Superior mechanical properties of multilayer covalent-organic frameworks enabled by rationally tuning molecular interlayer interactions

Significance

Two-dimensional (2D) materials are intrinsically strong materials, but most of them lack a strong interlayer interaction, leading to their brittle nature and low fracture toughness. Here, we synthesized covalent-organic frameworks (COFs), a new class of organic 2D materials that inherit the high strength of inorganic 2D materials and designability of polymers, and established the layer-dependent structure–mechanical property relationship. By performing nanoindentation on mono- to trilayer COF films, we determined the mechanical properties of COF films as a function of layer number. Changing the side groups in COF films allows the interlayer interaction, thus mechanical properties to be engineered at the molecular level. The designability of 2D COFs provides a platform to design strong and tough 2D materials.

Abstract

Two-dimensional (2D) covalent-organic frameworks (COFs) with a well-defined and tunable periodic porous skeleton are emerging candidates for lightweight and strong 2D polymeric materials. It remains challenging, however, to retain the superior mechanical properties of monolayer COFs in a multilayer stack. Here, we successfully demonstrated a precise layer control in synthesizing atomically thin COFs, enabling a systematic study of layer-dependent mechanical properties of 2D COFs with two different interlayer interactions. It was shown that the methoxy groups in COF_TAPB-DMTP provided enhanced interlayer interactions, leading to layer-independent mechanical properties. In sharp contrast, mechanical properties of COF_TAPB-PDA decreased significantly as the layer number increased. We attributed these results to higher energy barriers against interlayer sliding due to the presence of interlayer hydrogen bonds and possible mechanical interlocking in COF_TAPB-DMTP, as revealed by density functional theory calculations.

As graphene has been proven to be one of the strongest materials (1, 2), two-dimensional (2D) materials have emerged as high-performance building blocks for next-generation engineering materials (3). Although most 2D materials exhibited high strength, they are intrinsically brittle, resulting in low fracture toughness (4–7). More importantly, because of the weak interlayer van der Waals (vdW) interactions in 2D materials, the exceptionally high strength achieved at the monolayer limit cannot be sustained with increasing layer numbers. Additionally, the low fracture toughness of 2D materials generally persists from monolayer to multilayers (4, 6, 8). These limitations become major hurdles for widespread applications of 2D materials in structural applications. One possible strategy to overcome these limitations is to enhance the interlayer interactions by introducing non-vdW interactions, such as hydrogen bonds or electrostatic forces (9–12). It has already been shown that hydrogen bonds between functional groups in graphene oxide nanosheets can restrict the growth of cracks, leading to increased fracture toughness (13–15). However, because of the inertness of the surfaces in most inorganic 2D materials and the inevitable trade-off between chemically functionalizing such inert surfaces and maintaining their high intrinsic strength, such a strategy remains material-specific and synthesis-challenging.

2D polymers are organic analogs to graphene that can be designed and synthesized with molecular-level structural control. They combine excellent mechanical properties of conventional 2D materials with low density, good processability, and designability and therefore provide a new platform for designing strong and tough 2D materials (16, 17). Notably one representative family of 2D polymers is the covalent-organic frameworks (COFs), in which organic monomers precisely assemble into highly ordered porous structures. The superior mechanical properties of 2D COFs have been demonstrated by Feng et al. for monolayer porphyrin-based COFs, and an average Young’s modulus of 267 ± 30 GPa was obtained (18). However, similar to inorganic 2D materials, the lack of strong interlayer interaction makes it challenging for stacked COFs to inherit the superior mechanical properties of their monolayer counterparts. As a result, studies of multilayer COFs generally showed much smaller Young’s modulus in the range of 10 to 12 GPa (19, 20). Distinct from inorganic 2D materials, the tunable surface chemistry of COFs enables the facile formation of strong bonds in-between neighboring layers (21, 22). Recently, COFs with interlayer hydrogen bonds have been successfully synthesized (23–25). The presence of these hydrogen bonds can significantly enhance the chemical stability of COFs and provide an opportunity to tune the mechanical properties of COFs at the molecular level by engineering interlayer interactions.

Results and Discussion

As discussed earlier, retaining outstanding mechanical properties of various monolayers in their multilayer structures poses a major challenge for scale-up engineering applications of 2D materials. To better address such a challenge and understand the effects of interlayer interactions on the mechanical properties of 2D COFs, we compare COF_TAPB-DMTP [TAPB, 1,3,5-tris(aminophenyl)benzene; DMTP, 2,5-dimethoxyterephthaldehyde] and COF_TAPB-PDA (PDA 1,4-phenylenedialdehyde) (Fig. 1A). These two COFs are chosen because they have similar molecular structures, with the only difference being whether or not methoxy groups are present. The methoxy groups in COF_TAPB-DMTP can reduce the interlayer repulsion effect and form strong interlayer hydrogen bonds (25, 26). Our density function theory (DFT) calculations indicated that COF_TAPB-DMTP has a higher interlayer bonding energy (4.66 eV per unit cell) than that of COF_TAPB-PDA (3.77 eV per unit cell). Meanwhile, hydrogen bonds and $\pi -\pi$ interactions can be found in the differential charge density of COF_TAPB-DMTP, which does not exist in COF_TAPB-PDA. Fig. 1B plots the DFT-optimized molecular structure (Upper) of COF_TAPB-DMTP with AA stacking and the corresponding differential charge densities (Lower). Interlayer hydrogen bonds ( $C-H\cdots N$ ) form at a bond length of 3.41 $\mathrm{\AA }$ , which are also evident in the charge density pattern (highlighted by red triangles): The charge density on methoxy hydrogen decreases significantly (light blue shade), and there is an increase in charge density on the nitrogen of the adjacent layer (yellow shade). Similar hydrogen bonds are also found in other COFs materials (24, 25). Meanwhile, the electrostatic interactions between neighboring layers avoid charge repulsion (24) and result in the $\pi -\pi$ interactions between the benzene rings of the upper and lower layers. These two types of bonds (hydrogen bonds and $\pi -\pi$ interactions) synergistically enhance the interlayer interaction of COF_TAPB-DMTP. Fig. 1C illustrates the DFT-optimized molecular structure (Upper) of COF_TAPB-PDA with AA stacking and the corresponding differential charge densities (Lower). No methoxy exists in the molecular chains of each layer, and the charge deflection mainly occurs within each layer. The profiles of the loss and gain of electrons in the upper and lower layers are nearly the same, indicating the existence of electrostatic repulsion between the two layers and the vdW force nature of the interlayer interaction in COF_TAPB-PDA. These findings provide strong motivations to precisely synthesize layer-controlled 2D COFs as suggested by the DFT results and experimentally evaluate the interlayer interaction effects on their mechanical properties down to the monolayer limit.

Fig. 1.

(A) COFs thin films prepared by the condensation reaction between TAPB and dialdehydes. (B) DFT-optimized molecular structure (Upper) of COF_TAPB-DMTP with AA stacking and the corresponding differential charge densities (Lower). (C) DFT-optimized molecular structure (Upper) of COF_TAPB-PDA with AA stacking and the corresponding differential charge densities (Lower). Oxygen, nitrogen, carbon, and hydrogen atoms are represented with red, blue, grey, and white beads, respectively. Light blue and yellow shades in the pattern of charge density represent the loss and gain of electrons, respectively. The isosurface is set as 0.0005 e/Bohr³. (D) Optical microscope image of a monolayer COF_TAPB-DMTP thin film. (E) Tapping mode AFM image of monolayer COF_TAPB-DMTP thin film on a 300 nm SiO₂/Si. (F) corresponding height profile of the solid lines in E. (G) Raman spectra of monomers TAPB, DMTP, and COF_TAPB-DMTP thin film.

To achieve layer-controlled 2D COFs synthesis, Langmuir–Blodgett (LB) trough method has first been developed as an efficient way to prepare large-area monolayers. Very excitingly, we found that by carefully controlling the density of monomers on the water surface, i.e., the surface pressure in the LB trough, large-area and precisely layer-controlled (mono- to trilayers) COFs thin films can be prepared at the air/water interface by a dynamic Schiff-base reaction between TAPB and dialdehydes (DMTP and PDA, respectively) (18, 27). The proposed reaction between two monomers, TAPB and dialdehydes (DMTP and PDA), was schematically illustrated in Fig. 1A. The unit cells of these two COFs were shown in SI Appendix, Fig. S1. SI Appendix, Fig. S2 schematically illustrated the synthesis of monolayer COFs thin films in an LB trough. A chloroform/methanol (volume ratio = 3:1) solution of monomer TAPB was gently spread on the water surface by a microsyringe and compressed to a surface pressure of 1 mN/m to form a densely packed layer. Followed by adding an acetic acid solution of dialdehydes, a freestanding monolayer COFs thin film was formed at the air/water interface. In this case, acetic acid served as an effective solution to dialdehydes (SI Appendix, Fig. S3) and a catalyst for imine bond formation. The reaction had been kept for 24 h, and the resulting thin film was transferred to solid substrates for further characterization. Under an optical microscope, a millimeter-sized, continuous, homogeneous thin film with some small cracks was observed (Fig. 1D). The thickness of the atomically thin film was 0.6 nm, close to previously reported monolayer (1L) 2D COFs (18, 27), with a smooth and uniform surface measured by tapping mode atomic-force-microscope (AFM) (Fig. 1 E and F). Since the film thickness can be controlled by the surface pressure using the LB trough method, the thicknesses of as-synthesized 2D COFs were measured to be 1.0 and 1.4 nm with the surface pressure of 3 and 10 mN/m, corresponding to bilayer (2L) and trilayer (3L) COFs, respectively (SI Appendix, Fig. S4 A and B). The chemical structures of the obtained COFs thin films were characterized by Raman spectra, as shown in Fig. 1G. Compared to the Raman spectra of monomers (TAPB and DMTP), after condensation, the NH₂ wagging band at around 1,360 cm⁻¹ and C=O stretching band at around 1,670 cm⁻¹ vanished, and a peak at around 1,600 cm⁻¹ corresponding to the vibration mode of –C=N– bonds appeared, indicating the formation of imine bonds. A high-resolution transmission electron microscopy study was conducted to characterize the detailed structures of these porous thin films. Due to the instability of polymers under electron beam irradiation, only an amorphous porous structure was observed (28). The diameter of the pores was measured to be around 3.6 nm, closed to the theoretic pore size of 3.7 nm (SI Appendix, Fig. S5).

Next, to experimentally evaluate the effects of interlayer interaction on the mechanical properties of layer-controlled 2D COFs, the as-synthesized COFs thin films (1L-3L) were then transferred onto a prepatterned Transmission Electron Microscopy (TEM) sample grid for subsequent nanoindentation test using a Langmuir–Schaefer technique (SI Appendix, Fig. S6). As demonstrated in SI Appendix, Fig. S7A, this particular sample was a monolayer. According to the tapping mode AFM image, it spanned over the holes and attached well to the substrate (Fig. 2 A, Inset). Scanning electron microscope image showed that the sample was suspended with high yield over the entire holey substrate without visible cracks or voids (SI Appendix, Fig. S7B). The schematic illustration of the nanoindentation test is shown in the inset SI Appendix, Fig. S8A, where the AFM tip was centered at the suspended region and pressed downward to deflect the atomically thin COFs films with different thicknesses. The load versus displacement curves were recorded (1, 29). Two cycles of loading and unloading curves with a loading increment of 50 nN were obtained from the same sample. As shown in SI Appendix, Fig. S8B, the loading and unloading curves retraced with each other well at different load levels, indicating the elastic deformation of the thin film and no occurrence of sample sliding during the indentation process. The load–deflection (F-δ) relationship of the sample can be obtained after subtracting the deflection of the AFM cantilever. Considering the polycrystalline nature of the COFs thin films, we modeled the deformation of the COFs thin films to be isotropic, which consists of two parts: 1) Under a small load, the deformation was linear as it was dominated by the deflection of the prestrained thin film and 2) At a large load, the nonlinearity in-plane stretching of a suspended film was characterized by a cubic F~δ³ relationship (SI Appendix, Fig. S8C). The F-δ relationship can be expressed as: (30)

$$\mathrm{F}=\left({\sigma }_0^{2D}\pi \right)\delta +\left(\frac{{\mathrm{E}}^{2\mathrm{D}}{q}^3}{{\mathrm{r}}^2}\right){\delta }^3,$$ [1]

Fig. 2.

(A) Load–deflection and fitting curves of 1-3L COF_TAPB-DMTP thin films. (Inset) Tapping mode AFM image of monolayer COF_TAPB-DMTP thin film suspended over a 750-nm circular hole. (B) Histogram of 2D modulus of 1-3L COF_TAPB-DMTP and COF_TAPB-PDA thin films. (C) Young’s modulus of 1-3L COFs thin films. Dash lines are the value of average Young’s modulus of monolayer COFs thin films. (D) Load-deflection curve (up to fracture) and fitting curve of a monolayer COF_TAPB-DMTP. (Inset) Tapping mode AFM image of a fractured COF_TAPB-DMTP thin film after indentation. (E) Histogram of the 2D strength of 1-3L COF_TAPB-DMTP and COF_TAPB-PDA thin films. (F) Breaking strength of 1-3L COFs thin films. Dash lines are the value of the average breaking strength of monolayer COFs thin films.

where F is the applied load, δ is the deflection at the center of the thin film, ${\mathrm{\sigma }}_0^{2D}$ is the prestrain, E^2D is the 2D elastic modulus, r is the radius of the hole, and q is a constant determined by Poisson’s ratio (ν) as q = 1/(1.05–0.15ν−0.16ν²). The values of the Poisson’s ratio of both COFs are calculated to be around 0.3 (31). The value of ${\mathrm{\sigma }}_0^{2D}$ and E^2D can be derived by fitting the F-δ data using Eq. 1. Typical loading curves of mono- to trilayer COF_TAPB-DMTP and COF_TAPB-PDA thin films, with a deflection of 50 nm and the corresponding fitting curves (R² > 0.99), were shown in Fig. 2A and SI Appendix, Fig. S9, respectively. In our experiment, for each atomically thin COFs film with different thicknesses, we measured more than 25 experimental points from three batches of samples, and their E^2D distribution was shown in Fig. 2B. The E^2D of 1-3L COF_TAPB-DMTP were 16.2 ± 0.8 N/m (number of tests, N = 52), 31.4 ± 1.5 N/m (N = 47), 50.4 ± 2.2 N/m (N = 54), and 15.6 ± 1.1 N/m (N = 27), 28.8 ± 1.7 N/m (N = 33), 41.6 ± 1.7 N/m (N = 32) for 1-3L COF_TAPB-PDA, respectively. The narrow distribution of E^2D indicates good reproducibility and high quality of these thin films since reversible reactions could minimize the defect density with extended reaction time (18). Time-dependent experiments were performed on a bilayer COF_TAPB-DMTP sample to help confirm this effect, and the results showed that with a reaction time of 6 h, the sample was defective and too fragile to be measured. As the reaction time extended to 12 h, the E^2D of the sample was 22 N/m, and the distribution was wide. While after 24 h reaction, no significant change of the value and distribution of E^2D was found, indicating efficient defect repairing under dynamic reversible reaction (SI Appendix, Fig. S10). To further validate our measurements, we also compared the E^2D of thin film samples suspended on holes with different radii (750 and 1,000 nm); the results showed negligible differences (SI Appendix, Fig. S11). The value of ${\mathrm{\sigma }}_0^{2D}$ were in the range of 0.1 to 0.4 N/m (SI Appendix, Fig. S12), suggesting strong interaction between COFs thin films and substrates. SI Appendix, Fig. S13 summarizes the 2D moduli of COF_TAPB-DMTP and COF_TAPB-PDA of different thicknesses. The dash lines were obtained by multiplying the E^2D of the monolayer with the layer number. Compared to COF_TAPB-PDA (black line), where the value of E^2D decreased more than 10% as the thickness increased from 1L to 3L, E^2D increased along with the red dash line for COF_TAPB-DMTP. The Young’s modulus of COFs thin films can be obtained by dividing E^2D by thin film thickness. For Young’s modulus calculation, we used the thickness of a COF_TAPB-DMTP and COF _TAPB-PDA monolayer of 0.344 nm and 0.363 nm, respectively, as supported by DFT calculation. For 1-3L COF_TAPB-DMTP thin films, Young’s modulus was 47.1 ± 2.3 GPa, and 45.6 ± 2.2, and 48.8 ± 2.1 GPa, respectively, indicating that Young’s modulus was independent of layer numbers. While for COF_TAPB-PDA, Young’s modulus significantly decreased as the layer number increased (Fig. 2C). Such a decreasing in Young’s modulus has been reported in many 2D materials (e.g., graphene, MoS₂) and was attributed to the interlayer slippage (32–34). For 2D materials with increasing interlayer interaction (e.g., h-BN, graphene oxide, MXenes) (34–36). Young’s modulus was independent of the thickness. These results strongly suggested the enhanced interlayer interactions in COF_TAPB-DMTP thin films.

We further measured the breaking strength of COFs thin films. Fig. 2D illustrates an F-δ curve of a fractured monolayer COF_TAPB-DMTP thin film with a peak force of around 70 nN. When the F-δ curve was fitted using Eq. 2 (black dash line), a 2D modulus of 16.1 N/m was obtained. After indentation, the thin film fractured and it still hung around the hole, suggesting the rupture occurred in the area with direct contact with the AFM tip (Fig. 2 D, Inset). SI Appendix, Fig. S14 shows the comparison of the F-δ curves of 1-3L COF_TAPB-DMTP and COF_TAPB-PDA. For monolayer thin films, the area under the curves was similar. However, with the layer number increased, together with the increased deflection and breaking force, the area under the F-δ curves of COF_TAPB-DMTP significantly increased, indicating enhanced toughness of COF_TAPB-DMTP as the layer number increased.

The maximum fracture stress at the center part of the hole can be extracted using the formula of a clamped, linear elastic, circular suspended membrane under a spherical indenter: (37)

$${\mathrm{\sigma }}^{2\mathrm{D}}={\left(\frac{\mathrm{F}{\mathrm{E}}^{2\mathrm{D}}}{4{\mathrm{\pi r}}_{\mathrm{t}\mathrm{i}\mathrm{p}}}\right)}^{\frac{1}{2}},$$ [2]

where σ^2D is the maximum stress at the center of the film, F is the fracture load, r_tip is the radius of the AFM tip. In our case, r_tip = 70 nm (SI Appendix, Fig. S15). Fig. 2E illustrates the statistical histogram of the σ^2D of COFs thin films with different thicknesses. The obtained σ^2D of COF_TAPB-DMTP and COF_TAPB-PDA thin films with different thicknesses are shown in SI Appendix, Fig. S16. Similar to SI Appendix, Fig. S13, the dash lines were obtained by multiplying the value of σ^2D of monolayer COFs with layer number. The σ^2D of COF_TAPB-PDA decreased more significantly as the layer number increased. σ^2D of a trilayer COF_TAPB-PDA thin film was 30% smaller than three times of σ^2D of a monolayer, while the σ^2D of COF_TAPB-DMTP with different layer thicknesses were close to the red dash line. The breaking strength of COFs thin films can be obtained by dividing σ^2D with the film thickness. As shown in Fig. 2F, the breaking strength of 1L-3L COF_TAPB-DMTP thin films was 4.62 ± 0.29 GPa, 4.59 ± 0.19 GPa, and 4.83 ± 0.16 GPa, respectively, and for COF_TAPB-PDA thin films, the values were 4.16 ± 0.22 GPa, 3.66 ± 0.19 GPa and 3.30 ± 0.16 GPa, respectively. These values of breaking strength were 8 to 10% of the Young’s modulus of corresponding thin films, approaching the theoretical breaking strength. According to our previous work, at small length scales, COFs could become flaw insensitive, that is, defects or grain boundaries do not significantly affect its mechanical strength (19). Note that even though these thin films were polycrystalline, high failure strains were observed, which might be attributed to the porous skeleton of COFs compared to more compact 2D materials. Based on the previous theoretical calculations, because of the rotation and tilting of the flexible bonds, COFs can reach very high strain levels (as high as 25%) before failure (19).

Our experimental results clearly show that the Young’s modulus (as well as strength) of multilayer COF_TAPB-PDA is lower than the sum of those of individual layer, e.g., a two-layer COF_TAPB-PDA has a 2D Young’s modulus lower than twice that of a monolayer COF_TAPB-PDA. This is similar to the phenomenon observed from MoS₂/graphene and MoS₂/WS₂ heterostructures (38), which was attributed to the interlayer slippage in between neighboring layers during indentation (39). However, we have also experimentally shown that two-layer and three-layer COF_TAPB-DMTP have a 2D Young’s modulus (as well as strength) comparable or even higher than twice and three times of that of a monolayer COF_TAPB-DMTP, respectively (Fig. 2). To further assess possible roles played by the side chains in COF_TAPB-DMTP, we prepared layer-controlled COF_TAPB-DETP (SI Appendix, Fig. S17). COF_TAPB-DETP has similar side chains to COF_TAPB-DMTP, but without strong interlayer hydrogen bonds. The layer-dependent indentation results are shown in SI Appendix, Fig. S17. Similar to COF_TAPB-PDA, COF_TAPB-DETP shows a decreasing Young’s modulus as the layer number increases although with a slightly weaker layer thickness dependence, indicating that mechanical interlocking of the side chains is a possible factor that influences the layer-dependent mechanical properties of COFs in this study (detailed in SI Appendix). While the overall structure in the synthesized COF films is polycrystalline, we hypothesize that the orientation of the interlayer stacking is not completely random but with short-range order. Such a hypothesis is justified as follows. Recent studies show that planar monomers similar to TAPB tend to assemble and stack together by strong π–π interaction (40). In our case, we hypothesize that after compression of the barriers, TAPB monomers tend to stack together and then react with the other monomers to form multilayers, in which there’s a strong trend to be aligned and form AA stacking. As the reversible reaction in COF synthesis gives rise to the “self-error checking” of the dynamic bonds and the resulting structure of COFs, a thermodynamically favored, self-sorting structure with short-range order (e.g., local stacking) can be formed (41–43). The verification of the hypothesis is detailed in the SI Appendix, Figs. S19 and 20.

To better understand the enhanced mechanical properties of multilayer COF_TAPB-DMTP, first-principle DFT calculations are performed using the Vienna Ab initio Simulation Package (VASP) (44, 45) to investigate its interlayer interactions (Detailed in Materials and Methods). Fig. 3A plots the variation of energy per unit cell ( $\mathrm{\Delta }E$ ) from the lowest energy states of COFs as the function of interlayer sliding displacement in the zigzag (ZZ) direction. The highest energy variation of COF_TAPB-PDA (defined as the energy barrier of the sliding process) is 2.96 eV/ unit cell, corresponding to a critical interlayer sliding displacement of 8 $\mathrm{\AA }$ . For COF_TAPB-DMTP, the energy variation is slightly higher than that of COF_TAPB-PDA at the beginning of the sliding process and the energy barrier reaches 3.08 eV/ unit cell at a critical interlayer sliding displacement of 10 Å. As the sliding displacement further exceeds the critical value, the energy variation starts to decrease until the sliding displacement leads to the formation of AB stacking of the COFs. The energy variation due to further sliding is generally mirrored symmetrically to the above process until the COFs form AA stacking again, with a slight deviation that could be attributed to the sliding-induced rotation of the chains. Fig. 3B shows the plots of energy variation of COFs as the function of interlayer sliding displacement in the armchair (AC) direction. The energy variation of COF_TAPB-DMTP is slightly higher than that of COF_TAPB-PDA at the beginning of the sliding process and the energy barrier of COF_TAPB-DMTP (4.00 eV/ unit cell) is much higher than that of COF_TAPB-PDA (3.28 eV/ unit cell).

Fig. 3.

Sliding Behaviors of COFs along AC and ZZ Directions. Energy evolution of two different COFs as a function of sliding displacement along (AC) direction (A) and (ZZ) direction (B), respectively. (C–E) Differential charge densities of COF_TAPB-DMTP in the sliding process along ZZ direction at three stages labeled as II, III and IV, respectively. The Upper panels in (C–E) represent the Top view of differential charge densities. The Lower panels in (C–E) illustrate the side views of differential charge densities at the locations labeled by the red square in the corresponding Upper panel. Light blue and yellow shades in the Lower panels represent the loss and gain of electrons, respectively. Differential charge densities are visualized using isosurfaces of 0.0005 e/Bohr³.

To quantitively compare the interlayer interaction of the two types of COFs, we calculate the differential charge densities between each layer during the sliding process (Fig. 3 C–E and SI Appendix, Figs. S21–S24). The charge density of pristine COFs (Stage I) is illustrated in Fig. 1 B and C. Fig. 3 C–E plot the atomic structures and charge densities of COF_TAPB-DMTP with three representative stages in the sliding process along the ZZ direction. As discussed above, pristine COF_TAPB-DMTP forms hydrogen bond ( $C-H\cdots N$ ) and $\pi -\pi$ interaction between each layer. In Stage II (Fig. 3C), $C-H\cdots N$ bonding and $\pi -\pi$ interaction disappear gradually, and the hydrogen bond is reformed simultaneously between methoxy hydrogen and methoxy oxygen in different layers ( $C-H\cdots O$ ). The repeated breaking and reforming of hydrogen bonds increase the energy dissipation during the sliding process until the energy barrier is reached (Stage III, Fig. 3D) at which there is no electrostatic interaction between the upper and lower layers. Further sliding leads to decreasing energy due to the forming of the hydrogen bond between methoxy hydrogen and nitrogen ( $C-H\cdots N$ ), as shown in Fig. 3E (Stage IV). From the global perspective view of charge density of COF_TAPB-DMTP in the sliding process along the ZZ and AC directions (SI Appendix, Figs. S21 and S22), the dipole-dipole interaction mainly occurs around the methoxy of each layer. It is worth noting that such an interaction binds neighboring layers tightly and thus results in the inheritable 2D modulus and strength of layered COFs. In contrast, there is no hydrogen bond formation in COF_TAPB-PDA (Fig. 1C) and the intralayer electron transfer leads to the repulsion of each layer. SI Appendix, Figs. S23 and S24 plot the evolution of differential charge density of COF_TAPB-PDA in the sliding process along ZZ and AC directions. The electron migration of the intralayer reduces significantly and finally disappears (Stage IV), which suggests that the interlayer interaction is mainly vdW force. Such a weak interlayer interaction in multilayer COF_TAPB-PDA leads to a 2D modulus (as well as strength) of multilayers lower than the sum of the property of each monolayer. Local AA stacking in COF layers can significantly enhance the hydrogen bond formation between neighboring COF layers, even if they are not so well aligned globally. To further verify such a mechanism, we use the fully relaxed structures of the COF bilayers with various misorientation angles ranging from 0° to 90° (e.g., those shown in SI Appendix, Figs. S19 and S20) to investigate the energy variation of the system as Layer 1 slides on Layer 2 in the AC direction of Layer 2 (SI Appendix, Fig. S25A, Movies S4, and S5). The misorientation between the top and bottom layers causes the distortion of the sidewall and thus deforms the hexagon rings to release the inner stress. Due to the thermal oscillation, both COF_TAPB-DMTP and COF_TAPB-PDA have a large out-of-plane deformation (around 3.5 nm in COF_TAPB-DMTP and 2.8 nm in COF_TAPB-PDA), as shown in SI Appendix, Fig. S26 A and B, which is much larger than their interlayer distance (0.344 nm and 0.366 nm). It is found that the thermal oscillation, as well as the large size of the sidewalls of the hexagonal rings of the COF_TAPB-DMTP and COF_TAPB-PDA, can facilitate the local distortion of some sidewalls, which in turn results in local physical interlocking between the Upper and Bottom layers during the sliding process of both COF_TAPB-DMTP and COF_TAPB-PDA (SI Appendix, Fig. S27, Movies S6, and S7). Although such physical interlocking only occurs in a small portion of the sidewalls (0.9 to 5.4% and SI Appendix, Table S1), it could still play a role in affecting mechanical properties of layered COFs. SI Appendix, Fig. S25B shows the plots of energy variations of the COF_TAPB-DMTP and COF_TAPB-PDA bilayers with 10 different misorientation angles (from 0° to 90° with an increment of 10°) as a function of interlayer sliding distance (in terms of simulation time). The range of the energy variations of all COF_TAPB-DMTP simulation cases is shaded in red, and that of all COF_TAPB-PDA simulation cases is shaded in blue. It is clearly shown that the energy variation during interlayer sliding of a COF_TAPB-DMTP bilayer with any misorientation angle is higher than that of a COF_TAPB-PDA bilayer with any misorientation angle. In other words, the interlayer interactions in neighboring COF_TAPB-DMTP layers are significantly higher than that in neighboring COF_TAPB-PDA layers, regardless of the extent of alignment (or misorientation). Such a significantly higher interlayer interaction results from the effective hydrogen bond formation in neighboring COF_TAPB-DMTP layers due to the existence of methoxy groups that are otherwise not available in COF_TAPB-PDA layers. Above said, we further confirm that the stacking nature (perfectly or imperfectly) is not a crucial factor that affects the interlayer sliding in COF films, and the deformation and failure mechanism revealed by the relatively well-aligned multilayer systems still holds in not-so-well–aligned multilayers of COFs. To further clarify the key mechanism of interlayer sliding in COF_TAPB-DMTP films, we also prepared multilayer COF_TAPB-DMTP films by manual stacking monolayers, as shown in SI Appendix, Fig. S28A. Such manually stacked films are layered structures without perfect alignment or entangling of adjacent layers. They are shown to have comparable mechanical properties as the LB through synthesized films (SI Appendix, Fig. S28 B–D) and similar layer number independence of mechanical properties. These results further demonstrate that the interlayer sliding of COF_TAPB-DMTP films is strongly affected by the presence of interlayer hydrogen bonds.

Given the outstanding mechanical properties and high porosity nature of 2D COFs (density of COF_TAPB-DMTP and COF_TAPB-PDA are 0.47 and 0.38 g/cc, respectively), COFs can be very promising materials for lightweight structural materials. The Young’s modulus and the breaking strength of 2D COF_TAPB-DMTP atomically thin films with designed interlayer hydrogen bonds were found to be 47.1 ± 2.3 GPa and 4.62 ± 0.29 GPa, respectively, which impressively are in the same range of many inorganic 2D materials. More importantly, compared to the reference COF_TAPB-PDA thin films with no interlayer hydrogen bonds, in which the mechanical strength significantly decreased as the layer number increased from 1L to 3L, we found that the mechanical properties of COF_TAPB-DMTP thin films were independent of the layer numbers. Furthermore, corresponding DFT simulations suggested that the interlayer hydrogen bonds significantly increase the sliding energy between adjacent layers and unveiled the interlayer interaction dictated strengthening mechanism in 2D COFs. Fig. 4A compared the specific Young’s modulus and specific strength of 2D COFs with other bulk polymers. It clearly shows that the specific Young’s modulus and breaking strength of COF_TAPB-DMTP and COF_TAPB-PDA are among the highest for polymers even compared to well-known strong polymers such as polyimide and polybenzobisoxazole. We also compared the layer-dependent strengths of 2D COFs and other 2D materials. As shown in Fig. 4B, for most 2D materials, such as graphene and MoS₂, the strength drastically decreases as the layer number increases. Such a drawback could be attributed to relatively weak interlayer interactions and significantly limits their practical applications. Unlike those 2D materials, there is no significant change in the strength of 2D COF_TAPB-DMTP as the layer number increase. The layer-independent Young’s modulus and strength enabled by designing and engineering interlayer interactions in 2D COFs strongly suggest that they could potentially retain their remarkable mechanical properties from monolayer to the bulk form, representing a significant step forward for scale-up adoption of 2D materials in a wide range of applications.

Fig. 4.

(A) Comparison of the specific strength and specific modulus of COFs and other polymers suggests the superior mechanical performance of the two types of COFs in the present study. (B) Layer-dependent strength of various 2D materials showing common decreasing trend with increasing layer numbers, in sharp contrast to the interlayer hydrogen bonds strengthened COF_TAPB-DMTP, which shows a highly desirable feature of inheritable mechanical properties from a monolayer to a multilayer stack.

In summary, we successfully synthesized large-area and high-quality 1-3L COF_TAPB-DMTP and COF_TAPB-PDA thin films by an LB trough method and measured their mechanical properties by AFM indentation. Monolayer COF_TAPB-DMTP and COF_TAPB-PDA had a Young’s modulus of 46.3 ± 2.3 GPa and 44.6 ± 3.3 GPa with a fracture strength of 4.67 ± 0.45 and 4.50 ± 0.40 GPa, respectively. Few-layer COF_TAPB-DMTP maintained similar mechanical properties as their monolayer counterparts, which is in sharp contrast to few-layer COF_TAPB-PDA. Our DFT study revealed that the methoxy groups in COF_TAPB-DMTP could enhance the interlayer interaction and lead to a higher sliding barrier and thus enhance the mechanical properties of multilayer COF_TAPB-DMTP thin films. These results indicate that COFs can be a very promising platform for the preparation of strong, tough, and lightweight high-performance 2D polymeric materials by precise molecular design.

Materials and Methods

Materials.

TAPB and DMTP were purchased from Jilin Province Yanshen Technology Co., Ltd. PDA was purchased from Sigma-Aldrich. All materials were used as received.

Preparation of Layer-Controlled COFs Thin Films.

COFs thin films were prepared at air–water interface in a Langmuir-Blodgett trough equipped with a platinum Wilhelmy plate, Taflon dipper, and barriers. TAPB was dissolved in a mix solution of chloroform and methanol (volume ratio = 3:1) with a concentration of 1 mg/mL, and 100 μL above solution was carefully spread on water surface with a micro-syringe. After 30 min evaporation of the solution, the water surface was compressed to desired surface pressure (1, 3, 10 mN/m for 1 to 3 L COFs thin films, respectively). One milliliter of dialdehyde acetic solution was added to the subphase without disturbing the TAPB layer. After 24 h of polymerization, COFs thin films can be transferred onto desired substrates by a Langmuir–Schaefer method.

AFM Indentation Test.

Single-crystal diamond tips (K-Tek, D80) were used for indentation test (1, 2, 29). The spring constant of each tip was calibrated by a thermal noise method (46). During the indentation test, the indentation speed was controlled at 0.1 μm/s.

DFT Simulation.

First-principle DFT calculations are performed using the (VASP) (44, 45). The projector augmented wave pseudopotential (47) and the generalized gradient approximation of the Perdew-Burke–Ernzerhof functional (48) are used. A plane-wave basis is set with a kinetic-energy cutoff of 500 eV. The Monkhorst–Pack (49) k-point mesh of 2 × 2 × 10 is used for structure optimization and electronic calculations of COFs. Periodic boundary conditions are applied in the x-, y-, and z-directions for all calculations. The primitive unit cells in our simulation are illustrated in SI Appendix, Fig. S1 and the lattice constants are obtained as 37.33 × 37.33 × 3.44 ų and 37.17 × 37.17 × 3.63 ų for COF_TAPB-DMTP and COF_TAPB-PDA, respectively, which is in good agreement with our experimental results. To model the sliding process, 1 × 1 × 2 primitive unit cells are constructed to demonstrate the mechanism of interfacial interaction, which contains 300 atoms for COF_TAPB-DMTP and 252 atoms for COF_TAPB-PDA. The VDW correction, which is described by DFT-D2 method of Grimme (50), is considered in all the calculations. All structures are relaxed using a conjugate gradient algorithm until the atomic forces are converged to 0.01 eV/Å.

We define the variation of energy per unit cell as $\mathrm{\Delta }E=E_{\mathit{total}}-E_{\mathit{origin}}$ , where $E_{\mathit{total}}$ is the total energy of the COFs in the interlayer sliding process and $E_{\mathit{origin}}$ represent the energy of pristine COFs. The maximum value of $\mathrm{\Delta }E$ is defined as the energy barrier of the interlayer sliding process. In the electronic calculation, the differential charge density is defined as $\mathrm{\Delta }\rho ={\rho }_{\mathit{total}}-{\rho }_{layer1}-{\rho }_{layer2}$ , where ${\rho }_{\mathit{total}}$ is the total charge density of layered COFs. ${\rho }_{layer1}$ and ${\rho }_{layer2}$ represent the charge density of upper and lower layers, respectively.

Molecular Dynamics Simulation.

The full molecular dynamics simulations using the ReaxFF potential (51) are implemented in the Large-scale Atomic/Molecular Massively Parallel Simulator simulation package (52). ReaxFF potential can be widely used to describe chemical bonds and weak interactions, such as hydrogen bonds and van der Walls force (53). Time step was set as 0.5 fs and a Nose–Hoover thermostat was used to maintain the canonical ensemble (NVT ensemble). The energy of system was minimized until the total atomic forces was converged to less than 10^–9 eV/Å. To avoid the large energy oscillation, we set the initial temperature of the layered COFs at 10K to equilibrate the system. In the process of sliding, the initial velocity of layer 1 was set as 0.002 Å/fs.

Supplementary Material

Appendix 01 (PDF)

Movie S1.

For the case of COF_TAPB-DMTP with zero misorientation angle, Layer 1 is initially placed at a random location on Layer 2. As the simulation model is relaxed sufficiently, Layer 1 slides slightly to form a perfect AA stacking with the underlying Layer 2.

Movie S2.

For the case of COFTAPB-DMTP with non-zero α (α=10°), the initial structure of the simulation model is formed by randomly placing Layer 1 on Layer 2 with the prescribed misorientation angle, so that there is minimum alignment between the two layers. As the simulation model is relaxed sufficiently, the molecular structure of both layers deforms locally by bond stretching and rotation so that the two layers can be partially aligned to form AA stacking.

Movie S3.

For the case of COF_TAPB-PDA with zero misorientation angle, Layer 1 is initially placed at a random location on Layer 2. As the simulation model is relaxed sufficiently, Layer 1 slides slightly to form a perfect AA stacking with the underlying Layer 2.

Movie S4.

After the sufficient relaxation of COF_TAPB-DMTP with zero misorientation angle, Layer 1 slides on Layer 2 in the armchair direction of Layer 2.

Movie S5.

After the sufficient relaxation of COF_TAPB-DMTP with non-zero misorientation angle (α = 10°), Layer 1 slides on Layer 2 in the armchair direction of Layer 2.

Movie S6.

Perspective view of the sliding process and bond breaking in the case of COF_TAPB-DMTP with a misorientation angle α=10°.

Movie S7.

Perspective view of the sliding process and bond breaking in the case of COF_TAPB-PDA with a misorientation angle α=10°.

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.