Construction and characterization of 2DWPN-1BDBB was synthesized by employing dehydration between commercially available 1,4-phenylenediboronic acid and catechol. Subsequently, the 2DWPN labelled 2DWPN-1 was obtained by heating a colourless mixture of BDBB and commercially available BPE in methylbenzene (PhMe) at 90 °C for 2 h and slowly cooling the reaction mixture to room temperature (Fig. 1a and Supplementary Figs. 1–9). An optical microscopic image revealed that centimetre-scale yellow single crystals of 2DWPN-1 were obtained with a schistose structure (Fig. 1b).Fig. 1: Illustration of the 2DWPN labelled 2DWPN-1.a, Synthesis of 2DWPN-1. b, Optical microscopic images of its single crystals. c,d, Scanning electron microscopy (c) and AFM (d) images of 2DWPN-1 crystal cross-section. e, Top view of a monolayer of the crystal packing in 2DWPN-1. The independent chains are distinguished by different colours. Solvent molecules, H atoms and catechol groups were removed for the sake of presenting a clearer woven display. The inset shows a graphical representation of the biaxial weaving structure in 2DWPN-1. f, View along the b axis of the crystal packing in 2DWPN-1; here PhMe molecules (yellow) were retained between layers and H atoms were removed for the sake of providing a clearer display. g, View along the a axis of the crystal packing in 2DWPN-1; all solvent molecules and H atoms were removed for the sake of clarity.The micromorphology of a 2DWPN-1 crystal was examined using field-emission scanning electron microscopy. The macroscopic schistose crystal is composed of molecule-thick sheets as a result of well-organized stacking (Fig. 1c). Furthermore, atomic force microscopy (AFM) was employed to illustrate the structure associated with the accumulation. We found that the thickness of each layer is approximately 1.3 nm (Supplementary Fig. 8) and the arrangement between layers is loose, a situation that favours mechanical exfoliation to obtain monolayers (Fig. 1d). In addition, single-crystal X-ray diffraction was employed to directly observe precise molecular structure information regarding 2DWPN-1, which crystallizes in the orthorhombic Pbca space group with lattice constants a = 23.4671(6), b = 19.6536(5) and c = 28.7688(5) Å (Supplementary Table 1). Further analysis revealed that BDBB molecules are linked by dative B–N bonds to BPE molecules to form 1D helical chains.Density functional theory (DFT) calculations based on the crystal data for 2DWPN-1 were used to analyse the nature of the dative B–N bonds. The large electron localization function value between the B and N atoms indicates (Supplementary Figs. 10–12) the formation of dative B–N bonds with strong bond energies of 150.7 kJ mol–1 (Supplementary Table 7). The dative B–N bonds provide excellent stability to 2DWPN-1 and serve as the foundation for its ability to be exfoliated and applied in some scenarios. From the top view of the crystal structure (c axis), a 2D polymer network with a two-over and two-under weaving pattern is generated through the entanglement of the 1D helical threads. For clarity, hydrogen atoms and catechol groups, which do not affect the main chain topology, are concealed. Distinct colours were assigned to molecular chains, facilitating an intuitive representation of the intricate woven topology within the crystal structure of 2DWPN-1 (Fig. 1e and Supplementary Fig. 13). Notably, the woven nodes in the 2D networks are formed (Supplementary Fig. 14) by the crossover of a pair of double bonds in the BPE moieties in the molecular warp and weft, with a plane-to-plane distance of 3.79 Å and a cross angle of 42° between two BPE moieties. These covalent entanglements exert the influence of the woven topology upon the mechanical properties of these self-standing networks. Such a 2D woven monolayer can stack to form a final 3D multilayer structure, although the filling of PhMe solvent molecules between these woven layers prevents them from being densely packed (Fig. 1f). Furthermore, given their distribution between layers, these solvent molecules have no impact on the freedom of movement associated with chain segments within the woven layers. Importantly, when the solvent is removed at 120 °C, the stacking superstructure is disrupted, causing the crystals to lose their crystallinity (Supplementary Fig. 15). A side-on view (bc planes; Fig. 1g) reveals an anti-direction arrangement mode between adjacent layers in which the thickness of the 2D woven monolayer is 13.2 Å from direct measurement of the single-crystal structure, a distance that is consistent with the observed results from AFM measurements.Mechanism of topology formationTo gain a deeper insight into the mechanism of formation of the woven topology in the case of this polymer network, we synthesized, as a means of comparison (Supplementary Figs. 16–21), a non-woven polymer network (NWPN-1) in m-xylene from the same starting materials and used the same method of preparation as that used in the production of 2DWPN-1. In essence, the synthetic conditions employed in making 2DWPN-1 and NWPN-1 differ only in the nature of the solvent, suggesting that it plays a key role in the formation of these polymer topologies. Single-crystal X-ray diffraction analysis showed that each asymmetric unit of NWPN-1 contains only one m-xylene molecule. We also investigated the interactions between the solvent molecules and the polymer network, showing (Supplementary Fig. 22) the existence of C–H···π interactions (3.26−3.27 Å) between a m-xylene molecule and the four catechol groups located on different polymer main chains. The conformation of each polymer chain (Fig. 2a) is not affected by the solvent molecules in NWPN-1, and each dative B–N bond in the chains has the same bond length of 1.67 Å. We can see clearly (Fig. 2b) that the symmetrical distribution of electron clouds and solvent molecules has no effect on the electron cloud distribution associated with the polymer chains, by inspection of the electrostatic potential map. Therefore, the polymer chains in NWPN-1 exhibit a classical extended zigzag conformation (Fig. 2c). Finally, the zigzag polymer chains of NWPN-1 adopt a grillage accumulation as a consequence of the balanced interactions of each m-xylene residue with its four adjacent polymer chains (Supplementary Figs. 23 and 24).Fig. 2: Mechanism of topology formation in NWPN-1 and 2DWPN-1.a, Illustration of the interactions between the solvent molecules and the polymer chains in NWPN-1, and the lengths of the dative B–N bonds in NWPN-1. b, Surface electrostatic potentials of NWPN-1. c, The conformation of a polymer chain in NWPN-1. d, The interaction between solvent molecules (Toluene-1 and Toluene-2) and molecular chains, the lengths of dative B–N bonds and the dihedral angle in a unit of 2DWPN-1. e, Surface electrostatic potentials of 2DWPN-1. f, The helical conformation of the polymer chains in 2DWPN-1. g, The woven nodes of the polymer chains in 2DWPN-1.In the case of 2DWPN-1 obtained in PhMe, an asymmetric unit consists of two BDBB moieties, two BPE moieties and two PhMe molecules (Toluene-1 and Toluene-2). In the formation of the 2D woven topology, the two PhMe molecules are thought to play different roles. Specifically, Toluene-1 forms a π···π stacking interaction (3.66 Å) with BPE-1 (Fig. 2d). As a result, the bond energies of the two related dative B–N bonds in BPE-1, with bond lengths of 1.65 and 1.66 Å, are enhanced on account of their higher electron densities. The interaction between Toluene-2 and BPE-2 is weaker. Accordingly, the two related dative B–N bonds possess longer bond lengths of 1.68 and 1.67 Å. Based on electrostatic potential analysis, Toluene-1 enters into space conjugation with BPE-1, resulting in the enhancement of the electron cloud density around BPE-1. On the contrary, Toluene-2 has no obvious space conjugation with BPE-2 and the electron cloud around BPE-2 exhibits no obvious change (Fig. 2e). These asymmetric interactions result in a dihedral angle of 53° between the B1–B4–B3 and B2–B3–B4 planes related to two BPE units and one BDBB moiety during the polymerization (Fig. 2d). The repeated extension of the twisted fragments results in helical polymer chains (Fig. 2f). Interestingly, each helical chain is composed of two helical segments with the same helical pitch of 23.4 Å that are arranged alternately in opposite helical directions.The adjacent reverse spirals make up two alternating bays with up and down openings, providing the over-and-under stereoscopic space for the weaving of warps and wefts (Supplementary Figs. 25–30) guided by the π···π stacking interactions between the BPE moieties in different threads. Each bay contains two BPE moieties, allowing two polymer chains to cross by an over-or-under space to form a typical biaxial woven topology (Fig. 2g). Therefore, the PhMe molecules change the conformation of the polymer chains to form a 2DWPN as a result of the π···π stacking interactions with BPE moieties, while the steric m-xylene residues are unable to do so. We suspect that the construction of the woven topology is influenced strongly by the steric effect of benzene solvent molecules. To verify this hypothesis, we prepared single crystals of the dative B–N polymer in p-xylene and the more sterically bulkly o-xylene using the same method as that employed in the production of 2DWPN-1 and NWPN-1, respectively. In a manner similar to the preparation of 2DWPN-1 and NWPN-1, a 2DWPN (2DWPN-2) was generated in p-xylene by π···π interactions between p-xylene residues and BPE moieties, and a parallel stacking non-woven polymer (NWPN-2) was obtained in o-xylene (Supplementary Figs. 31–56), an observation that is consistent with the expected result. Overall, the conformation of the dative B–N bond is impacted by the steric and electronic properties of the interacting components41,42. This kind of intrinsic conformational flexibility is the central element for the acquisition of the woven topology in the presence of selected solvents.Preparation and characterization of 2DWPN-1 nanosheetsFrom the packing of 2DWPN-1 (Fig. 1g), the adjacent layers interact only through solvent molecules. We adopted a computational method to analyse the difference between the interactions of adjacent layers and the interactions among different components inside the same layer in the crystal structure of 2DWPN-1. Theoretical calculations showed that the packing energy (ΔEab) of 2DWPN-1 along the ab plane is 4.5 times stronger than that (ΔEc) along the c axis (Supplementary Fig. 57). The much stronger interactions among different components in the same 2D, ab plane, compared with the interactions between 2D layers along the c axis, make the layered 2DWPN-1 crystals easy to exfoliate into large-size crystalline woven monolayers (Supplementary Figs. 58–66).Inspired by the weak interlayer interactions in the layered 2DWPN-1 crystals, we exfoliated a bulk crystal using a Scotch Magic Tape-assisted mechanical method43, with reference to the exfoliation (Fig. 3a,b) of traditional van der Waals 2D materials. Micrometre-scale 2DWPN-1 flakes were obtained on a SiO2/Si substrate (Fig. 3c). Notably, we obtained a free-standing monolayer woven flake of 2DWPN-1 with a thickness of 1.3 nm, and introduced the woven polymer network materials successfully to the 2D limit (Fig. 3d). The AFM images of a bilayer woven flake with a thickness of 2.6 nm and a trilayer woven flake with a thickness of 3.9 nm are shown in Fig. 3e,f, respectively. Moreover, the root-mean-square roughness (Rq) values of these 2DWPN-1 flakes (Supplementary Fig. 64) were less than 0.2 nm, indicating atomically flat surfaces. Compared to the 2D woven nanosheets obtained by the liquid exfoliation method (Supplementary Figs. 64 and 65), the clean and atomically flat surfaces of these as-prepared 2DWPN-1 flakes are more favourable for exploring intrinsic properties. As a result, the layered 2DWPN-1 crystal was exfoliated successfully to the molecular thickness limit. The 2DWPN-1 flakes have a micrometre-scale edge size and atomically flat surfaces, paving the way for applications of woven polymer networks in the field of 2D materials.Fig. 3: Fabrication and characterization of atomically thin 2DWPN-1 flakes.a, Schematic diagram of Scotch Magic Tape-assisted mechanical exfoliation and transfer of a layered 2DWPN-1 bulk crystal. b, Schematic illustration of the exfoliation of 2DWPN-1 nanosheets. c, Optical microscopic image of exfoliated monolayer (1 L), bilayer (2 L), trilayer (3 L) and multilayer (Multi-L) 2DWPN-1 flakes on a SiO2/Si substrate. d–f, AFM images of monolayer (d), bilayer (e) and trilayer (f) 2DWPN-1 flakes on a SiO2/Si substrate and corresponding thicknesses (bottom) of 1.3 nm, 2.6 nm and 3.9 nm, respectively. The thickness plots are along the green arrows. g, Raman spectra of bulk 2DWPN-1 and 2DWPN-1 flakes. h,i, The corresponding Raman mapping images based on the intensities at 1,210 cm−1 (h) and 1,633 cm−1 (i) of a 2DWPN-1 flake. The inset shows the corresponding optical microscopic image.Source dataRaman spectroscopy was performed to obtain structural and phonon-vibration information for 2DWPN-1. Raman spectra (Fig. 3g) of a bulk 2DWPN-1 crystal show two main peaks at 1,210 and 1,633 cm−1, and a few weaker peaks at 1,050, 1,072, 1,264, 1,336 and 1,552 cm−1. The analysis, combined with the structure of 2DWPN-1, indicates that the Raman peaks at 1,050 and 1,072 cm−1 are attributable to the dative B–N vibrations, and the signals at 1,336, 1,633 and 1,552 cm−1 arise from C=C vibrations. The peaks at 1,210 and 1,264 cm−1 correspond to the C–O and C–B vibrations, respectively. The experimental data match well with the calculated results (Supplementary Figs. 67–70). Without an obvious peak position shift, the Raman peak intensities gradually decrease with the thickness of 2DWPN-1 decreasing to the 2D limit. The Raman mapping images based on the intensity at 1,210 cm−1 (Fig. 3h) and 1,633 cm−1 (Fig. 3i) of a 2DWPN-1 flake (inset of Fig. 3h) show homogeneous colour distribution, demonstrating the uniform and high-quality 2D sheets of 2DWPN-1.We also performed transmission electron microscopy (TEM) to investigate the real-space superstructure information of 2DWPN-1 nanosheets. From low-magnification TEM images (Supplementary Fig. 71), we observed that the 2DWPN-1 crystals have a nanoplate morphology with homogeneous distributions of B, C, N and O as probed by electron energy-loss spectroscopy (Supplementary Figs. 72 and 73). The selected-area electron diffraction patterns of 2DWPN-1 nanosheets match well with the simulated electron diffraction patterns based on bulk single-crystal structures (Supplementary Fig. 71), confirming that the 2D woven nanosheets retain the periodic structures of the ab plane in the 3D crystals. The explicit molecular-level real-space structural elucidation of the purely organic networks requires high-resolution (HR) TEM imaging. Organic materials are extremely vulnerable to electron-beam irradiation and are subject to radiolytic structural damage under the high accumulated electron dose required for traditional HRTEM imaging. Recent advances in low-dose electron microscopy allow the direct imaging of beam-sensitive materials such as metal–organic frameworks44,45,46,47. Nevertheless, the low-dose imaging of purely organic networks has greater challenges on account of the low image contrast in addition to the high beam sensitivity of the networks.It has been widely acknowledged that electron microscopy carried out under cryogenic conditions (that is, cryo-EM) can minimize radiolytic beam damage effects and thus alleviate the dose-limited resolution problem48,49,50,51. To this end, by integrating low-dose and cryo-EM imaging techniques together with a state-of-the-art electron direct detector and a custom-designed ultrastable cryo-transfer holder, we were able to image the 2DWPN-1 nanosheets with high spatial resolution and superstructure integrity. A low-dose cryo-EM image of the woven polymer networks projected along the [001] direction of 2DWPN-1 crystals is shown in Fig. 4a. From the denoised image after correcting for contrast inversion effects arising from the contrast-transfer function (CTF) of the objective lens, the topological features of the woven polymer networks can be visualized clearly with structural information transfer up to 2.2 Å. From the false-colour motif-averaged and symmetry-imposed images in Fig. 4b, the woven polymer networks exhibit contrast closely resembling interconnected hollow ovals that match well with the projected electrostatic potentials simulated based on the proposed structural model for 2DWPN-1 (Fig. 4c). These observations confirm unambiguously the identical polymer chain conformations and associate the woven topologies of 2DWPN-1 with those determined by X-ray crystallography.Fig. 4: Cryogenic low-dose HRTEM image of the crystalline 2DWPN-1.a, Cryogenic low-dose HRTEM image of the crystalline 2DWPN-1 taken along the [001] direction. The image was denoised and the contrast inversion effects caused by the objective lens CTF were corrected. Insets show a raw image (upper left) and the fast Fourier transform pattern (upper right) of the HRTEM image with structural information transfer up to 2.2 Å. b, A false-colour motif averaged along the [010] direction. c, The simulated projected electrostatic potential with a point spread function width of 2.2 Å. The inset of b shows the p2gm projection symmetry-imposed image, and the projected structural model of 2DWPN-1 is embedded in c.Mechanical properties of 2DWPN crystalsApart from their aesthetically appealing superstructures, the woven topology also furnishes materials with unique mechanical behaviour. To reveal quantitatively the influence of the woven structure on the mechanical response of 2DWPN-1 crystals, PeakForce quantitative nanomechanical mapping and nanoindentation were employed. For a comparison, we also investigated the mechanical properties of the NWPN-1 crystals. Here 2DWPN-1 and NWPN-1 differ only in their polymer topologies, so the differences in their mechanical properties reflect the influence of their woven topologies. The uniform surface textures (Fig. 5a,b) of crystals of 2DWPN-1 and NWPN-1 ensure a homogeneous modulus distribution. The Derjaguin–Müller–Toporov moduli are 2.1 and 3.2 GPa for 2DWPN-1 and NWPN-1, respectively. Therefore, the surface flexibility of 2DWPN-1 with the woven topology is superior to that of non-woven NWPN-1.Nanoindentation was employed to investigate the deformation properties of 2DWPN-1 and NWPN-1. Load–displacement (P–h) curves were obtained (Fig. 5c) by indenting crystals of 2DWPN-1 or NWPN-1 with a Berkovich tip in a load-controlled mode of 4 mN. Under the same load, 2DWPN-1 (D = 1,159 nm, H = 0.11 GPa) exhibits a greater displacement depth (D) and a lower hardness value (H) than NWPN-1 (D = 868 nm, H = 0.23 GPa) along the c axis, which is perpendicular to the multilayer stacking plane of the crystal structures. Moreover, load–displacement (P–h) curves with similar slopes obtained by indenting 2DWPN-1 and NWPN-1 crystals (Fig. 5d and Supplementary Figs. 74–76) confirm the good homogeneity of these crystals using a load-controlled mode of different loads from 1 to 6 mN. Overall, the quantitative nanomechanical mapping and nanoindentation results both show that the woven topology endows the rigid crystal with excellent flexibility, an observation that may be attributed to the unique and effective mechanisms to disperse external forces as a result of the conformational freedom of polymer chains and sliding space in the topological networks. In particular, the molecular warp and weft strands experience relative sliding through woven nodes with an interplanar distance of 3.79 Å under external stimuli while the interlaced structures of the polymer chains maintain the integrity of woven networks (Supplementary Fig. 14). As a result, the stress is dispersed throughout the polymer network, preventing stress concentration that could lead to structural damage while imparting exceptional flexibility upon the crystal (Supplementary Videos 1 and 2). These findings highlight that topological control is a simple and effective method to develop crystalline materials with exotic mechanical performance.Fig. 5: Study of the mechanical properties of 2DWPN-1 and NWPN-1.a,b, Mapping of the Derjaguin–Müller–Toporov moduli of 2DWPN-1 (a) and NWPN-1 (b). The insets show the corresponding Derjaguin–Müller–Toporov modulus distribution. c, P–h curves for 2DWPN-1 on the (001) facet, and NWPN-1 on the (110) facet of the multilayer stacking plane at a fixed load (4 mN). d, P–h curves obtained using various loadings from 1 to 6 mN.Source data
