Seeded growth of single-crystal two-dimensional covalent organic frameworks

See allHide authors and affiliations

Science  06 Jul 2018:
Vol. 361, Issue 6397, pp. 52-57
DOI: 10.1126/science.aar7883

Covalent organic frameworks writ large

Covalent organic framework (COF) materials have been difficult to characterize structurally and to exploit because they tend to form powders or amorphous materials. Ma et al. studied a variety of three-dimensional COFs based on imine linkages (see the Perspective by Navarro). They found that the addition of aniline inhibited nucleation and allowed the growth of crystals large enough for single-crystal x-ray diffraction studies. Evans et al. describe a two-step process in which nanoscale seeds of boronate ester–linked two-dimensional COFs can be grown into micrometer-scale single crystals by using a solvent that suppresses the nucleation of additional nanoparticles. Transient absorption spectroscopy revealed superior charge transport in these crystallites compared with that observed in conventional powders.

Science, this issue p. 48, p. 52; see also p. 35


Polymerization of monomers into periodic two-dimensional networks provides structurally precise, layered macromolecular sheets that exhibit desirable mechanical, optoelectronic, and molecular transport properties. Two-dimensional covalent organic frameworks (2D COFs) offer broad monomer scope but are generally isolated as powders comprising aggregated nanometer-scale crystallites. We found that 2D COF formation could be controlled using a two-step procedure in which monomers are added slowly to preformed nanoparticle seeds. The resulting 2D COFs are isolated as single-crystalline, micrometer-sized particles. Transient absorption spectroscopy of the dispersed COF nanoparticles revealed improvement in signal quality by two to three orders of magnitude relative to polycrystalline powder samples, and suggests exciton diffusion over longer length scales than those obtained through previous approaches. These findings should enable a broad exploration of synthetic 2D polymer structures and properties.

Polymerizing monomers into periodic two-dimensional (2D) networks yields macromolecular sheets linked by robust covalent bonds (14). Such polymers differ from established linear, branched, or cross-linked polymer architectures and promise distinct combinations of properties that emerge from their designed structures and uniform, nanometer-scale pores (4, 5). This promise has been hindered by the limited materials quality provided by current 2D polymerization methods (4, 6). Previous work has resulted in 2D polymers with large crystalline domains, obtained by crystallizing tritopic monomers into layered crystals capable of topochemical photopolymerizations, some of which have been characterized as single-crystal to single-crystal transformations (7, 8). But the strict geometric requirements of topochemical polymerizations and the difficulty of designing molecules that crystallize appropriately have limited the generality of this approach to just nine monomers across three design motifs.

In contrast, the simultaneous polymerization and crystallization of monomers into 2D polymers known as covalent organic frameworks (COFs) is more general, with more than 200 reported examples (920). However, 2D COFs typically form as insoluble, polycrystalline powders or films with small crystalline domains (typically smaller than 50 nm), which we attribute to poorly controlled nucleation and growth processes (2123). Wuest and co-workers reported three 3D COFs isolated as large single crystals (24), but these are linked by weak azodioxy bonds, are unstable to elevated temperatures and solvent removal, and have not been generalized to 2D COF single crystals. Here, we report 2D COFs as discrete particles comprising single-crystalline domains with sizes ranging from 500 nm to 1.5 μm. They were prepared through a two-step approach that separates the nucleation and growth processes. These 2D COF samples have superior properties and can be studied more rigorously than polycrystalline samples, as demonstrated by a transient absorption study that provides improved signal quality by three orders of magnitude and offers evidence for exciton delocalization at length scales larger than the crystalline domains of the powders. These findings represent a major advance in 2D COF materials quality that will greatly broaden the monomer scope available for accessing well-defined 2D polymers.

2D COF formation is poorly understood because effective reaction conditions are identified by empirically screening for the formation of the desired materials as polycrystalline powders. The appropriate rates of polymerization and exchange processes needed for defect correction are unknown, and we postulate that any attempt to identify reactions that produce polycrystalline powders is inherently predisposed to identify conditions with uncontrolled nucleation. Our recent mechanistic studies of boronate ester–linked 2D COF formation revealed a nucleation-elongation process that was interrupted by the aggregation and precipitation of microcrystalline COF powders (22). We later found that nitrile-containing cosolvents prevented aggregation and precipitation, resulting in stable colloidal suspensions of 30-nm crystalline 2D COF nanoparticles that could be solution-processed into films (25). Here, we enlarged the COF colloids into faceted single crystals with lateral dimensions greater than 1.5 μm through a second growth step in which additional monomers were introduced slowly to suppress further nucleation. This approach is shown to be general for three boronate ester–linked COFs, including a pyrene-containing structure previously shown to be photoconductive (26). These seeded 2D COF polymerizations provide 2D COFs as nonaggregated single crystals and demonstrate a means to obtain high-quality 2D polymers using modular and versatile directional bonding approaches.

The solvothermal condensation of 1,4-phenylenebis(boronic acid) (PBBA) and 2,3,6,7,10,11-hexahydroxytriphenylene (HHTP) in a mixture of 1,4-dioxane and mesitylene (4:1 v/v) affords a boronate ester–linked 2D COF (COF-5), which precipitates within minutes (22). However, when 80 volume percent (vol %) of CH3CN is included as a cosolvent, a stable colloidal suspension of COF-5 nanoparticles is obtained instead (25). Having identified 2D COF polymerization conditions that prevent aggregation and precipitation, we hypothesized that newly introduced monomers would add to the existing nanoparticles rather than nucleating new particles under appropriate conditions. Indeed, adding monomers gradually to the reaction mixture limited the steady-state monomer concentration, and the colloids grew without forming new particles. In contrast, when the monomers were introduced quickly, their concentration increased above a critical nucleation concentration, and the reaction was dominated by the formation of new particles (Fig. 1).

Fig. 1 Schematic of controlled 2D polymerization.

A two-step seeded growth approach provides 2D COF single crystals. When HHTP and a linear bis(boronic acid) monomer are condensed in a solvent mixture containing CH3CN, crystalline 2D COF nanoparticles are formed as stable colloidal suspensions. These nanoparticles are enlarged in a second polymerization step in which the monomers are slowly added to the solution. If the monomers are added more quickly, their concentration increases above a critical nucleation threshold, which leads to uncontrolled nucleation and smaller average particle size.

We prepared an initial COF-5 colloidal suspension by heating HHTP and PBBA to 90°C in CH3CN:1,4-dioxane:mesitylene (80:16:4 vol %) for 18 hours. The formation of crystalline COF-5 colloids with an average diameter of 30 nm was confirmed by dynamic light scattering (DLS) and synchrotron small-angle and wide-angle x-ray scattering (SAXS/WAXS) experiments. At first, we screened conditions that provide seeded growth by adding monomers at various concentrations and rates and monitoring colloid size by DLS. These exploratory experiments were successful and provided COF-5 nanoparticles that had grown from 30 nm to 400 nm (see supplementary materials) (fig. S2). The 400-nm particles were used in more careful growth studies (depicted in Fig. 2, A to C) because they are sufficiently large to enable easy differentiation of newly formed colloids from the preexisting seeds using DLS and other tools.

Fig. 2 The monomer addition rate defines seeded growth and nucleation regimes.

(A) The DLS number-average size of COF-5 particles as a function of added monomer equivalents. (B) Wide-angle x-ray scattering of COF-5 particles as a function of the amount of added monomers at the two monomer addition rates. (C) DLS number-average size distributions obtained at three points shown in (A). (D) DLS number-average size distributions for the initial and final particle sizes of COF-10 during slow monomer addition. (E) WAXS traces of COF-10 particles as a function of the amount of added monomers. (F) DLS number-average size distributions for the initial and final particle sizes of TP-COF during slow monomer addition. (G) WAXS traces of TP-COF particles as a function of the amount of added monomers.

We observed both seeded growth and new particle formation regimes by adding monomers at different rates to dispersions of the 400-nm colloids. A COF-5 colloidal suspension was heated to 85°C and separate solutions of PBBA (6 mM) and HHTP (4 mM) were simultaneously added using a syringe pump. The size distributions of the COF-5 nanoparticles were monitored by DLS and are plotted as a function of molar equivalents of the monomers added relative to those present in the initial colloids (equiv HHTP added/equiv HHTP in the COF-5 seed solution; Fig. 2). When monomers were added slowly (0.10 equiv hour−1), the average particle size steadily increased (Fig. 2A); the size distribution remained monomodal as particle size increased (Fig. 2C). The added monomers were incorporated into the existing colloids, which grew from 400 to 1000 nm after addition of 4.0 equiv HHTP, and no evidence for new nanoparticle formation was observed by DLS (Fig. 2C), x-ray diffraction, or microscopy (see below). In contrast, when the rate of monomer addition was increased by a factor of 10 (1.0 equiv hour−1), the average particle size began to decrease after addition of 0.5 equiv of monomers (Fig. 2A). The particle size distribution was bimodal with a new population at 50 nm after addition of 0.8 equiv HHTP, and this population of smaller particles developed into the dominant signal in the DLS measurements as more monomers were added (Fig. 2C). Control experiments in which the reaction solvent lacking HHTP and PBBA was added to the colloids resulted in no change of the size distribution of the colloids (fig. S3), indicating that the increased particle size observed during slow monomer addition is not attributable to Ostwald ripening or other particle fusion processes.

These experiments indicate that the faster addition rate causes the monomer concentration to exceed that needed to nucleate new nanoparticles. As indicated by the point at which the DLS curves for the fast and slow addition experiments diverge, HHTP and PBBA concentrations of ~1 mM represent an upper bound needed for nucleation of new COF-5 particles under these conditions. However, this estimate of the critical nucleation concentration is likely to be influenced by several reaction parameters. These experiments demonstrate that COF-5 colloidal suspensions remain stable and available for continued 2D polymerization when the monomer concentrations are sufficiently low.

Simultaneous WAXS, which characterizes the crystallinity of the colloids, and SAXS, which tracks the average particle size, also differentiate the seeded growth and new particle formation regimes. In the seeded growth regime, the average crystallite size and particle size increase as monomers are added slowly. In contrast, the average crystallite size and particle size decrease when monomers are added quickly. The WAXS data indicate that the initial COF seeds diffracted at q = 0.24 and 0.42 Å−1, corresponding to the 100 and 010 Bragg directions of a hexagonal lattice with in-plane lattice parameters a = b = 29.9 Å (fig. S6). These parameters match well with calculated (30.0 Å) and measured (29.7 Å) values previously reported for COF-5 powders (27). When monomers were added slowly, the 100 peak intensified and sharpened, and the 200 and 210 Bragg peaks became visible at q = 0.48 and 0.64 Å−1 after addition of 0.8 equiv HHTP. These observations indicate that the size of the COF-5 crystalline domains and the crystallinity within the domain both increased as the monomers were added to the solution. However, when monomers were added at 1.0 equiv hour−1, the WAXS peaks broadened and decreased in intensity; this observation is consistent with the nucleation of new particles with smaller crystalline domains. The SAXS results were fully consistent with the DLS studies described above, validating that the WAXS experiments were done under conditions relevant to each growth regime. The SAXS traces (fig. S13) trended toward higher intensities at low scattering angles at the slow monomer addition rate (0.10 equiv hour−1), which indicates an increased average particle size. SAXS of solutions with rapid monomer addition (1.0 equiv hour−1) trended to lower intensities at low scattering angles, consistent with decreased average particle size. Together, these experiments demonstrate that the average crystalline domain size of the COF-5 colloids increases along with the particle size as monomers are added sufficiently slowly to suppress nucleation.

We demonstrated the generality of the seeded growth approach for two other boronate ester–linked 2D COFs, COF-10 and TP-COF (Fig. 1), which are synthesized by the condensation of HHTP with 4,4′-biphenylbis(boronic acid) and 2,7-pyrenebis(boronic acid), respectively. Both systems exhibited growth behavior similar to that of COF-5. For each network, an initial colloidal suspension was generated by heating HHTP and the corresponding boronic acid in a mixture of CH3CN:1,4-dioxane:mesitylene (80:16:4 vol %) for 18 hours at 90°C. Once the colloids were formed, separate solutions of HHTP and the diboronic acid were added at 0.10 equiv hour−1 to the 2D COF nanoparticle suspensions. DLS of the reaction solutions showed monomodal particle size distributions that shifted to larger average sizes as monomers were added. The DLS particle size of the COF-10 colloids shifted from 80 nm to 190 nm after 1.5 equiv HHTP and 2.25 equiv 4,4′-biphenylbis(boronic acid) were added (Fig. 2D). The TP-COF particle size (Fig. 2F) increased from 400 nm to 750 nm after 1.50 equiv HHTP and 2.25 equiv 2,7-pyrenebis(boronic) acid were added at 0.10 equiv HHTP hour−1. These experiments suggest that nucleation is suppressed under these conditions and that the existing COF domains are enlarged through addition of the added monomers. Furthermore, nucleation predominated when the monomer solutions were added quickly (1.0 equiv hour−1), resulting in smaller COF particle sizes of approximately 50 nm (figs. S4 and S5).

We also used in-solution x-ray diffraction techniques to interrogate the crystallinity of these colloidal suspensions. For TP-COF, WAXS peaks were observed at q = 0.19, 0.34, 0.39, and 0.52 Å−1 (Fig. 2G), corresponding to the Bragg directions 100, 110, 200, and 210 that were reported in its powder pattern (fig. S8) (26). Likewise, COF-10 colloid diffraction peaks were observed at q = 0.20, 0.34, and 0.52 Å−1 (Fig. 2E), which correspond to the 100, 110, and 210 Bragg diffractions (fig. S8) (28). These observations are consistent with those of the seeded growth of COF-5, in which the intensity of the diffraction peaks increased and the width (as judged by their full width at half of the maximum intensity) decreased. Here again, the 200 and 210 diffraction peaks became visible after addition of 0.8 equiv HHTP to each initial sample. These observations indicate that as monomers are introduced to COF nanoparticles sufficiently slowly, they add to and enlarge existing crystalline domains.

The enlarged COF nanoparticles were heated at 85°C for 14 days and then analyzed by low-dose, high-resolution transmission electron microscopy (HRTEM) to visualize their morphology, size, aspect ratio, and crystallinity (Fig. 3). TEM imaging of discrete particles was possible for COF-5 and COF-10, with TP-COF appearing aggregated when prepared on TEM substrates; this precluded close analysis of its lattice structure by microscopy (fig. S9). TEM imaging at low magnification revealed that the COF-5 particles (Fig. 3A) have uniform six-fold symmetry and hexagonal faceting in projection. Most individual particles were 300 to 500 nm in diameter, with some reaching >1 μm. The COF-5 particles were observed at random orientations by TEM (no preferential orientation when drop-cast on TEM substrates), and all orientations appeared dimensionally isotropic. Lattice-resolution images of individual COF-5 particles show that they are single-crystalline (Fig. 3, B to D), as consistent and continuous lattice fringes extend throughout the particles. For the particle selected in Fig. 3, B to D, which is tilted just off a zone axis, the fringe spacing is ~10 Å (fitting d210 for COF-5), as measured by fast Fourier transform (FFT; Fig. 3B, inset) and the left-to-right intensity profile (Fig. 3D) of the particle. The magnified lattice images of vertically aligned regions in this particle show the continuous single-crystalline structure that extends vertically (parallel with the fringes) from edge to edge, as shown in several regions of a particle juxtaposed in Fig. 3C. The intensity profile (Fig. 3D) similarly shows that the single-crystalline structure is continuous horizontally (perpendicular to the fringes) from edge to edge.

Fig. 3 Low-dose HRTEM characterization of COF single-crystalline particles.

Cumulative dose per image: ~25 e Å–2 s–1. (A and E) Low-magnification images of COF-5 and COF-10 particles, respectively. (B) Lattice-resolution HRTEM image of a COF-5 particle with consistent lattice fringes extending across the entire particle. Inset: FFT of the image, cropped at the predominant fringe spacing (~10 Å, d210). (C) Four regions of interest at higher magnification corresponding to the green, teal, red, and magenta boxed regions of the particle in (B), which are aligned vertically and parallel with the 10 Å fringes. (D) Intensity profile plot (left to right across the particle) of the image in (B) after applying a bandpass filter. The periodicity of the intensity profile is constant and continuous with a period of ~10 Å. (F) Overview image of one COF-10 particle, from which higher-magnification images were acquired at four regions of interest (green, teal, magenta, and red). (G) Lattice-resolution HRTEM image at the lower right edge of the particle in (F) where 8.3 Å lattice fringes are resolved, corresponding to d400. (H) The HRTEM image in (G) after applying a Fourier filter to select the central spot and the two 8.3 Å spots in the FFT. (I) The FFTs of the four high-magnification images from the four color-coded regions of interest in (F).

The COF-10 particles (Fig. 3, E to I) were also highly uniform in size and morphology. They were hexagonally faceted in projection yet lacked the six-fold symmetry of the COF-5 particles; their sizes were predominantly 4 to 5 μm across their major length and ~3 μm across their minor length (Fig. 3E). When prepared on TEM substrates, the particles preferentially oriented with their intersheet stacking dimension normal to the substrate, which was likely caused by their z-dimension (particle thickness) being smaller than their lateral dimension (fig. S10). Because of the large size of these particles (thickness ~0.5 to 1 μm), analysis was limited to their edges; many particles were too large for high-resolution TEM characterization. Therefore, we acquired high-magnification images at various regions of interest around the perimeter of a single particle that was sufficiently thin (Fig. 3, F to H). Lattice fringes (Fig. 3, G and H) could not be visually resolved in the images for all regions, but the FFTs all contained more subtle periodic information, which revealed that all regions were consistently crystalline throughout the particle (Fig. 3, H and I). The FFTs of all four regions contained ~8.3 Å fringe spots (fitting d400 for COF-10) at the same radial location (Fig. 3I), indicating consistency of the crystal structure and crystal orientation at each of the four regions of interest (fig. S11). These observations strongly suggest that the particle is a single continuous crystalline domain. We note that low-magnification images of the COF-10 particles contain very prominent diffraction contrast with no detectable grain boundaries, which is consistent with the particles being single crystals (Fig. 3E). Collectively, the TEM characterization of the individual, single-crystalline COF-5 and COF-10 particles further indicates that the well-controlled, seeded growth procedure represents a major advance in COF materials quality and the modular synthesis of 2D polymers.

The well-dispersed, single-crystalline COF colloids enable characterization of emergent electronic properties of these 2D layered assemblies that were not previously possible in polycrystalline aggregates. The optical signatures and exciton dynamics of the COF-5 colloids were characterized using optical transient absorption (TA) spectroscopy measurements. Delocalized excitons and charge carriers in 2D COFs have been of interest since the earliest reports of these materials, but their small crystalline domain sizes and aggregated powder forms have severely limited their characterization to qualitative photoconductivity measurements (11, 29). Previous TA experiments on 2D COFs were performed either on polycrystalline samples dispersed in solvent or on polycrystalline thin films (3032). Although the first reported TA measurements showed reasonable data quality, they were inconsistent with more recent results from the same group on the same systems, which were noisy and dominated by scattering (30, 32). Similar noise sources were also observed in our TA measurements on COF-5 powder samples obtained by solvothermal growth (fig. S12). In comparison, the TA data from our dispersed COF-5 colloids show signal-to-noise enhancement by approximately three orders of magnitude relative to the powder sample measured under the same experimental conditions (Fig. 4B). The well-defined structures of these colloids now make it possible to correlate their optical TA spectra to exciton dynamics and decay pathways inherent in the nature of the COF itself.

Fig. 4 Exciton diffusion studies of COF-5.

(A) Depiction of exciton distribution in COF-5 single crystals of different sizes. Exciton interactions at boundaries are highlighted. (B) Transient absorption spectra (excitation wavelength 360 nm) as a function of indicated probe delay time for 110-nm COF-5 colloids. (C) Exciton decay kinetics (dots) and fits (lines) of COF-5 colloids of different sizes at high photon fluence (91 μJ/pulse) observed at 410 nm. Inset: Exciton decay kinetics (dots) and fit (line) at low photon fluence (3 μJ/pulse) observed at 410 nm.

TA spectroscopy was performed on COF-5 colloids of varying particle size and as a function of pump fluence. These spectra were dominated by a broad excited-state absorption that decayed on a time scale of hundreds of picoseconds while remaining spectrally consistent, as well as a stimulated emission feature that was longer than the time scale of the experiment and was subtracted to ensure consistent baselines for the time scale of observation. The identity of this feature is likely related to the singlet excitons of the triphenylene cores. However, dynamics clearly differed according to the particle size for high photon fluence (Fig. 4C), with the emergence of faster decays that were not observed at the low photon fluences (Fig. 4C, inset). We attribute this high fluence, fast exciton decay kinetics, and particle size dependence at the high excitation photon fluences to exciton-exciton annihilation, which occurs when multiple excitons within a diffusion length of each other are excited within the same particle sufficiently close to each other so as to diffuse and interact (3335). At a given number density, excitons in smaller COF crystallites more readily undergo exciton-exciton annihilation because they are confined to a smaller effective volume (Fig. 4A). These combined observations demonstrate that the optical quality of our COF colloid nanoparticles enables high-quality spectroscopic measurements that were previously inaccessible. The fluence-sensitive exciton decay dynamics convey size-dependent photophysical properties and will lead to an improved understanding and leveraging of emergent electronic processes in these materials.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S13

Tables S1 to S4

References and Notes

Acknowledgments: We thank J. Bower and the cryo-electron microscopy facility at UC San Diego for assistance in low-dose HRTEM imaging. Funding: Supported by the Army Research Office for a Multidisciplinary University Research Initiatives (MURI) award under grant W911NF-15-1-0447. A.M.E. is supported by the NSF Graduate Research Fellowship under grant DGE-1324585, the Ryan Fellowship, and the Northwestern University International Institute for Nanotechnology. This study made use of the IMSERC and EPIC at Northwestern University, both of which have received support from the Soft and Hybrid Nanotechnology Experimental (SHyNE) Resource (NSF NNCI-1542205 and NSF ECCS1542205, respectively), the State of Illinois, and the International Institute for Nanotechnology (IIN). L.R.P. is supported by the National Institute of Biomedical Imaging and Bioengineering under award F32EB021859. N.C.G. was supported through a MURI through the Army Research Office under award W911NF-5-1-0568. Portions of this work were performed at the DuPont-Northwestern-Dow Collaborative Access Team (DND-CAT) located at Sector 5 of the Advanced Photon Source (APS). DND-CAT is supported by Northwestern University, E.I. DuPont de Nemours & Co., and the Dow Chemical Company. This research used resources of the Advanced Photon Source and Center for Nanoscale Materials, both U.S. Department of Energy (DOE) Office of Science User Facilities operated for the DOE Office of Science by Argonne National Laboratory under contract DE-AC02-06CH11357. N.C.F. and L.X.C. are partially supported by Basic Energy Science, CBG Division, DOE through Argonne National Laboratory under contract DE-AC02-06CH11357. Resources at the Advanced Photon Source were funded by NSF under award 0960140. Author contributions: A.M.E., N.C.F., R.P.B., E.V., and W.R.D. performed and interpreted the COF growth and structural characterization experiments; L.R.P. and N.C.G. performed and interpreted the TEM experiments; N.C.F., M.S.K., R.D.S., and L.X.C. performed and interpreted the time-resolved spectroscopy experiments; and all authors wrote and revised the manuscript. Competing interests: None declared. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or the supplementary materials.
View Abstract

Navigate This Article