Microscopic origins of poor crystallinity in the synthesis of covalent

43 mins ago - The hidden signs of transplant rejection. On TV medical dramas, a common suspenseful story line follows a person in need of an organ tra...
1 downloads 17 Views 15MB Size
Subscriber access provided by UNIVERSITY OF TOLEDO LIBRARIES

Article

Microscopic origins of poor crystallinity in the synthesis of covalent organic framework COF-5 Vu Nguyen, and Michael Gruenwald J. Am. Chem. Soc., Just Accepted Manuscript • DOI: 10.1021/jacs.7b12529 • Publication Date (Web): 02 Feb 2018 Downloaded from http://pubs.acs.org on February 3, 2018

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Journal of the American Chemical Society is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 17 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of the American Chemical Society

Microscopic origins of poor crystallinity in the synthesis of covalent organic framework COF-5 Vu Nguyen and Michael Grünwald∗ Department of Chemistry, University of Utah, Salt Lake City, Utah 84112 E-mail: [email protected] Abstract Covalent organic frameworks (COFs) are porous crystalline materials that are entirely composed of organic building blocks and can be assembled straightforwardly from solution. The main synthetic challenge associated with COFs, compared to other porous materials like zeolites or metal-organic frameworks, is their poor long-range order; typical sizes of crystal domains do not exceed a few tens of nanometers. Here, we develop a model of the molecular constituents of COF-5 and follow the early stages of its assembly dynamics from dilute solution. Our simulations indicate that under typical experimental conditions COF-5 formation happens not through nucleation, but far from equilibrium through spinodal decomposition. This rapid assembly mode leads to a plethora of defects that are difficult to anneal and that are likely responsible for the limited crystallinity observed in the synthesis of many COFs. We analyze the driving forces for COF-5 formation and find that stacking interactions between aromatic molecular constituents are too strong. When these interactions are weakened, assembly proceeds through single nucleation events followed by slow growth. The COF-5 crystallites obtained in this way are essentially defect-free. These results suggest experimental strategies for growing COFs with enhanced crystalline quality.

1

ACS Paragon Plus Environment

Journal of the American Chemical Society 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Introduction The self-assembly of ordered structures usually works best when interactions between components are weak. 1–3 With only a small bias towards structure formation, bonds between building blocks are frequently broken and defective patterns can anneal before further growth increases their lifetime. The range of experimental conditions over which high quality crystals can be achieved depends on the nature of the interactions between building blocks and the associated time scales for bond making and breaking. Furthermore, the growth of crystallites can occur via a spectrum of mechanisms that range from monomer addition to the incorporation of small clusters, to the attachment of large crystallites. 4,5 Accordingly, the nature of defects and the time scales needed to anneal them can vary greatly. To develop rational strategies aimed at improving the quality of crystals a clear microscopic picture of formation mechanisms and associated defect populations is needed. Two-dimensional covalent organic frameworks (2D-COFs) consist of porous planar polymers that stack on one another to form extended three-dimensional crystals. 6–11 To achieve large crystals of 2D-COFs, the covalent interactions responsible for 2D-polymerization should be weak and allow for frequent bond formation and breaking. In the first synthesis of a 2DCOF, Yaghi and coworkers addressed this problem through the use of condensation reactions with equilibrium constants that are tunable through the choice of solvent and additives. 12 Other, similarly reversible chemistries have since been successfully used to synthesize a wide range of 2D-COFs. 11 Nevertheless, these syntheses typically yield not single crystals but powder products with crystallite sizes on the order of tens of nanometers. 13 Powder COFs can display porosity that is inferior to theoretical predictions based on crystal structure and are unsuitable for applications that require large, near-macroscopic single crystals. 14 A growing body of experimental work suggests that non-covalent intermolecular interactions, in particular stacking interactions of planar molecules, can determine the structure and crystallinity of 2D-COFs. 15–22 While this insight has led to increasing crystallite size for 2D-COFs, the goal of reliably producing large crystals suitable for single X-ray diffraction studies has not 2

ACS Paragon Plus Environment

Page 2 of 17

Page 3 of 17 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of the American Chemical Society

been achieved.

Figure 1: (a) Excerpt from the COF-5 crystal structure. (Carbon atoms are shown in blue, boron in pink, and oxygen in red; hydrogen atoms not shown.) (b) Pair potential used to model covalent bond formation (red curve). Inset: Models of COF-5 building blocks, with red arrows indicating pairwise interactions. Special interaction sites (transparent spheres) on HHTP molecules interact with boron atoms to facilitate the formation of rigid 5-membered rings.

Methods COF-5, one of the two COFs first discovered, self-assembles under suitable conditions from the precursor molecules hexahydroxytriphenylene (HHTP) and phenylenebis(boronic acid) (PBBA) to form a honeycomb-like porous crystal, as illustrated in Figure 1a. 12 Dichtel and coworkers recently showed that COF-5 formation is essentially irreversible in experiments. 13 A recent kinetic Monte Carlo study revealed COF-5 nuclei with various morphologies. 23 In this work, we use molecular dynamics computer simulations to study the early stages of COF-5 formation. We have built a coarse-grained force-field for the molecular constituents of COF-5 (HHTP and PBBA) that treats solvent effects implicitely at synthetic conditions 3

ACS Paragon Plus Environment

Journal of the American Chemical Society 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

suggested by Dichtel and coworkers, which involve a temperature of 90◦ C, concentrations of [HHTP] = 8 mM and [PBBA] = 12 mM, and a solvent mixture of 4:1 dioxane:mesitylene with added methanol. The force-field treats molecules as partially rigid and includes the usual pairwise additive non-bonded interactions, including short-range repulsion and attractive van der Waals forces. To generate the damped molecular dynamics of our model, we use a Langevin thermostat as implemented in the simulation package LAMMPS. 24 (Force-field and simulation details are specified in the SI.) We designed our force-field to match the binding and stacking equilibria under experimental conditions. To reproduce the stacking behavior of molecules, we calculated the free energy of stacking of two HHTP molecules in explicit solvent, modeled with the TraPPE force-field. 25 We find that our implicit-solvent model reproduces this stacking free energy if all van der Waals interactions between molecules are scaled by a factor of s = 0.45. This scaling factor is consistent with recent calculations by Clancy and coworkers using the OPLS force-field. 26 We also calculated the stacking free energy for a solvent mixture of 1:1 dioxane:mesitylene, which is frequently used in the synthesis of COF-5. Monomers of HHTP and PBBA are not fully soluble in this solvent. Consistently, we observed stronger effective stacking interactions in this solvent, corresponding to a scale factor of s = 0.5 (see SI). Boronic ester formation, as required for 2D polymerization of COF-5, is modeled schematically. We model OH groups on HHTP and PBBA molecules as effective interaction sites, or "patches", that have mutual attractions described by a short-ranged pair potential that has its minimum of energy Ebind at zero separation, as illustrated in Figure 1b. To ensure the rigidity of the HHTP-PBBA dimer, we introduce an additional binding site on HHTP molecules that only interacts with Boron atoms on PBBA molecules. In addition, pair potentials responsible for binding are augmented with a barrier of height Ebar . The magnitude of Ebar controls the activation energy of bond formation, while Ebind sets the equilibrium constant. We adjust the parameters Ebind and Ebar to match experimental results by Dichtel and

4

ACS Paragon Plus Environment

Page 4 of 17

Page 5 of 17 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of the American Chemical Society

coworkers, who report equilibrium yields for boronic ester formation in a model reaction of PBBA and 4-tert-butylcatechol (TCAT). 15 From their data, we estimate an equilibrium boronic ester yield of 50% at 90◦ C. The value of Ebind = 6.2 kcal/mol used in our simulations produces the same yield in simulations of this model reaction. The activation energy Ea of the formation of a HHTP-PBBA dimer, which was measured as Ea = 7 kcal/mol, 15 can be tuned via the parameter Ebar . Based on rates of COF-5 formation in our simulation, we estimate that Ea ≈ Ebar (see SI). For values of Ebar > 6 kcal/mol, COF-5 formation becomes inaccessible to our straightforward molecular dynamics simulations. In our simulations, we thus use values of Ebar = 2 − 6 kcal/mol; these values are small enough to allow the formation of large COF-5 fragments in our simulations, but are sufficiently large to introduce a substantial separation of time scales between bond formation and stacking. (Stacking proceeds essentially barrier-free, as illustrated in Figure S5.) While our model is able to approximate experimentally measured activation energies of boronate ester formation, the reaction rates in our simulations are orders of magnitude larger compared to experiments, where boronate ester formation proceeds on the timescale of minutes. 15 Below, we present results that suggest that our conclusions are valid for structure formation on experimental time scales, too. Two additional simplifications of our model need to be addressed. First, in this schematic model of covalent bonding, boronic ester formation does not produce water molecules and bond breaking does not require the presence of water. In experiments, on the other hand, ester formation is initially irreversible, unless water is added. Therefore, our simulations effectively model experiments performed with sufficiently large amounts of added water. Experimentally, the addition of water has been shown to only modestly increase the crystalline quality of COF-5, 13 suggesting that other effects are responsible for the limited crystallinty of COFs. As a second caveat, the mechanism of bond formation is not accurately captured by our model. Differences in rate constants in dilute solution and the more crowded environment of a growing COF-5 cluster might therefore not be reproduced by our model. However,

5

ACS Paragon Plus Environment

Journal of the American Chemical Society 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Clancy and coworkers recently showed that activation energies for ester formation do not strongly depend on the molecular environment. 26

Results and Discussion Straightforward molecular dynamics initiated from an initially dispersed collection of 686 HHTP and 1029 PBBA molecules results in the rapid formation of a large number of COF5-like clusters. In the course of the simulation, these clusters grow via addition of monomers and attachment of other clusters, as illustrated in Figure 2a. In Figure 2b, we plot the potential energy and the sizes of the three largest connected cluster of molecules as a function of time. The immediate and continuous decrease of the potential energy indicates that under experimental conditions formation of large clusters does not await a nucleation event, but proceeds essentially barrier-less and is limited by diffusion and bond formation. These fast assembly dynamics generate clusters that are highly defective, as illustrated in Figure 2c. Typical defects include 5-membered rings (i.e., pentagonal pores formed by 5 HHTP and 5 PBBA molecules), screw dislocations, and interpenetrating rings. These defects persist over the course of our simulations; their annealing likely requires the correlated breaking of many bonds and we therefore expect them to be metastable also on experimental time scales. On the timescale accessible to our simulations, we can observe the growth of clusters up to a maximum diameter of about 25 nm, which is comparable to the typical domain size observed in experimental COF-5 powders. Due to the large concentration of these clusters, they will further grow by coalescence on longer time scales. As discussed below, large cluster are likely to attach without crystallographic alignment, even when the clusters themselves are defect free. To asses the ability of our force field to accurately describe defects, we have calculated the energy difference between molecules in isolated 5- and 6-membered rings with DFT at the B3LYP level and our force-field (see SI). We find that 5-membered rings are disfavored by

6

ACS Paragon Plus Environment

Page 6 of 17

Page 7 of 17 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of the American Chemical Society

Figure 2: (a) Time series of snapshots from a MD simulation of COF-5 formation at 90◦ C, 8 mM HTTP and 12 mM PBBA, and Ebar = 2 kcal/mol. Structure formation occurs rapidly and results in a large concentration of defective COF-5 fragments. A movie of this trajectory is available (file exp.mpg). (b) Potential energy (red curve) and number of molecules of the three largest clusters (blue, orange, and green curves) as a function of time. (c) Close-up view of five clusters observed after 500 ns of dynamics. Typical defects include interpenetrating rings, five-membered rings, and screw dislocations. 0.37 kcal/mol in DFT, and by 0.48 kcal/mol in our force field. Our force field thus somewhat overestimates the energy penalty associated with 5-membered rings in comparison to DFT, indicating that the experimental density of these defects might be even higher than in our

7

ACS Paragon Plus Environment

Journal of the American Chemical Society 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

simulations. This notion is supported by TEM images of single layers of COF-5 that show substantial concentrations of these topological defects. 27 The observed irreversible assembly dynamics does not depend strongly on the kinetics of bond formation. We have repeated our simulations with values of Ebar ranging from zero to 6 kcal/mol (see SI). Structure formation proceeds with different rates in these cases, but the number of observed nuclei and the concentration and types of defects are comparable to the results presented in Fig. 2. For values of Ebar > 6 kcal/mol, bond formation is rare on the timescale accessible to our simulations, and we cannot observe the formation of COF-5 from monomers under these conditions. However, we have probed the structures that form during the very early stages of COF-5 formation when realistically large kinetic barriers for bond formation are in effect. To this end, we first simulated a system of 686 HHTP and 1029 PBBA molecules with Ebar = 2 kcal/mol and weak stacking interactions at a scale factor s = 0.1. COF-5 formation is avoided in this simulation due to the weak stacking interactions. Instead, we observe the formation of an ensemble of small oligomers of different sizes (see SI). From this ensemble of oligomers we then initialized long molecular dynamics runs using the experimentally relevant strength of stacking interactions (s = 0.45) and an activation energy (Ebar = 20 kcal/mol) that is large enough to render bond formation and breaking highly unlikely on the simulation timescale, consistent with experimental bond formation rates. The results were qualitatively similar to the early stages of COF-5 formation observed with much smaller values of Ebar , as illustrated in Fig. 2. More precisely, we observed the aggregation of oligomers driven by stacking interactions (see SI). We have confirmed that no bonds are made or broken during these simulations; stacking interactions alone drive the formation of the observed structures. These simulations suggest that even relatively small oligomers are unstable against aggregation into defective clusters, and that there is no substantial nucleation barrier for COF-5 formation in experiments. The irreversible assembly and growth dynamics observed here strongly suggest that COF5 synthesis happens under conditions of large supersaturation in experiments. The degree

8

ACS Paragon Plus Environment

Page 8 of 17

Page 9 of 17 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of the American Chemical Society

of supersaturation can be reduced by decreasing solute concentration, by increasing temperature, or by modifying the solvent. The first two strategies are unlikely to be productive in the case of COF-5. A decrease in solute concentration does not change the timescale for bond breaking. Accordingly, simulations performed at ten times lower concentrations of monomers decrease the number of defects insignificantly. Temperature control is limited in experiments since the relevant solvents are already close to their boiling points in typical experiments and competing reactions between PBBA molecules become more frequent at higher temperatures. 12 The choice of solvent, on the other hand, can have a strong influence on COF formation. While COF-5 was originally synthesized in a 1:1 mixture of dioxane and mesitylene, HHTP and PBBA monomers are not fully soluble in this mixture. A 4:1 mixture of these solvents with methanol, however, provides a homogeneous solution of precursors and results in much faster COF-5 formation. 13 Recently, Dichtel and coworkers have furthermore shown that colloidal COF-5 particles with narrow size distribution can be produced by adding different amounts of acetonitrile. 28 In all these studies, however, the typical length scale of crystalline order does not exceed tens of nanometers. Our simulations suggest that much larger COF-5 crystallites could be grown in solvents that more effectively control stacking interactions between monomers in solution. We have screened different solvent conditions by simulating the self-assembly of COF-5 with varying strengths of covalent and stacking interactions, as parameterized by the parameter Ebind and the scaling factor s, respectively. We have identified conditions that lead to single nucleation events followed by slow and defect-free growth of COF-5 crystals. These conditions are shown as red data points in Figure 3a and are mostly characterized by substantially weaker stacking interactions (and slightly lower strength of covalent bonds) compared to experimental conditions. 13 In fact, at the covalent bond strength found in experiment, stacking interactions need to be reduced by half to yield COF-5 crystals of the highest quality. In Figure 3b, we plot the potential energy and size of the largest three clusters as a function of time at these improved conditions. As evident from the figure, a single COF crystal nucle-

9

ACS Paragon Plus Environment

Journal of the American Chemical Society 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Figure 3: (a) Kinetic phase diagram of COF-5 (T = 90◦ C, [HHTP] = 8 mM, [PBBA] = 12 mM), as a function of the binding strength Ebind (or, equivalently, the percent yield of esters in a model reaction of PBBA and TCAT) and the strength of stacking interactions s. Red points mark simulation conditions conducive of defect-free COF nucleation and growth. Experimental interaction strengths are indicated as a blue square. Green points indicate the saturation curve below which COF-5 is not stable (see SI). (b) Potential energy (red curve) and number of molecules of the three largest clusters (blue, orange, and green curves) as a function of time, for simulation conditions (4) in (a). The dashed line indicates the average potential energy in the supersaturated solution before nucleation. A movie of this trajectory is available (file opt.mpg). (c) Snapshots illustrate different crystallite morphologies obtained at interaction strengths as labeled in (a). ates from the metastable supersaturated solution. Even modestly stronger interactions (i.e., simulation conditions right above the red curve in Fig. 3a) lead to multiple nucleation events and higher concentrations of defects. Weaker interactions (i.e., simulation conditions below the red curve) do not lead to structure formation on the simulation time scale due to large nucleation barriers, but should result in high-quality crystals on longer, experimental time scales. We have verified that these results are insensitive to the activation energy of bond formation (see SI). Weaker stacking interactions lead to improved crystalline quality of COF-5 for two rea-

10

ACS Paragon Plus Environment

Page 10 of 17

Page 11 of 17 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of the American Chemical Society

sons. First, the free energy barrier of nucleation, ∆Gnuc , increases strongly when stacking interactions are weakened. Classical nucleation theory predicts that ∆Gnuc ∝ |∆µ|−2 , where ∆µ is the chemical potential difference between monomers in solution and in the solid, 29 which depends sensitively on the strength of stacking interactions between monomers (and also on the strengths of the covalent bonds). Reduced stacking interactions thus increase the nucleation free energy barrier, leading to exponentially smaller nucleation rates, knuc ∝ exp[−∆Gnuc /(kB T )]. The conditions marked in red color in Fig. 3a effect values of knuc that are just large enough to allow for a single nucleation event to occur on simulation timescales. Smaller interaction strengths result in nucleation rates too small for straightforward simulation. In fact, ∆µ is zero and the nucleation barrier infinitely large, along the saturation curve shown in green color in Fig. 3a. Second, weakened stacking interactions allow for more frequent dissociation of defective aggregates, thus increasing the ability of growing COF-5 fragments to anneal. High quality COF-5 crystals can be obtained for a range of suitable combinations of bond and stacking strength, as evident from Figure 3a. Across this range, the morphology of crystallite changes markedly, from compact (snapshot 2 in Figure 3c) to laterally extended (snapshot 4). These crystallite shapes likely reflect the equilibrium shapes at the respective conditions. Beyond this range, formation of well-defined COF-5 crystallites gives way to defective compact aggregates dominated by stacking interactions (snapshot 1) and large single-layer 2D polymers (snapshot 5), respectively. Control over the relative strength of binding and stacking thus potentially allows one to tune the morphology of COF crystals. Because of computational cost, system sizes in our simulations are limited to thousands of molecules, approximately corresponding to the size of typical COF-5 crystallites observed in experiments. On longer time scales, crystallites will continue to grow through coalescence and Ostwald ripening. 4,30 Imperfect coalescence of large crystallites could pose further bottlenecks in the production of long-range ordered COFs. To gauge the ability of well-ordered COF crystallites to undergo oriented attachment, we have performed several simulations

11

ACS Paragon Plus Environment

Journal of the American Chemical Society 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Figure 4: Time series of snapshots from a simulation of two parallel COF-5 crystallites undergoing imperfect attachment. under experimental conditions, initiated by placing two parallel, disc-shaped COF crystallites of 15 nm diameter at a distance of 5 nm. The crystallites typically aggregate within a few nanoseconds. About half our simulations result in oriented attachment, the other half produce defective aggregates, as illustrated in Figure 4. We estimate that the free energy barrier for annealing such stacking defects is many tens of kcal/mol, thus rendering this type of defect formation essentially irreversible. Attachment of well-ordered crystallites with other relative orientations result in long-lived defects with similar probability (see SI). Defective COF-5 fragments likely have an even lower propensity for oriented attachment. Achieving smaller nucleation rates, and thus a lower density of COF-5 nuclei, is therefore a prerequisite for improving the size of crystalline domains in the synthesis of COFs.

Conclusions While we have focused on the assembly of COF-5 in this paper, it is likely that strong stacking interactions are at least partially responsible for small crystallite sizes of other 2D-COFs, too, since most 2D-COFs are made from planar monomers that guarantee ordered stacking of 2D polymers. Bein and coworkers have recently demonstrated a new strategy to increase crystallite size in COFs by using catechols with chiral, non-planar shapes that restrict the configurations in which two monomers can stack. 16 The resulting COF crystallites have 12

ACS Paragon Plus Environment

Page 12 of 17

Page 13 of 17 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of the American Chemical Society

typical sizes of a few hundred nanometers, a significant improvement over typical COF-5 products. The success of this strategy can be understood in the context of our simulations, as the stacking free energy of monomers is likely reduced for non-planar molecules. While our work indicates that the most direct way to increase crystallite size is to identify solvents that diminish effective stacking interactions, our simulations do not directly suggest particular solvents that would achieve this. To this end, spectroscopic methods could be used to probe the stacking propensity of monomers in different solvents. A computational screening of different solvents could also furnish promising solvent candidates for experimental investigation. In a recent study, Clancy and coworkers tested several solvents for their ability to screen stacking interactions of COF-5 fragments using the OPLS force field. 26 Consistent with our results, these authors found that dioxane and mesitylene do not weaken stacking interactions particularly effectively. Interestingly, the authors identified acetone and dimethyl acetamide as the two solvents that provide the weakest effective stacking interactions. Another possible strategy to grow large COF crystals might be to forego the nucleation stage entirely and seed a solution at low supersaturation with a small concentration of conventionally obtained crystallites. Defective aggregation of large numbers of small crystallites might be avoided in this fashion.

Supporting Information Available The following files are available free of charge. Methods and simulation details, additional figures, and discussion.

Author Information Corresponding author [email protected] 13

ACS Paragon Plus Environment

Journal of the American Chemical Society 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Acknowledgement The authors thank Will Dichtel, Jon Rainier, and Xuchen Zhao for useful discussions. The support and resources from the Center for High Performance Computing at the University of Utah are gratefully acknowledged.

References (1) Hagan, M. F.; Chandler, D. Biophys. J. 2006, 91, 42–54. (2) Grant, J.; Jack, R. L.; Whitelam, S. J. Chem. Phys. 2011, 135, 214505. (3) Whitelam, S. Phys. Rev. Lett. 2010, 105, 088102. (4) De Yoreo, J. J.; Gilbert, P. U. P. a.; Sommerdijk, N. a. J. M.; Penn, R. L.; Whitelam, S.; Joester, D.; Zhang, H.; Rimer, J. D.; Navrotsky, A.; Banfield, J. F.; Wallace, a. F.; Michel, F. M.; Meldrum, F. C.; Colfen, H.; Dove, P. M. Science 2015, 349, aaa6760– aaa6760. (5) De Yoreo, J.; Whitelam, S. MRS Bull. 2016, 41, 357–360. (6) Zwaneveld, N. A. A.; Pawlak, R.; Abel, M.; Catalin, D.; Gigmes, D.; Bertin, D.; Porte, L. J. Am. Chem. Soc. 2008, 130, 6678–6679. (7) Han, S. S.; Furukawa, H.; Yaghi, O. M.; Goddard, W. A. J. Am. Chem. Soc. 2008, 130, 11580–11581. (8) DeBlase, C. R.; Dichtel, W. R. Macromolecules 2016, 49, 5297–5305. (9) Wang, Y.; Xie, S.; Liu, J.; Park, J.; Huang, C. Z.; Xia, Y. Nano Lett. 2013, 13, 2276–81. (10) Colson, J. W.; Dichtel, W. R. Nat. Chem. 2013, 5, 453–465. (11) Waller, P. J.; Gándara, F.; Yaghi, O. M. Acc. Chem. Res. 2015, 48, 3053–3063. 14

ACS Paragon Plus Environment

Page 14 of 17

Page 15 of 17 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of the American Chemical Society

(12) Cote, A. P.; Benin, A. I.; Ockwig, N. W.; O’Keeffe, M.; Matzger, A. J.; Yaghi, O. M. Science 2005, 310, 1166–1170. (13) Smith, B. J.; Dichtel, W. R. J. Am. Chem. Soc. 2014, 136, 8783–8789. (14) Bisbey, R. P.; Dichtel, W. R. ACS Cent. Sci. 2017, 3, 533–543. (15) Smith, B. J.; Hwang, N.; Chavez, A. D.; Novotney, J. L.; Dichtel, W. R. Chem. Commun. 2015, 51, 7532–7535. (16) Ascherl, L.; Sick, T.; Margraf, J. T.; Lapidus, S. H.; Calik, M.; Hettstedt, C.; Karaghiosoff, K.; Döblinger, M.; Clark, T.; Chapman, K. W.; Auras, F.; Bein, T. Nat. Chem. 2016, 8, 310–316. (17) Thompson, C. M.; Occhialini, G.; McCandless, G. T.; Alahakoon, S. B.; Cameron, V.; Nielsen, S. O.; Smaldone, R. A. J. Am. Chem. Soc. 2017, 139, 10506–10513. (18) Kandambeth, S.; Shinde, D. B.; Panda, M. K.; Lukose, B.; Heine, T.; Banerjee, R. Angew. Chemie Int. Ed. 2013, 52, 13052–13056. (19) Chen, X.; Addicoat, M.; Irle, S.; Nagai, A.; Jiang, D. J. Am. Chem. Soc. 2013, 135, 546–549. (20) Chen, X.; Addicoat, M.; Jin, E.; Zhai, L.; Xu, H.; Huang, N.; Guo, Z.; Liu, L.; Irle, S.; Jiang, D. J. Am. Chem. Soc. 2015, 137, 3241–3247. (21) Xu, H.; Gao, J.; Jiang, D. Nat. Chem. 2015, 7, 905–912. (22) Koo, B. T.; Dichtel, W. R.; Clancy, P. J. Mater. Chem. 2012, 22, 17460. (23) Li, H.; Chavez, A. D.; Li, H.; Li, H.; Dichtel, W. R.; Bredas, J.-L. J. Am. Chem. Soc. 2017, 139, 16310–16318. (24) Plimpton, S. J. Comput. Phys. 1995, 117, 1–19.

15

ACS Paragon Plus Environment

Journal of the American Chemical Society 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

(25) Keasler, S. J.; Charan, S. M.; Wick, C. D.; Economou, I. G.; Siepmann, J. I. J. Phys. Chem. B 2012, 116, 11234–11246. (26) Koo, B. T.; Heden, R. F.; Clancy, P. Phys. Chem. Chem. Phys. 2017, 19, 9745–9754. (27) Zwaneveld, N. A. A.; Pawlak, R.; Abel, M.; Catalin, D.; Gigmes, D.; Bertin, D.; Porte, L. J. Am. Chem. Soc. 2008, 130, 6678–6679. (28) Smith, B. J.; Parent, L. R.; Overholts, A. C.; Beaucage, P. A.; Bisbey, R. P.; Chavez, A. D.; Hwang, N.; Park, C.; Evans, A. M.; Gianneschi, N. C.; Dichtel, W. R. ACS Cent. Sci. 2017, 3, 58–65. (29) Jungblut, S.; Dellago, C. Eur. Phys. J. E 2016, 39, 77. (30) Ratke, L.; Voorhees, P. W. Growth and Coarsening: Ostwald Ripening in Materials Processing; Springer-Verlag, 2002.

16

ACS Paragon Plus Environment

Page 16 of 17

Page 17 of 17 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of the American Chemical Society

Graphical TOC Entry

17

ACS Paragon Plus Environment