Colloidal Covalent Organic Frameworks.

Covalent organic frameworks (COFs) are two- or three-dimensional (2D or 3D) polymer networks with designed topology and chemical functionality, permanent porosity, and high surface areas. These features are potentially useful for a broad range of applications, including catalysis, optoelectronics, and energy storage devices. But current COF syntheses offer poor control over the material's morphology and final form, generally providing insoluble and unprocessable microcrystalline powder aggregates. COF polymerizations are often performed under conditions in which the monomers are only partially soluble in the reaction solvent, and this heterogeneity has hindered understanding of their polymerization or crystallization processes. Here we report homogeneous polymerization conditions for boronate ester-linked, 2D COFs that inhibit crystallite precipitation, resulting in stable colloidal suspensions of 2D COF nanoparticles. The hexagonal, layered structures of the colloids are confirmed by small-angle and wide-angle X-ray scattering, and kinetic characterization provides insight into the growth process. The colloid size is modulated by solvent conditions, and the technique is demonstrated for four 2D boronate ester-linked COFs. The diameter of individual COF nanoparticles in solution is monitored and quantified during COF growth and stabilization at elevated temperature using in situ variable-temperature liquid cell transmission electron microscopy imaging, a new characterization technique that complements conventional bulk scattering techniques. Solution casting of the colloids yields a free-standing transparent COF film with retained crystallinity and porosity, as well as preferential crystallite orientation. Collectively this structural control provides new opportunities for understanding COF formation and designing morphologies for device applications.

Introduction

Covalent organic frameworks (COFs) are crystalline porous polymers with predictable bonding in two or three dimensions, tunable molecular structures, and high specific surface areas.[1−5] COFs are potentially attractive for selective membranes, catalyst supports, organic electronic devices, and electrical energy storage devices.[6−15] One of the major challenges to harnessing their properties is that COFs are typically formed as insoluble microcrystalline powders, which are difficult to process or not useful for these applications. Although some progress has been made in preparing COF thin films selectively and with specific crystalline orientations,[16−18] we attribute the small average crystallite sizes and tendency to form as powders to empirical screening approaches used to identify polymerization conditions for new frameworks. We recently reported the first studies into the mechanism of COF growth using boronate ester systems as a model network, which identified an irreversible precipitation during the reaction that inhibited further structural error correction and crystallite growth.[19] We hypothesized that this aggregation of crystallites was a function of both rapid, facile nucleation and intermolecular attractive forces between the crystallites.

Here we report a synthetic approach to arrest irreversible crystallite aggregation and precipitation, resulting in the first stable colloidal suspensions of COF nanoparticles. The resulting colloids show characteristic COF crystallinity, with adjustable particle size, low polydispersity, and long-term stability. We demonstrate the first use of in situ variable-temperature liquid-cell transmission electron microscopy (VT-LCTEM) with on-chip heating to directly image hundreds of individual COF nanoparticles as they form in solution. The colloidal COFs also allow postsynthetic solution casting to produce free-standing, porous COF films. We anticipate this synthetic approach to both provide a fundamental understanding of COF nucleation and growth as well as facilitate device incorporation through increased morphological control of solution processable films. Stable suspensions of porous nanoparticles also offer functional internal surfaces that may prove attractive for drug delivery, chemical sensing, or site-isolated catalysts.[20,21] More broadly for the field of solvothermal nanomaterials synthesis, these findings establish VT-LCTEM as a powerful analytical tool that complements in situ bulk scattering (light or X-ray) methods for characterizing the size, morphology, and growth processes of nanostructures in liquid.

Results and Discussion

Colloid Synthesis

Stable COF colloids form under conditions that disrupt crystallite aggregation yet do not inhibit polymerization. In our previous studies of 2D boronate ester-linked COF formation, the reaction temperature, monomer identity and concentration, and the addition of competitors each influence the rate of COF formation,[19,22] but the resulting microcrystalline powders exhibit reasonably consistent crystalline domain sizes and morphologies. Upon evaluating various cosolvents to determine their interaction with the formation of COF-5, a prototypical 2D boronate-ester linked COF, we found that CH3CN, even at modest volume fractions, prevents precipitation and provides stable colloidal COF-5 suspensions. Specifically, the condensation of 2,3,6,7,10,11-hexahydroxytriphenylene (HHTP) and 1,4-phenylenebis(boronic acid) (PBBA) to form COF-5 under homogeneous conditions (2 mM HHTP, 3 mM PBBA, 4:1 dioxane/mesitylene, 90 °C), modified to include 15–95 vol % CH3CN, provides a translucent solution that exhibits a pronounced Tyndall effect (Figures and S1A). No precipitation occurs from the reaction mixture upon cooling, even after standing for weeks. However, it is possible to sediment these particles through extended centrifugation, suggesting transient and/or reversible interactions between particles. Such interactions are not unexpected given that acetonitrile stabilizing agents stand in notable contrast to bulky or amphiphilic ligands often used in nanoparticle syntheses. Colloid formation is specific to nitrile cosolvents, whereas other cosolvents (e.g., CH2Cl2, THF, toluene) result in COF-5 precipitation (Figure S1B and Table S1). Benzonitrile also stabilizes colloid growth, suggesting that a direct interaction between the nitrile functional group and the COF is responsible for colloid formation. Control experiments, in which each monomer was subjected to the polymerization conditions in the absence of the other, show no evidence of nanoparticle formation, ruling out the formation of simple monomer aggregates. Taken together, these observations indicate that nitrile solvents uniquely inhibit the precipitation of COF-5 as a microcrystalline powder. These reaction conditions provide a previously unattainable means to characterize COF formation and allow for subsequent solution processing.

Figure 1

(A) Typical growth conditions for boronate ester-linked 2D COFs provide insoluble, polycrystalline powders. CH3CN cosolvents instead yield stable colloidal nanoparticles of the crystalline polymer networks. (B) Colloid formation as a function of solvent CH3CN percentage (2 mM HHTP, 1.5 equiv of PBBA, 15 equiv of CH3OH, remaining solvent 4:1 dioxane/mesitylene, 90 °C).

In-Situ Small-Angle and Wide-Angle X-ray Scattering

The crystallinity and size of the COF-5 colloidal nanoparticles were characterized by small-angle and wide-angle X-ray scattering (SAXS/WAXS). Aliquots of polymerizations performed in the presence of 55% CH3CN (2 mM [HHTP]0, 90 °C) were monitored at various time points to simultaneously characterize the increase of particle size and crystallization (Figure ). COF crystallization is observed within 15 min, as determined by the appearance of a peak at 0.24 Å–1, corresponding to the <100> diffraction peak of COF-5. The integrated area of this peak increases for the initial 2 h of the polymerization, which matches the time scale of the increase in particle size calculated from the SAXS data, suggesting that the earliest observable boronate ester-linked particles are already crystalline. This behavior is consistent with our previous study of boronate ester-linked COF precipitation, in which powders isolated in the first few minutes of the polymerization were both microcrystalline and showed high BET surface areas (>2000 m2/g).[19] In contrast, 2D imine-linked COFs form an initial amorphous phase that crystallizes slowly over extended reaction times.[23]

Figure 2

(A) SAXS/WAXS of in situ COF-5 colloid formation over time, with inset of the highlighted <100> diffraction peak (growth conditions: 2 mM HHTP, 1.5 equiv of PBBA, 15 equiv of CH3OH, solvent 55% vol % CH3CN, 90 °C). (B) Particle size and <100> diffraction integrated area as a function of time. (C) WAXS diffraction peaks of boronate ester COF colloids prepared using various boronic acid linkers (red: COF-5; green: COF-10; blue: Py-COF; purple: HHTP-DPB COF), indicating the generality of this procedure.

The colloidal COF nanoparticles have relatively uniform size distributions, as is evident from the measurable SAXS form factor (Figures and S4). Particle sizes are quantifiable at 25 min at a mean radius of 21 nm, which increases to 30 nm in the initial 2 h. This increased particle size is accompanied by an increase in the intensity of the <100> reflection, which suggests that the particle growth occurs through polymerization and crystallization processes rather than particle aggregation. Furthermore, the particle size distribution narrows as the average crystallite size increases, characterized by the full-width half max (fwhm) of the <100> peak, which is also inconsistent with particle aggregation (Figure S4). These observations indicate that CH3CN stabilizes discrete crystallites in solution and inhibits their aggregation. The earliest observed particle sizes are comparable to the 20–40 nm crystalline domains determined by X-ray diffraction of COF-5 precipitated powders previously synthesized in the absence of CH3CN.[19] Concurrently, the standard deviation of the particle size distribution narrows as the particles grow. COF crystallites undergo further growth after nucleation, which slows over time, presumably due to monomer consumption. CH3CN stabilizes colloids of several boronate ester-linked COFs, indicating the generality of this procedure. HHTP was reacted with 4,4′-biphenylbis(boronic acid), 2,7-pyrenebis(boronic acid), and 4,4′-diphenylbutadiynebis(boronic acid) under similar conditions to yield colloids of COF-10,[24] TP-COF,[25] and HHTP-DPB COF,[26] respectively. The crystallinity of each COF colloidal suspension was confirmed by the presence of its <100> diffraction peak in the WAXS pattern (Figure C). The feature at 0.16 Å–1 corresponds to the <100> of HHTP-DPB COF, in addition to the <110> peak observed at 0.27 Å–1. The WAXS patterns for COF-10 and TP-COF are consistent with the expected structures and resolve the slight difference in the length between the biphenyl and pyrene linkages. The kinetics of DPB-HHTP COF colloid formation were similar to that of COF-5, with framework crystallinity again observed at the earliest time points at which particle size is detectable (Figure S5). These combined observations demonstrate the generality of the colloidal stabilization of 2D boronate ester-linked COFs by the CH3CN cosolvent. It is therefore likely that CH3CN interacts directly with the boronate linkages to attenuate attractive forces between crystallites.

Colloid Size Characterization and Control

Dynamic light scattering (DLS) was used to probe the long-term stability of COF-5 colloidal suspensions and the effect of the polymerization conditions on particle size. The average particle size and polydispersity index were measured for reaction mixtures of varying monomer and CH3CN concentrations after their polymerization at 90 °C for 20 h (Figure and Table S2). For all CH3CN concentrations examined (15–95 vol % CH3CN), the colloids exhibit Gaussian size distributions and low polydispersity indices. The average particle size is relatively invariant at CH3CN concentrations above 55 vol %, between 45–60 nm. This size is comparable to the 20–40 nm crystallite sizes observed by powder X-ray diffraction (PXRD) of COF-5 precipitates and suggests stabilization of discrete crystallites at high CH3CN concentrations. Larger colloids (100 and 240 nm) are obtained at 15 and 35 vol % CH3CN, respectively. These larger particle sizes may originate from the aggregation of discrete, smaller crystallites at lower CH3CN content, which would represent a transition from insoluble material with no added CH3CN to dispersed, smaller colloids at higher CH3CN content (Figure ). COF colloid growth conditions yield monodisperse particle distributions, with DLS PDI < 0.2. The initial monomer concentration has a minor influence on the particle size, with an average of 60 nm at 2 mM [HHTP]0 and 75 vol % CH3CN. This average decreases to 40 nm at 1 mM [HHTP]0 under similar reaction conditions (Figure B).

Figure 3

Average size of COF-5 colloidal nanoparticles prepared under different reaction conditions, characterized by the DLS Z-average, where error bars correspond to the polydispersity width. (A) Size dependence as a function of the volume fraction of CH3CN (remainder is 4:1 dioxane/mesitylene). (B) Size dependence as a function of [HHTP]0 at 75% vol % CH3CN.

Once formed at a given CH3CN concentration, the colloids are stable for at least one month at room temperature (Table S2). Furthermore, the particle sizes do not change if the CH3CN concentration is varied after the colloids have been formed. For example, 230 nm colloids formed at 15 vol % CH3CN do not shrink when additional CH3CN is added to the solution. Likewise, 50 nm colloids do not increase in size when diluted by non-nitrile containing cosolvents, even upon heating at 90 °C (Figure S6). These experiments demonstrate that COF colloids are stable and robust once formed. Therefore, the particles are likely not in rapid equilibrium but rather kinetically trapped without appreciable monomer exchange. This suggests that crystallite size limitation in COF growth is not due to precipitation, but rather a lack of Oswald ripening postnucleation. Colloid stability provides the opportunity to explore a variety of postsynthetic processing to characterize and control overall morphology.

To further explore the size of colloidal particles, solvated colloidal nanoparticles were dried and analyzed by atomic force microscopy (AFM). COF-5 colloids are readily drop cast onto mica, yielding discrete surface-adsorbed species (Figure ). COF nanoparticles size trends are consistent with DLS measurements, with smaller particles observed at CH3CN solvent levels >55%. In particular, average heights at 95% CH3CN are <6 nm, suggesting few-layer thick COF domains. The nanoparticle shape at low CH3CN vol % (<55%) is consistent with crystallite aggregation (Figures A and S7). In general, particle heights are comparable to but smaller than the solvated sizes obtained by DLS (Figure C). Size distributions by AFM also show a narrow dispersity. Collectively, the size, shape, and surface density results suggest that the larger particles observed at low CH3CN concentrations are aggregates of the growing crystallites, whose aggregation is inhibited at higher CH3CN concentrations.

Figure 4

COF-5 colloidal nanoparticles, size characterized by AFM. (A) Example COF-5 nanoparticles (35% CH3CN). (B) Example COF-5 nanoparticles (75% CH3CN). (C) Particle height as a function of solvent CH3CN percent, where the error bar indicates the Gaussian fit RMS width.

Liquid-Cell Transmission Electron Microscopy

We correlate the bulk particle-size characterization of COF colloid growth/stabilization with VT-LCTEM, whereby we image/video individual nanoparticles in real-time under the polymerization conditions. First, we performed electron-beam sensitivity and damage tests of the COF precursor solutions (see Supporting Information), which indicated that the solution is electron beam sensitive at elevated temperatures (i.e., conditions at which COF growth is initiated). The precursor solutions were found to be comparatively stable at room temperature under low/modest electron dose rates (Figure S11, Figure S12, and Table S3).[27−30] No COF nanostructure nucleation or growth is observed during prolonged, continuous illumination at 27 °C (no radiolysis-induced growth). However, continuous e– illumination influences the thermally activated COF nanostructure growth once the temperature of the liquid-cell has been heated above 80 °C (Video S2 and Figures S13 and S14). Because of the acute sensitivity of the solution to radiolysis damage at elevated temperatures, we employ time-lapse VT-LCTEM, whereby single snapshot images are periodically acquired, with the beam off between acquiring images, instead of continuous video (Figure , Figure S17 and Table S4).

Figure 5

VT-LCTEM experiment. (A) Time series of bright-field LCTEM images as a COF precursor solution (55 vol % CH3CN) is heated to 80 °C. The first frame is the COF precursor solution at room temperature (no nanostructures), and subsequent frames show a region of interest upon heating the solution to 80 °C using on-chip in situ variable-temperature control. (B) Discrete particle diameter, as measured from each LCTEM image using multiobject image analysis. The average particle size (red square) is ∼15–20 nm, which is reached within 10 min at 80 °C. These particle sizes remain stable over extended time periods at 80 °C. Error bars indicate standard error (narrow particle-size distribution, ∼ ±3 nm).

The average discrete colloidal particle size (17.5 ± 3.5 nm diameter) directly measured by LCTEM imaging (Figure and Figure S17) is consistent with the particle size calculated from the in situ SAXS/WAXS bulk-scattering data (Figure ) obtained after 20 min at 90 °C (35 nm diameter). The discrete VT-LCTEM-formed nanoparticles match the COF-5 particles grown via bulk-solution synthesis with respect to their average size and rounded morphology (Figure S15). However, the average VT-LCTEM particle-size remains constant over time at elevated temperatures, while the particle size measured by SAXS/WAXS increases for the first few hours before plateauing at ∼60 nm diameter (2 h at 90 °C). This discrepancy might arise from differences in how these measurements are conducted (analysis of reaction aliquots compared with particles adhered to the windows of a microfluidic setup) and will be investigated further in future experiments. The uniform, radial growth of individual seed particles without particle–particle coalescence or ripening, as well as the time-lapse image-series of the size-stabilized, discrete particles (Figure and Figure S17) suggest that COF-5 nanoparticles grow by monomers or oligomers adding to particles attached to the TEM window. Similar behavior was observed in the room temperature growth of mesoporous MOF nanoparticles (zeolitic imidazolate framework-8, ZIF-8).[31] The continued development of low-dose methods is needed to enable continuous VT-LCTEM imaging, as many other systems of interest are likely to experience similarly increased sensitivity to beam damage at elevated temperatures.

Solution Casting of Freestanding COF Films

These stable colloidal suspensions of COFs offer a new means to process these materials into thin films from solution, in contrast to the poor processability of the microcrystalline powder form. 2D COFs can be obtained as thin films by including a substrate in the polymerization mixture.[16,32,33] However, this practice requires that the substrate to be compatible with the polymerization conditions and is inherently inefficient—most of the building blocks lost to the powder form. In contrast, the solution casting of COF colloids represents an efficient means to form free-standing films of controlled size and thickness. COF-5 films were obtained by evaporating solvent from a colloidal suspension at 90 °C, resulting in a translucent material upon near-complete evaporation (Figure ). Films >10 μm in thickness are readily obtained, which delaminated from the glass substrate to provide a freestanding, brittle film. Films generated this way retain the crystallinity of the original colloidal solution, as determined by PXRD (Figure S10). The COF-5 film had a BET surface area of 840 m2/g, as determined by N2 adsorption (Figure ), compared to ca. 2000 m2/g for the highest reported microcrystalline powders.[19] Nonlocal density functional theory (NLDFT) analysis indicated the expected pore size of 2.5 nm, consistent with COF-5 nanocrystalline powders.

Figure 6

(A) Solution casting of colloid yields a coherent, free-standing COF film. (B) Optical image of transparent freestanding COF-5 film. (C) SEM of freestanding film.

The free-standing COF film exhibited a preferential orientation opposite to that of films grown directly on the substrate, as determined by grazing incidence X-ray diffraction (GI-XRD). COF-5 film displays increased electron density near Q∥ = 0 for peaks corresponding to <100>, <110>, and <200> diffractions (Figure C). Moreover, the <001> diffraction, corresponding to interlayer stacking of the 2D COF sheets, shows decreased electron density near Q∥ = 0. These observations indicate a crystallite preferred orientation with stacking direction of the 2D sheets perpendicular to the plane of the film. In contrast, COF films grown on graphene or other substrates are typically vertically oriented, with the stacking direction normal to the substrate.[16] Thus, the printing of colloid films through solution casting yields a structural morphology dramatically different from surface-supported thin film growth. Such orientations are of interest for applications in which charge or molecular transport occurs parallel to a substrate, such as transistors or microfluidic devices. Through the further study of the film aggregation mechanism, established rules for targeting either pore arrangement will provide increased control over COF morphology and expand the scope of accessible applications.

Figure 7

Characterization of freestanding COF-5 film prepared from 55% vol % CH3CN conditions. (A) Nitrogen absorption surface area. (B) NLDFT pore size distribution. (C–D) GI-XRD showing preferred crystallite orientation data. (E) Projection of film diffraction near Q∥ = 0. (F) Illustration of the preferred orientation of COF pores parallel to the film plane.

Conclusion

Nitrile-functionalized cosolvents profoundly change boronate ester-linked COF polymerizations to provide colloidal suspensions instead of insoluble microcrystalline powders. The colloids remain dispersed in the growth solvent for at least one month with no evidence of aggregation. In situ SAXS/WAXS analysis indicated that the colloids show the expected crystalline 2D lattices from the earliest measurable time points with complementary VT-LCTEM experiments confirming the growth kinetics and morphology of the resulting nanoparticles. These stable colloidal suspensions can be used as an ink to solution process COF films, which show preferred orientation opposite those of films grown directly on substrates. The films retain their crystallinity and much of the permanent porosity. As such, these findings represent an important development for the convenient solution processing of polymers with 2D topologies. One of the most important challenges in the synthesis of COFs is to increase the average crystallite size and understand their nucleation and growth processes rigorously. Further study of COF colloid formation and growth will address these questions.