isotherms indicated that the samples contain micropores with diameters of ~0.7 and 1.5 nm (Fig. 1d).After complexing COF with the less porous CNT, the obtained c-CNT@COFs showed smaller micropore volumes than that of the pure COF.Consistently, the BET surface area decreased from 1288.3 m 2 g À1 for the pure COF to 834.3, 709.8 and 576.7 m 2 g À1 , for c-CNT@COF-1, 2, 3, respectively.Remarkably, the isotherms of c-CNT@COFs show significant N 2 uptakes and hysteresis loops at relative pressures of P/P 0 > 0.5, suggesting the existence of mesopores that appear to be generated by the aggregation of hybrid c-CNT@COF nanofibers.
The pure COF shows a typical lamellar layer structure consisting of closely stacked flakes (Fig. 2aec).Obviously, the flakes tended to be converted into nanofibers by adding c-CNT during the synthesis, which led to the formation of the c-CNT@COFs nanocomposites assisted by a templating effect.The c-CNT@COF-1 with the lowest c-CNT content shows a mixed morphology containing both nanofibers and nanoflakes (Fig. 2def).By increasing the amount of c-CNTs that provided more anchoring sites and higher interfacial surface areas, c-CNT@COF-2 and -3 consisting mainly of nanofibers were eventually formed (Fig. 2g-n).Consistently, the diameter of the nanofibers decreased with increasing c-CNT content (Fig. S5).In addition, energy dispersive X-ray mapping analysis (carbon, nitrogen, and oxygen) shows the uniform distribution of the COFs on c-CNTs (Fig. S6).As mentioned above, the use of carboxylic acid functionalized CNT and the stepwise synthesis procedure are crucial for the formation of the homogenous nanofibrous structure.In contrast, by either using unmodified CNTs or a one-pot synthesis method resulted in a large amount of COF flakes and particles in the composites (Fig. S7).Noteworthy, highly dispersed single nanofibers can easily be obtained by sonication of the nanocomposites in water, which greatly facilitates the processing procedure afterward (Fig. S8).Fig. 2oeq shows the transmission electron microscope (TEM) images of dispersed c-CNT@COF-3 nanofibers, displaying typical tube-type core-shell structures, in which COF nanolayers compactly wrapped the c-CNT backbones and the contrasts at the interface between the core and shell are clearly observed.The nanofibers have an average diameter of ~30 nm and the COF nanolayers have an average thickness of ~10 nm.The high resolution TEM image shows the stacking of COF layers with interlayer spacing of ~0.35 nm (Fig. S9), which is agreement with the d spacing of the 001 planes of COF.These results demonstrate that the composition and morphology of the c-CNT@COF nanocomposites can be finely controlled at the nanoscale by the interfacial synthesis method, which allows for a fundamental understanding of their structure-property relationships and also opens up for regulation of their performances in electrochemical energy storage.
Owing to their conductive, porous, and nanofibrous structure and accessible redox-active units (Fig. 3a), c-CNT@COFs may be promising as electrode materials in electrochemical energy storage.The electrochemical performances of c-CNT@COFs were evaluated in a three-electrode system in an aqueous electrolyte (0.5 M H 2 SO 4 ).The working electrodes were prepared by mixing the c-CNT@COF (80 wt%) with CNT (10 wt%) and binder (10 wt% polyvinylidene fluoride (PVDF)), the mixture of which was coated onto a glassy carbon (GC) electrode.A piece of platinum foil was used as counter electrode and an Ag/AgCl electrode was used as reference electrode.The electrodes were first evaluated by cyclic voltammetry (CV) over a voltage range of À0.30e0.30V vs. Ag/AgCl at different scan rates.As shown in Fig. 3b, the CV curves of c-CNT@COFs recorded at a scan rate of 50 mV s À1 have symmetric redox peaks at 0 and À0.15 V vs. Ag/AgCl, suggesting that the c-CNT@COFs are battery-type materials.The redox reactions can be assigned to the reversible switching of anthraquinone (AQ) to anthrahydroquinone (AHQ) (Fig. 3a).There is a clear trend that the nanocomposites with higher content of c-CNT show stronger current responses because the enhanced conductivity results in more accessible redox active sites (Fig. 3b; Fig. S10).In order to reveal the influence of the nanostructure on the electrochemical performance, we prepared a c-CNT-COF electrode by physical mixing of pure COF (30 wt%) and c-CNT (70 wt%) and evaluated it by CV.Although the c-CNT-COF electrode has an almost identical composition to the c-CNT@COF-3 electrode, it displays a distinctly different CV curve (Fig. 3b).The c-CNT-COF electrode exhibits a rectangular-shaped CV curve with very modest redox peaks and the current is an order of magnitude smaller than that of the c-CNT@COF-3 electrode.Comparison of the nanostructures in SEM images gives an insight into the differences of their electrochemical performances.The c-CNT-COF electrode consists of phaseseparated COF flakes and c-CNT agglomerates (Fig. S11), which could cause sluggish electron transfer and concomitant slow ion diffusion preventing most of the redox active sites in the COF to be accessed during the electrochemical evaluation.In contrast, the compactly integrated nanofibers of c-CNT@COF-3 possess much higher interfacial area between the COF nanolayer and c-CNT backbone favouring efficient electron transfer along the nanofibers.As well, the fabricated thin COF nanolayers expose more redox active sites on the surface or near-surface region thereby shortening the ion transportation path.
In addition, the kinetics of the redox reactions in c-CNT@COFs were investigated by studying the dependency of the CV peak current (i p ) on the scan rate (v) according to the empirical expression of i p ¼ av b , where a and b are adjustable parameters [37].In general, there are two well-defined conditions with b ¼ 0.5 and 1.0, associated with a diffusion-limited redox process and a process occurring in a monolayer of adsorbed redox species without diffusion-limited mass transport, respectively.A value between 0.5 and 1.0 is more difficult to interpret and can entail a range of different situations including diffusion limitation in porous systems and/or redox reactions occurring on irregular surfaces.We plotted lg(i p ) from the CV curves as a function of lg(v) and calculated the b values by fitting the slopes (Fig. 3c; Fig. S12).A b value of 0.72 was obtained for c-CNT@COF-1, in which both c-CNT@COF nanofibers and COF flakes were observed, indicating that the redox process cannot be fully described by one of the well-defined conditions outlined above.With the disappearance of COF flakes and the formation of maily nanofibrous structures, higher b values of 0.96 and 1.00 were observed for c-CNT@COF-2, and 3, respectively.This result demonstrates that the redox reactions in these were prone to occur on surface or near-surface region [38] of the COF nanolayers in c-CNT@COF-2 and -3, without diffusion-limited mass transport, hence enabling fast charge/discharge responses and efficient utilization of the redox active sites.
Galvanostatic chargeÀdischarge (GCD) measurements were carried out to investigate the electrochemical behaviour and energy storage performance of the electrodes.The GCD curves were recorded at different current densities over a voltage range of À0.3e0.3V vs. Ag/AgCl.(Fig. 3d; Fig. S10).In accordance with the CV studies, the c-CNT@COF electrodes exhibit symmetric charge/ discharge plateaus at 0e0.15 V vs. Ag/AgCl in the GCD curves corresponding to the reversible protonation/deprotonation reactions taking place in the COF.However, the triangular GCD curves of the c-CNT-COF electrode, similar to those of the c-CNT electrode, represent a double layer charge and discharge mechanism while the Faradaic process in the COF is negligible.Therefore, the c-CNT-COF electrode showed a low capacitance of 52.3 F g À1 at 0.5 A g À1 , slightly higher than the value of the CNT electrode (49.2 F g À1 at 0.5 A g À1 ).Upon growth of COF nanolayers on c-CNT, the c-CNT@COF electrodes showed dramatically increased capacitances up to 418.7 F g À1 at 0.2 A g À1 (376.2F g À1 at 0.5 A g À1 ) (Fig. 3e, Table S1), comparable to state-of-the-art COF and polymer-based electrodes, such as TpPae(OH) 2 eCOF (416 F g À1 at 0.5 A g À1 ) [39], polyaminoanthraquinones (165e576 F g À1 at 1.0 A g À1 ) [40] (Table S2).It was calculated that 38.4,83.0, 99.1% of the redox active sites in the COF were utilized in the c-CNT@COF-1, 2, 3 electrode, respectively.In contrast, only 6.3% of the redox active sites were accessed in the c-CNT-COF electrode (Table S1).In addition, the rate capability of the c-CNT@COF electrodes was significantly increased with increasing c-CNT content.When the current density was increased from 0.2 to 10 A g À1 (Fig. 3e), the electrodes retained 18.0, 38.6, and 51.4% of its capacitance for c-CNT@COF-1, 2, 3, respectively.Meanwhile, the Coulombic efficiency of the electrodes was increased from 70% for c-CNT@COF-1 to ~100% for c-CNT@COF-2 and -3 (Fig. S13).
The electrochemical impedance spectroscopy (EIS) results indicate that the charge transfer resistances (R ct ) at the electrode/ electrolyte interface for c-CNT@COF-3 is 1e3 orders of magnitude smaller than the values for c-CNT@COF-2 and c-CNT@COF-1 (Fig. 3f; Fig. S14).In addition, the onset frequencies for a diffusionlimited response and for a semi-infinite Warburg diffusion response are much higher for c-CNT@COF-3 than the values for the Fig. 2. Scanning electron microscopy images of the pure COF (aec), c-CNT@COF-1 (def), c-CNT@COF-2 (gei) and c-CNT@COF-3 (jen); (oeq) transmission electron microscopy images of c-CNT@COF-3.c-CNT@COF-2, and -1 (Fig. S14).Apparently, the electrolyte ion diffusion proceeds much faster through the thin COF nanolayer and the nanofibrous, hierarchical porous structure of c-CNT@COF-3.Based on the above observations, we could conclude that the combined effects of efficient electron transfer, highly accessible charge storage sites, and fast ion transportation in the conductive and nanofibrous c-CNT@COF-3 led to an efficient utilization of the redox active sites in COF and thus high capacitances.
Given the nanofibrous structure and excellent electrochemical performances, we expect that the c-CNT@COF nanocomposites may find applications in flexible energy storage devices.Importantly, the nanofibrous structure of c-CNT@COFs is a great advantage for the materials processing.Cellulose nanofibers (CNFs), from a naturally abundant biopolymer with good processability and with rich organic functional groups, are ideal templates or substrates for the fabrication of freestanding nanocomposites.Cladophora cellulose, a type of CNFs extracted from algae, has been recently applied as flexible substrate to process both metal-organic frameworks (MOFs) and conducting polymers into freestanding films or aerogels for energy and environmental applications [41e45].In this context, we employed a bottom-up interweaving approach to fabricate c-CNT@COFs into freestanding nanopapers as flexible electrodes (Fig. 4a).Specifically, c-CNT@COF-3 nanofibers were mixed with CNTs and Cladophora cellulose with a mass ratio of 5:5:2 in water, followed by sonication to form a homogeneous suspension.Vacuum filtration of the suspension on a membrane filter formed freestanding and flexible c-CNT@COF-3/CNT/CNF nanopapers.Cladophora cellulose was used as a flexible substrate to knit and reinforce the c-CNT@COF-3 nanofibers, and the CNTs served as external conductive bridges.The XRD pattern confirms the existence of the crystalline COF and cellulose, and CNT in the nanopaper (Fig. S15).N 2 sorption analysis indicated that the nanopaper possessed a hierarchical micro-mesoporous structure with a relatively high surface area of 331.8 m 2 g À1 (Fig. S16).The micropores originated from the COF while the mesopores formed by aggregation of the nanofibers.It is clearly observed that the c-CNT@COF-3 nanofibers are interwoven by CNF and CNTs with the formation of a homogeneous and completely connected nanostructure (Fig. 4b).
Furthermore, a hybrid capacitor was fabricated by using the c-CNT@COF-3/CNT/CNF nanopaper and CNT/CNF nanopaper as the negative and positive electrode, respectively.A piece of filter paper that had been immersed in aqueous H 2 SO 4 solution (0.5 M) was used as the separator (Fig. 4a).The electrodes were initially tested in a three-electrode setup and the mass of the active materials in both electrodes was optimized to maintain charge balance, aiming at maximizing the capacitance and the operating potential window (Figs.S17 and S18).At the same scan rate of 10 mV s À1 , the negative and positive electrode show a potential window of À0.3 to 0.3 V vs Ag/AgCl and 0e1.2 V vs Ag/AgCl, respectively (Fig. 4c).Therefore, the device could reach a high operating potential window of up to 1.5 V.The electrochemical performance of the device was evaluated by CV and GCD measurements (Figs.S19 and 20).The CV curves were recorded in a potential window ranging from 0 to 0.6 V to 0e1.5 V (Fig. 4d).The redox peaks observed at 0.3 and 0.5 V can be attributed to the redox reactions of the COF component, while the near-rectangular shaped CV curves at high operating potentials illustrated the double layer capacitance of the CNT.The specific capacitance of the device was calculated from the GCD curves recorded at different current densities (Fig. 4e).Despite of the low COF loading of ~0.32 mg cm À2 , the device could reach a relatively high areal capacitance of 123.2 mF cm À2 at a current density of 0.2 mA cm À2 .In addition, it retained its 100% and 41% capacitance when the current density was increased by 5 and 100 times, respectively (Fig. 4f).The excellent rate capability can be attributed to the conductive feature and nanofibrous and hierarchical porous structure of the electrode materials that enable fast charge transfer and electrolyte ion diffusion during the charge/discharge processes.Given the high specific capacitance and broad operating window, the device displayed high energy density and power density values up to 30.7 mWh cm À2 and 591.9 mW cm À2 , respectively (Fig. S21).
Noteworthy, the calculated values of areal capacitance, energy, and power density are comparable to those of top-performing COFbased energy storage devices [25,46e48] and state-of-the-art flexible devices based on MOF and carbon materials [49e51] (Table S3).
In addition to the high capacitance, energy storage devices are required to have a long lifetime and good flexibility for their uses in flexible electronics.As shown in Fig. 4g, the device retained more than 94.0% of its original capacitance after 10000 GCD cycles at a current density of 10 mA cm À2 .Meanwhile, the CV curves of the device only showed minor changes during the cycling experiments.The redox peaks for the AQ/AHQ transformation in the COF were apparently observed after 10000 cycles.At a lower charge/ discharge current density of 1 mA cm À2 , it retained ~94.5% of the original capacitance after 1500 cycles (Fig. S22).The excellent cycling performance of the device can be attributed to the high chemical stability of the COF, CNF and CNT in the electrodes.The crystalline structure and the interwoven nanofibrous morphology of the electrode materials were retained after the cycling experiments (Figs.S23 and S24).More importantly, bending (90 ) or folding (180 ) the device has no significant influence on the CV curves (Fig. 4h).Furthermore, we assembled two devices in series that can power both red and green LED lights, even under the folded state (Fig. 4i).The excellent flexibility and foldability of the device can be attributed to the interwoven fibrous network of the nanopaper electrodes.

Conclusion
Freestanding and flexible nanopaper electrodes based on redox active COFs have been successfully fabricated by a bottom-up nanoengineering approach, which involves interfacial synthesis of c-CNT@COF nanofibers and interweaving of the hybrid nanofibers with nanocellulose and CNTs.An efficient utilization of the redox active sites of the COF can be achieved in the electrodes due to the synergetic effects of efficient charge transfer along the integrated conductive backbones and fast electrolyte ion diffusion/transportation through the thin COF nanolayers and the hierarchical porous structure of the c-CNT@COF nanofibers.For the first time, we fabricated a flexible and foldable hybrid capacitor by using the COF-based nanopaper electrodes that may open up new opportunities for the development of sustainable and flexible energy storage devices.Future studies may focus on the design of intrinsically electrical conductive and redox active COF materials that could avoid using external conductive substrates to design fully organic devices with high gravimetric/areal capacity and high energy density.More importantly, the transfer of the nanoengineering techniques developed in this study to COF processing may lead to the development of freestanding, flexible and sustainable COFbased nanomaterials for the promotion of their practical applications in energy storage, nanofiltration, sensing, heterogeneous catalysis, etc.

CRediT authorship contribution statement
Xueying Kong: performed the materials synthesis, electrochemical measurements, IR, TGA and BET characterizations.Shengyang Zhou: the TEM characterization.Maria Strømme: Formal analysis, Data curation, Writing -original draft, analyzed the electrochemical data, wrote the paper.Chao Xu: Writing -original draft, designed the project, did the XRD, XPS and SEM characterizations, wrote the paper.

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

Scheme 1 .
Scheme 1. Schematic of the synthesis of c-CNT@COF.

Fig. 1 .
Fig. 1.(a) Powder X-ray diffraction patterns of the pure c-COF and CNT@COFs; (b) high-resolution N1s X-ray photoelectron spectra of the pure COF and c-CNT@COF-3 and the deconvolution results; (c) N 2 adsorption and desorption isotherms recorded at 77 K; (d) pore size distributions of the pure COF and c-CNT@COFs calculated from the adsorption isotherms.(A colour version of this figure can be viewed online.)

Fig. 3 .
Fig. 3. (a) The reversible quinine to hydroquinone transformation in DAAQ-TFP COF showing its redox (charge/discharge) mechanism; (b) comparison of cyclic voltammetry curves of the c-CNT, c-CNT@COFs and c-CNT-COF electrodes at a same scan rate of 50 mV s À1 ; (c) plots of lg (i p ) versus log (n) for c-CNT@COFs electrodes, where i p is the peak current and n is the scan rate; (d) comparison of galvanostatic chargeÀdischarge (GCD) curves and (e) gravimetric capacitances of the c-CNT, c-CNT-COF and c-CNT@COFs electrodes.The capacitances were calculated from the GCD curves; (f) electrochemical impedance spectra of the c-CNT@COFs electrodes.(A colour version of this figure can be viewed online.)

Fig. 4 .
Fig. 4. (a) Illustration of the flexible c-CNT@COF-3/CNT/CNF nanaopaper and the assembled hybrid capacitor; (b) the SEM image of c-CNT@COF-3/CNT/CNF nanaopaper; (c) cyclic voltammetry (CV) curves of the c-CNT@COF-3/CNT/CNF and CNT/CNF nanopaper electrodes recorded in a three-electrode setup at a scan rate of 10 mV s À1 ; (d) CV curves of the hybrid capacitor at different potential windows at the same scan rate of 50 mV s À1 ; (e) galvanostatic chargeÀdischarge (GCD) curves of device at different current densities; (f) areal capacitance and rate performance of the device; (g) cycling performance of device (the inset shows the CV curves of the device before and after the cycling experiment) (h) CV curves of device under different deformations; (i) photo of red and green LEDs powered by the flat or folded devices.(A colour version of this figure can be viewed online.)