The synthesis of a biphenyl-linked borate-based covalent organic framework (BPB-COF) was achieved in a two-step process from lithium borohydride as simultaneous boron and lithium source with 4,4’-biphenol as linker. First, the phenol groups of the linker are deprotonated irreversibly by lithium borohydride, releasing hydrogen gas as a by-product for the formation of a biphenyl-linked borate polymer (BPB-Poly). Since the polymerization occurs in a short time span, first, a kinetically controlled amorphous polymer network is generated (BPB-Poly, Fig. 1b). Following the concept of dynamic covalent chemistry13, a transformation to the crystalline compound requires the cleavage and reformation of the B–O bonds present, enabling the conversion of the kinetic and amorphous material to the thermodynamically favoured crystalline state. This can be achieved through elongated reaction times, high pressure, and high temperatures. Therefore, the resulting polymer-solvent mixture was transferred to a Teflon-lined steel autoclave and 1 equivalent triethylamine was added to the suspension in order to deprotonate any residual phenol groups and assist as a competing nucleophile in the cleavage and reformation of the B–O bond. The autoclave was sealed shut and placed in the oven at 150 °C for at least 2 days (Supplementary Table 1). After cooling, the resulting white precipitate was filtered under an inert atmosphere, washed with dry tetrahydrofuran (THF) and dried at 100 °C under vacuum to yield BPB-COF (Fig. 1b). Due to the susceptibility of BPB-COF towards moisture it was handled and stored under inert atmosphere (Supplementary Fig. 6). For characterization purposes lithium tetraphenoxyborate (TPB) was synthesized from lithium borohydride and phenol as model compound (Fig. 1a).Fig. 1: Synthetic strategy and compositional analysis.a Synthesis of tetraphenoxyborate (TBP) as a model compound. b Synthesis of BPB-Poly and conversion to BPB-COF. c FT-IR spectra of BPB-COF (blue) and BPB-Poly (red) along with the model compound TPB (black) and linker (cyan). d 11B MAS NMR of the model compound TPB (black), BPB-COF (blue), and BPB-Poly (red). e 7Li MAS NMR showing the presence of lithium ions offsetting the anionic net charges. f One-batch synthesis yielding 1.2 g BPB-COF.In the Fourier-transform infrared (FT-IR) spectrum, the B–O stretching mode at 963 cm−1 shows the successful linkage formation, together with the absence of the O–H stretching band at 3359 cm−1, indicating the full conversion of the 4,4’-biphenol (Fig. 1c). The incorporation of 4,4’-biphenol in both networks is confirmed through 13C cross-polarization magic-angle spinning nuclear magnetic resonance (CP-MAS NMR) (Supplementary Fig. 1). For BPB-Poly, residual THF can be seen within the network in contrast to BPB-COF, which is absent of a 13C signal that can be assigned to THF (Supplementary Fig. 1)14.To verify the presence of the tetratopic [BO4]− species, the materials were investigated by recording directly excited 11B MAS NMR spectra (Fig. 1d). In order to confirm the tetratopic borate species of BPB-Poly and BPB-COF, the chemical shift was compared to the molecular tetraphenoxyborate species of the model compound TPB. The tetraphenoxyborate species was found at a chemical shift of 1.70 ppm, which is consistent with values for molecular tetratopic borates found in literature15. The 11B MAS NMR spectra of BPB-Poly were in accordance with a resonance signal at 1.87 ppm. In addition to the desired signal of the tetrahedral [BO4]− species, the quadrupolar pattern signal can be attributed to trigonal [BO3] in the range of 0–20 ppm, evidencing the presence of defects within BPB-Poly16. After solvothermal treatment, the attenuation of the quadrupolar [BO3] signal can be observed, with the signal for the desired [BO4]− at δ = 2.3 ppm (Fig. 1d). The change in chemical shift suggests that while both materials are constructed from the desired tetratopic borate nodes, the ligand to boron interactions are slightly altered in the crystalline phase.The negative backbone charge of the networks is balanced by lithium counter ions, which can be evidenced through the directly excited 7Li MAS NMR spectrum. The 7Li resonance at 0.1 ppm observed for BPB-Poly is typical for solvated lithium ions, due to residual THF present within the network. Upon crystallization, the 7Li resonance signal of BPB-COF is found to shift to 1.09 ppm, evidencing a changed environment for the lithium ions (Fig. 1e). The quantitative amounts of boron and lithium determined through inductively coupled plasma optical emission spectrometry (ICP-OES) are in accordance with the theoretical amount. These are, to our knowledge, the highest boron and lithium contents reported for borate-based COFs (Calc. B, 2.80; Li, 1.80. Found: B, 2.78 ± 0.12; Li, 1.64 ± 0.07, Supplementary Fig. 24 and Supplementary Table 2).While for BPB-Poly no long-range order is detected through powder x-ray diffraction (PXRD) analysis, through solvothermal treatment, the polymer could be successfully converted to the crystalline BPB-COF (Fig. 2a). By prolonging the duration of the treatment an increase in crystallinity is observed in the diffractogram through the appearance of additional diffraction peaks at low 2θ values combined with a reduced full width at half maximum (FWHM), indicative of an increase in crystallite size (Fig. 2a)17.Fig. 2: Structural characterization of BPB-COF.a Collected PXRD pattern for BPB-Poly (red) and BPB-COF after 2 (black), 5 (purple), and 12 (blue) days. b Experimental PXRD pattern (red), obtained profile after Rietveld refinement (grey), the difference between the experimental and refined profile (blue) and positions of observed diffractions (green). c Two-dimensional sql structure viewed from the c-plane. d Structure of BPB-COF viewed from the b-plane showing the interlayer lithium ions (yellow) enabling pillaring through ionic interactions and highlighted diagonal in-plane oriented linker. e Unit cell of BPB-COF. f Nitrogen sorption isotherms for the amorphous BPB-Poly (red) and crystalline BPB-COF after 2 (black), 5 (purple), and 12 (blue) days of solvothermal treatment showing a similar uptake at low-relative pressures. g Pore size distribution with inset depicting the pore and calculated pore size.After the subtraction of the background (Supplementary Fig. 2), the PXRD pattern was indexed on a primitive tetragonal unit cell (P-lattice) with lattice constants of a = b = 16.7557 Å and c = 5.0126 Å. As the anionic framework is based on tetrahedral [BO4]− sites reticulated by linear biphenyl units, the formation of a three-dimensional (3D) net with diamond (dia) topology was assumed, as the dia-net is a regular 4-connected (4-c) net based on tetrahedral nodes. Therefore, we assumed BPB-COF should crystallize in a single (dia) or, most likely, intergrown diamond net (dia-cN, with N defined as the number of interpenetrating net components)18. However, no feasible result could be derived from the general formula that establishes the interpenetration degree of dia-nets (Supplementary Fig. 3)19. Moreover, fitting dia-nets with various interpenetration degrees within the indexed cell parameters was in poor accordance with the experimental PXRD pattern (Supplementary Fig. 5) due to the absence of the characteristic diffractions at 2θ = 14.9 and 17.6, leading to the investigation of other topologies. Interestingly, a two-dimensional square-lattice (sql) net was found to fit well within the indexed cell parameters and was in very good agreement with the experimental pattern (Fig. 2b, Supplementary Figs. 4 and 5). Both sql and dia are regular nets based on 4-c vertices, however, while the dia topology is based on tetrahedral vertices, the sql-net is constructed from planar rectangular building units resulting in a 2-periodic net. So far, COFs with a sql topology have been constructed from 4-coordinated rectangular building-units. Although the tetratopic borate is a tetrahedral building block, the sql topology is enabled due to the flexibility of the B–O–C linkages around the borate unit, allowing the 4 points of extensions of the node to be coplanar and thus form such a layered structure (Fig. 2d).As two-dimensional COFs are based on 2-periodic nets, non-covalent interactions such as π–π interactions and hydrogen bonding assist in the construction of the framework through stacking of the layers. For our sql-based model, this is unrealistic, due to the torsion between the phenyl rings and the biphenyl linker oriented diagonally in the cell. The interlayer distance of 5.016 Å is also found to be much larger than common values for two-dimensional COFs and outside the range for significant contributions through π–π interactions3,5,20. Furthermore, the negative charge of the borates and the additional electrostatic repulsion of the layers caused by the anionic net charge of the framework further refutes the possibility of layers stacked by π–π interactions. Instead, a pillaring of the layers is driven by electrostatic interactions between the lithium counter ion located in the interlayer space and the phenoxy moieties from two different layers. The coordination is based on the formation of LiO4 tetrahedra leading to the effective charge compensation of the framework (Fig. 2d). To further validate the environment of the lithium ions present in BPB-COF, we conducted first principles quantum mechanical NMR calculations to determine the theoretical chemical shielding. Using CASTEP, the isotropic chemical shift for the lithium-ion of BPB-COF was calculated to be 1.19 ppm (Supplementary Fig. 25)21,22. This is in accordance with the experimentally measured value of 1.09 ppm further supporting the tetrahedral coordination in the interlayer space. It is worth noting that the coordination of the counter-cation by the phenoxy groups of the borate was also observed in molecular tetraphenoxyborates and even found to act as structural support, linking molecular borate subunits to infinite one-dimensional polymeric structures similar to what is observed for the structure of BPB-COF23,24. Consequently, the lithium ions become structurally and topologically relevant, with the final topology being best described by a pillared sql-net (Fig. 2c). In a series of Rietveld refinements25,26 where not only the Rwp and Rp values but also the potential energy of the structure was considered (applying the universal force field27) a final model with good Rwp and Rp values of 6.51% and 3.39% together with sensible bond lengths of O–Li = 1.978 Å and O–B = 1.493 Å could be obtained (Fig. 2e, final cell parameters: P-4, No. 81, a = b = 16.729 Å and c = 5.016 Å, boron (Wyckoff positions 1b and 1d), lithium (Wyckoff positions 1a and 1c))28,29.Since both borate-based materials displayed good thermal stability as evidenced by thermogravimetric analysis (TGA, Supporting Fig. 9), the porosity and sorption capacities were assessed on samples activated at 150 °C under high vacuum (Supplementary Fig. 7). While BPB-Poly shows no significant uptake of nitrogen (N2), a remarkable increase in microporosity and N2 uptake is found for the BPB-COFs (Fig. 2f). Using the BETSI tool30, after 2, 5, and 12 days of solvothermal treatment, respective surface areas of 802, 784 and 819 m2 g−1 were calculated with a pore size distribution centering around 0.61–0.64 nm (Fig. 2g, Supplementary Figs. 10 and 11). Notably, while the elongation of the solvothermal treatment results in an improved crystallinity (Fig. 2a), no significant enhancement in surface area could be detected through N2 sorption analysis (Fig. 2f, inset). For partial pressures >0.5p/p0 (Fig. 2f), an increasing hysteresis loop was observed for more crystalline samples. Through investigation with scanning and transmission electron microscopy (SEM and TEM), the emerging hysteresis was found to be consistent with crystallite length and can be attributed to the additional meso/macroporosity created through the interstices of the agglomerated crystallites (Fig. 3).Fig. 3: Morphological change of the particles throughout the crystallization.SEM images of BPB-Poly and BPB-COF after 2, 5, and 12 days of solvothermal treatment show the crystallization process of the bulk material and the growth of rod-shaped crystallites on the surface alongside a model depiction of the particles. For BPB-COF (12 days), a broken particle was chosen to highlight the hollow nature of the sphere. TEM images of BPB-Poly and BPB-COF (2, 5, and 12 days) show the conversion from dense and diffuse to hollow spherical particles with increasing surface-crystallite length. Scale bars set to 200 nm.Starting with BPB-Poly, dense agglomerated spheres with an irregular surface are obtained (Fig. 3, 0 days, Supplementary Figs. 12 and 17). After 2 days of solvothermal treatment, the particles are found to retain their dense nature with a new uniform surface (Fig. 3, 2 days, Supplementary Figs. 13 and 18). Prolonging the solvothermal treatment led to continuous recrystallization of the material resulting in the hollowing of the spherical particles and growth of crystallites on the surface (Fig. 3, 5 days, Supplementary Figs. 14 and 19). Further treatment resulted in the elongation of these crystallites, as the process is driven by thermodynamics to minimize the overall free energy of the system, with larger crystals having a minimized surface energy. Due to this process, the bulk phase is further dissolved to recrystallize on the already formed crystallites located at the particle surface. This led to the final morphology of hollow spheres covered with agglomerated rectangular cuboid crystallites in the 100–200 nm range (Fig. 3, 12 days, Supplementary Figs. 15 and 20).Having the rectangular cuboid crystallites of BPB-COF, we hypothesized the exfoliation of crystallites in a top-down approach. As the layers are comprised of covalent bonds, disruption of the weaker ionic interactions pillaring the layers should yield borate-based covalent organic nanosheets (BPB-NS). We also hypothesized that in order to achieve a chemical exfoliation the solvent should have an affinity for the lithium ion in order to compete with the ionic interactions between the phenoxy moieties and the lithium ions. While ultrasonication in THF proved to be unsuccessful, instead the exfoliation could be achieved by using dry methanol. First, BPB-COF (12 days) was suspended in dry methanol, followed by sonication at 35 kHz for 30 minutes. After centrifugation to settle bigger particles and fragments, the desired BPB-NS were left natant in solution (Fig. 4a). For the analysis of the BPB-NS, the solution was simply dropped and cast onto the respective substrates. Through TEM, the obtained BPB-NS were found to have the expected dimensions of the prior observed rectangular cuboid crystallites of BPB-COF (12 days) in the approximate range of 50 × 50 nm (Fig. 4c and d, Supplementary Figs. 21 and 22). The thickness of the exfoliated layers was investigated through atomic force microscopy (AFM) in AC mode, where the BPB-NS were found to consistently exfoliate < 1 nm thickness (Fig. 4b, Supplementary Fig. 22). Additionally, BPB-NS could be clearly distinguished in the AFM phase image due to the difference in roughness compared to the SiO2 substrate (Fig. 4f).Fig. 4: Exfoliation and characterization of nanosheets.a SEM image of BPB-COF (12 days) highlighting the crystallites and model of the exfoliation, the scale bar is 100 nm. b Height profile measured by AFM along the three lines shown in Fig. 4e. c TEM image giving an overview of BPB-NS, scale bar is 500 nm. d Close-up of layered BPB-NS, scale bar is 50 nm. e AFM topography image of exfoliated BPB-NS and marked height profiles, scale bar is 400 nm. f AFM phase image with BPB-NS clearly distinguished through surface roughness, scale bar is 400 nm.Lastly, to determine the ionic conductivity of BPB-Poly and BPB-COF (12 days) electrochemical impedance spectroscopy was carried out on a tape-cast film. The tape-cast films were prepared under an argon atmosphere by mixing the active material powder and polyvinylidene fluoride in a 9:1 weight ratio in anhydrous N-methyl-2-pyrrolidone (Supplementary Fig. 23). After casting on a carbon-coated aluminium foil and drying, the respective thin-films were pouched and assembled in a coin cell, sandwiched between two stainless steel spacers. An anhydrous propylene carbonate solvent was dropped onto both thin films of BPB-Poly and BPB-COF electrodes for activation purposes before being assembled into a coin cell. Using ambient Nyquist Plots, the fitted R1 value (Fig. 5a) represented the resistivity and was used to calculate the ionic conductivity. The ionic conductivity of BPB-Poly is 3.6 ± 0.5 × 10−6 S cm−1, while BPB-COF displayed an approximately 10-fold increase in conductivity of 3.1 ± 0.3 × 10−5 S cm−1 (Fig. 5b). The ionic conductivity values of both materials were found to increase with temperature from 20 to 60 °C, which indicated the hopping mechanism of lithium-ion transport in the materials. Moreover, the activation energy of lithium-ion transport in BPB-Poly was higher compared to BPB-COF (Fig. 5c), which indicates that the crystalline framework structure offers a regulated diffusion channel that lowers the hopping energy compared to the amorphous network underlining the beneficial influence of crystallinity on conductivity.Fig. 5: Electrochemical measurements comparing BPB-Poly and BPB-COF.a Room-temperature Nyquist plot (normalized by the ration of the area to thickness with the equivalent circuit model. b Temperature-dependent ionic conductivity and c Arrhenius character and activation energy of BPB-Poly and BPB-COF.In conclusion, we have developed a strategy that allows the construction of a charged covalent organic framework from tetratopic borate nodes connected by biphenyl units using lithium borohydride and 4,4’-biphenol as reactants. The synthesis was achieved in a two-step process of polymerization and crystallization. The two-dimensional framework was found to crystallize in a sql-net not relying on conventional π–π stacking but pillaring enabled by the countercation connecting the anionic framework layers through ionic interactions. The crystallization process was investigated by PXRD, SEM and TEM, showing the transformation from dense spheres to a microporous crystalline material with hollow spheres of agglomerated rod-shaped crystallites. The crystallites were exfoliated into <1 nm high rectangular nanosheets by sonication in methanol. Through EIS the influence of crystallinity on lithium-ion conductivity was investigated, showing a conductivity increase by approximately 10-fold when compared to the amorphous material.
