Supramolecular Reinforcement of a Large-Pore 2D Covalent Organic Framework

Abstract

Two dimensional covalent organic frameworks (2D-COFs) are a class of crystalline porous organic polymers that consist of covalently linked, two dimensional sheets that can stack together through non-covalent interactions. Here we report the synthesis of a novel COF, called PyCOFamide, which has an experimentally observed pore size that is greater than 6 nm in diameter. This is among the largest pore size reported to date for a 2D-COF. PyCOFamide exhibits permanent porosity and high crystallinity as evidenced by the nitrogen adsorption, powder X-ray diffraction, and high-resolution transmission electron microscopy. We show that the pore size of PyCOFamide is large enough to accommodate fluorescent proteins such as Superfolder green fluorescent protein and mNeonGreen. This work demonstrates the utility of non-covalent structural reinforcement in 2D-COFs to produce larger, persistent pore sizes than previously possible.

Two-dimensional covalent organic frameworks (2D-COFs) are crystalline, porous, organic polymer networks built from organic monomers and linked together by dynamic covalent bonds.^1–3 Researchers have explored the landscape of COF structures for use in applications such as molecular separation,^4,5 sensing,^6,7 gas storage,^8,9 electronic devices,^10,11 and catalysis.^12,13 COFs can be designed for various applications by modifying their crystallinity, pore size, and surface area from the bottom up.^3,14–16

One fundamental challenge in porous materials design involves the synthesis of structures with large, and persistent pores. There are several obstacles in this pursuit, including the solubility of large organic molecules needed to generate the large-pore sizes,^17,18 as well as the propensity for pore collapse or structural damage upon solvent removal.^19–23 As a result, 2D-COFs with a pore diameter larger than 5 nm are not common (Table S1).^14,24–26 There are additional challenges in the design of large-pore 2D-COFs as they use more flexible organic linkages^27–30 in the covalent sheets compared with more rigid metal-organic frameworks (MOFs),¹⁷ along with the potential for inefficient stacking of the layers in an eclipsed orientation^31,32 These 2D sheets are held together by non-covalent interactions such as donor-acceptor complexes,¹⁵ aromatic stacking interactions,^15,33,34 dipole-dipole interactions,^35,36 van der Waals forces,³⁷ and hydrogen bonding.^38–43 A number of methods to improve the crystallinity and surface area of 2D-COFs have been extensively studied.^15,33,44,45 COF activation methods have evolved to include the use of fluorous liquids²⁷ or supercritical carbon dioxide (scCO₂)^28–30 to remove organic solvents used during the polymerization. This strategy is beneficial as it can be used on any material and does not require any synthetic modification of the COF structure. However, more recent work has shown that by designing non-covalent interactions that are directed specifically between the layers, COFs with significantly improved crystallinity and surface area can be attained, even without scCO₂ activation.^15,26,43,45 However, as pore sizes increase, even the gentlest solvent activation methods can cause damage to the structure. In these situations, the only way to preserve these large-pores may be through a combination of structural reinforcement and improved activation methods. Here we present a supramolecular reinforcement strategy that strengthens the adhesion between the stacked 2D COF layers through the use of directed hydrogen bonding and stabilizes the large pores against collapse.

Our group recently reported a class of 2D-COFs with secondary amide sidechains that can facilitate interlayer hydrogen bonding (COFamide 1–2).⁴³ Given our previous observation that the highly correlated interlayer hydrogen bonding of the COFamide series made them resistant to pore collapse using conventional solvent activation conditions, we aimed to expand the pore size with the expectation that the rigidified layer structure could support larger-pores. In this study, we polymerized a tritopic aldehyde with amide sidechains (1) and a ditopic amine linker 4,4’-(pyrene-2,7-diyl)dianiline (3) to produce the imine-linked PyCOFamide (Figure 1). A control polymer without amide groups was synthesized (4) with methyl groups in place of the secondary amide groups (2) to investigate the importance of the interlayer hydrogen bonding on the COF structure. Polymerization reactions were carried out in a solvent mixture of o-dichlorobenzene, n-butanol and acetic acid (3M, aq) in a ratio of 1.9:1:0.1 at 120 °C for 5d to produce PyCOFamide, or polymer 4. The insoluble polymers obtained were activated using scCO₂ before further characterization. The PyCOFamide was obtained as an ash-colored powder whereas 4 was a light-yellow powder. These are insoluble in common organic solvents such as acetone, methanol, dichloromethane, and hexane.

Figure 1.

Synthesis and structure of PyCOFamide.

The FT-IR spectrum of PyCOFamide (Figure 2A) showed the appearance of a signal at 1628 cm⁻¹, which is characteristic of an imine stretching vibration. The disappearance of the aldehyde C=O stretching modes at 1697 cm⁻¹ and amine N-H vibrations at 3464 cm⁻¹ confirm the absence of starting monomers in the final polymer. Additionally, the amide N-H stretching vibrations of PyCOFamide shift to 3309 cm⁻¹ compared to the amide N-H stretching modes of 1 at 3278 cm^-1. This shift can be attributed to the formation of interlayer hydrogen bonding interactions.⁴³ We digested PyCOFamide in acidic DMSO-d₆ to determine the monomer incorporation ratio. We found the aldehyde to amine monomer ratio to be 1:1.5 which is consistent with the initial feed ratio (Figure S11).

Figure 2.

(A) FT-IR spectra of PyCOFamide and starting monomers. (B) PXRD patterns of PyCOFamide (purple), Pawley refined (red), simulated PyCOFamide (green), difference plot (black) and experimental polymer 4 (blue). (C) Nitrogen adsorption (closed circles) and desorption (open circles) isotherm for PyCOFamide (purple) and polymer 4 (blue). (D) Pore size distributions for PyCOFamide and control polymer 4. (E) SEM of PyCOFamide. (F) HR-TEM of PyCOFamide.

Powder X-ray diffraction (PXRD) experiments were performed to determine the crystallinity. Sharp diffraction peaks were observed in the scCO₂ activated PyCOFamide. An intense, and narrow diffraction peak corresponding to the (100) crystal plane was observed at 3.1° 2θ in the PyCOFamide diffraction pattern (Figure 2B). Other peaks were observed at 4.8, 6.2, 7.8, 9.4, 10.9 and 18.5° which can be attributed to the diffraction from the (110), (200), (210), (120), (220), and (001) crystal planes. The experimental and simulated PXRD patterns of PyCOFamide match well and help confirm the presence of key structural features found in the eclipsed stacking arrangement (Figure S6). The (001) reflection for PyCOFamide appears at 18.5°, which is characteristic for COFamides as they have larger interlayer spacing distances owing to the steric hindrance caused by the out of plane phenyl rings and amide groups at the node positions.⁴³ The interlayer distance predicted from the simulated crystal structure (~5.1 Å), is similar to the experimental value (~4.8 Å) obtained from PXRD. No peaks were observed in polymer 4 PXRD pattern (Figure 2B) indicating its amorphous nature due to either pore collapse or poor 2D sheet stacking.

The Brunauer-Emmett-Teller (BET) surface area and pore size distributions of PyCOFamide and 4 were measured by nitrogen adsorption measurements (Figure 2C). scCO₂ activated PyCOFamide has a BET surface area of 1682 m²/g. However, under conventional activation (washing with organic solvents and heating under dynamic vacuum >10 umHg), the BET surface area is significantly lower (8 m²/g) (Figure S5). Pore collapse or decrystallization is common in conventional solvent activation because of the capillary effect that occurs during solvent evaporation under vacuum.³⁰ In comparison, the control polymer 4 has a low accessible surface area (169 m²/g) and no observable crystallinity regardless of activation method, indicating that the hydrogen bonding interactions are key to reinforcing the eclipsed stacking mode. PyCOFamide has a type-IV isotherm, which is characteristic of mesoporous materials (pores >20 Å). PyCOFamide has a narrow pore size distribution at 61 and 67 Å (Figure 2D) that agrees with the calculated pore size (65 Å) from the computational model with eclipsed stacking (Figure S6).

Scanning electron microscopy (SEM) images of the PyCOFamide powder reveal discotic morphology (Figure 2E) with aggregated discs about 70 nm in size. The highly ordered, periodic structure of PyCOFamide was observed in the high-resolution transmission electron microscope (HR-TEM) image (Figure 2F). This clearly shows the hexagonal pore arrangement indicating long-range order in PyCOFamide. These observations are consistent with the PXRD data. The size of the hexagonal pores is also in agreement with those obtained from the nitrogen adsorption isotherm.

To test the stability of PyCOFamide, the COF was immersed in various aqueous solutions and organic solvents such as N, N-dimethylformamide (DMF), dimethyl sulfoxide (DMSO), 1,4-dioxane, N-methyl-2-pyrrolidone (NMP), sulfuric acid (1M, aq), sodium hydroxide (1M, aq) and phosphate buffered saline (PBS, pH=7.4). The retention of crystallinity after immersion was studied by PXRD (Figure S10). The results revealed that the crystallinity of PyCOFamide is retained in DMF, sodium hydroxide and PBS. However, the crystallinity of PyCOFamide is damaged in the sulfuric acid, which may be due to hydrolysis of the imine linkages under acidic conditions. Additionally, PyCOFamide crystallinity was lost in 1,4-dioxane, DMSO, and NMP solvents likely due to their polarity that could exfoliate or displace the COF layers. Interestingly, the peak at 18.5°, characteristic of the interlayer stacking distance in COFamide-based materials, remains visible in the PXRD patterns even after the other peaks have disappeared. We hypothesize that this could arise from continued stacking between disordered sheets after decrystallization.

The large pores of PyCOFamide (Figure 3A), combined with its stability in aqueous solutions like PBS encouraged us to study its ability to host proteins. Porous materials, including MOFs,^17,46–48 COFs,^49–52 mesoporous silica,^53,54 hydrogen-bonded organic frameworks,⁵⁵ zeolites,⁵⁶ and cage compounds⁵⁷ have been used to adsorb a variety of biomolecules as guests into their pores. Once inside the pores of a polymer, the reactive properties or stability can be greatly affected. However, the inclusion of biomolecules into the individual pores of COFs is less common because the pore sizes of 2D-COFs are often too small to host an entire protein. As a proof-of concept, we selected two β-barrel fluorescent proteins, Superfolder green fluorescent protein (sfGFP) ^58,59 (Figure 3C) and mNeonGreen (mNG) ⁶⁰ (Figure 3D). The approximate dimensions of selected proteins would allow for encapsulation within PyCOFamide, and the infiltration of protein into the COF pores can be easily monitored using fluorescence spectroscopy and fluorescence microscopic imaging. COF-42 (Figure 3B),⁶¹ which has a smaller pore diameter (2.3 nm) than scCO₂ activated PyCOFamide was used as a control since both proteins are too big to fit into its pores. Additionally, previous work in MOF-based protein adsorption has shown that hydrophobic sidechains, like those found in COF-42 and PyCOFamide, are favorable for the adsorption of GFP.¹⁷ We prepared samples with different COF to protein ratios and monitored the fluorescence signal of both sfGFP and mNG in solutions before and after PyCOFamide or COF-42 were added to them. In the ratio of COF to protein at 9:1, we observed that the fluorescence of the supernatant of both proteins drastically decreased when PyCOFamide was added, whereas the solutions containing COF-42 do not (Figure 3E,3F) indicating both proteins are being drawn out of solution and into the pores of PyCOFamide.

Figure 3.

Structure of (A) PyCOFamide. (B) COF-42. Dimensions of (C) sfGFP (PDB ID: 2B3P). (D) mNG (PDB ID: 5LTR). Fluorescence spectra of (E) sfGFP loading into PyCOFamide and COF-42. (F) mNG loading into PyCOFamide and COF-42. Fluorescence microscopy images of (G) sfGFP loaded COF-42 pellet. (H) sfGFP loaded PyCOFamide pellet (I) mNG loaded COF-42 pellet. (J) mNG loaded PyCOFamide pellet.

We further confirmed the inclusion of protein in the pores of PyCOFamide by using fluorescence microscopy to directly observe changes in fluorescence of the solid COFs before and after addition of protein. Some fluorescence is observable in COF-42 samples (Figure 3G, S18, 3I, S21) which is likely attributed to surface adsorption of the proteins on the COF particles, through interactions of the proteins with the functional groups at the pore edges or trapping within interparticle voids. In contrast, PyCOFamide particles are highly fluorescent indicating a greater extent of protein inclusion (Figure 3H, 3J). Since the mass of COF powder used in each of these experiments is the same, the amount of particle surface should be similar for both COF-42 and PyCOFamide. Therefore, the difference in protein adsorption between these COFs can be attributed to the fact that the large pores of PyCOFamide are accessible to the sfGFP and mNG whereas the small pores of COF-42 are not. The retention of fluorescence for sfGFP and mNG in PyCOFamide indicates that the proteins are not denatured and retain their structure after adsorption within the COF.

In conclusion, we have designed and synthesized a novel large-pore COF whose structure is stabilized through interlayer hydrogen bonding. PyCOFamide exhibits large-pore channels of >6 nm in diameter, which are among the largest reported to date in a 2D-COF. The interlayer hydrogen bonding in PyCOFamide is key to its ability to maintain its structure after activation, as similar monomers incapable of hydrogen bonding do not produce ordered COFs. The design strategies for making large-pore 2D-COFs in the future will necessitate design approaches that consider both mild activation techniques (e.g., scCO₂) and the supramolecular interactions between the layers.^{3,24,26,36,45} We have also demonstrated that large biomolecules such as fluorescent proteins can be loaded into the pores of PyCOFamide without loss of their function, suggesting that these COFs could potentially be used in the future as hosts for enzymes or biosensing proteins, or as delivery vehicles for biomolecule based therapeutics. Taken together, our study sets the stage for expanding the scope of COF chemistry providing a supramolecular design strategy to synthesize COFs with large, persistent pores.