Mechanisms of nucleation

Very little is known about the mechanisms of COF and MOF nucleation. One study on COF-5^{Reference Smith and Dichtel18} suggests that monomers or small oligomers are involved in growth (Figure 3a), which could imply that the nucleation of COF-5 (and possibly other COFs) occurs by processes that are more closely aligned with classical nucleation theory (CNT) (see the Introductory article in this issue). MOFs are prepared using molecularly dispersed metals and ligands and thus have the potential to nucleate by a similar mechanism. Few reports in the literature address COF and MOF nucleation; to our knowledge, there are no studies identifying the presence of precursors (e.g., clusters or primary particles) in synthesis solutions that would allude to nonclassical pathways of COF or MOF crystallization.^{Reference De Yoreo, Gilbert, Sommerdijk, Penn, Whitelam, Joester, Zhang, Rimer, Navrotsky, Banfield, Wallace, Michel, Meldrum, Colfen and Dove19}

Figure 3.

Putative pathways of nucleation for open framework crystals. (a) Dichtel and co-workers propose that COF-5 (covalent–organic framework) nucleation occurs through the assembly of monomers and oligomers, which also serve as growth units for 2D crystal growth. Adapted with permission from Reference Reference Smith and Dichtel18. © 2006 American Chemical Society. (b) One pathway of zeolite nucleation involves the aggregation of primary particles (step 1) followed by their structural rearrangement (step 2) into the crystalline product. During this process, the nuclei maintain a size and shape that is similar to the amorphous aggregate. The crystals may contain voids or defects (e.g., misaligned domains). The primary particles are composed of (alumino)silicate polymers with entrained solvent (callout illustration). (c) A second pathway of zeolite nucleation involves the formation of bulk (sol-gel) amorphous aggregates (step 3). These particles serve as sites for heterogeneous nucleation, either on the exterior surface of gels or within their interior. Gels contain entrained solvent (callout illustration), and the (alumino)silicate within the gel serves as a nutrient for zeolite growth. During crystallization, the amorphous gel is consumed (step 4), often leading to crystal sizes that are commensurate with those of the original gel particles.

On the other hand, there is more information on zeolite nucleation; however, molecular-level details are generally lacking, and few structures have been studied, leaving many fundamental questions unanswered. One defining feature of zeolite synthesis is the formation of precursors, which include primary particles (1–6 nm), bulk aggregates (0.1–1 μm), and gels. Zeolite nucleation does not appear to involve the association of monomers/oligomers into clusters of critical size, as per CNT.^{Reference Kashchiev20} The ubiquitous presence of precursors in zeolite growth solutions^{Reference Cundy and Cox21} suggests a two-step mechanism of nucleation,^{Reference Galkin and Vekilov22,Reference Vekilov23} wherein the assembly of precursors provides a substrate for heterogeneous nucleation or potentially generates localized regions of high-solute (silica and alumina) concentration, which lowers the energetic barrier for nucleation.

The amorphous-to-crystalline transition in zeolite precursors is not well understood. This process can be subdivided into two nucleation pathways (Figure 3b–c) that are distinguished on the basis of the initial (alumino)silicate connectivity. Primary particles are composed of branched polymeric networks with many uncoordinated bonds (e.g., SiOH groups) and a significant fraction of solvent.^{Reference Rimer, Vlachos and Lobo24–Reference Hould and Lobo26} As shown in Figure 3b, nucleation often involves the aggregation of primary particles, followed by structural rearrangement into a nucleus through unknown molecular events and dynamic sequences. The majority of zeolite growth solutions are composed of amorphous gel particles (50 nm–1 μm), see Figure 3c. Gel particles often contain heterogeneous compositions with spatially segregated Si- and Al-rich domains,^{Reference Maldonado, Oleksiak, Chinta and Rimer27,Reference Ren, Subotic, Bronic, Tang, Sikiric, Misic, Svetlicic, Bosnar and Jelic28} suggesting that nucleation within gels involves significant bond breakage to form zeolite crystals. There are two general locations where nucleation occurs in gel particles: within the gel interior and on its exterior surface.^{Reference Mintova, Olson and Bein29,Reference Mintova, Olson, Valtchev and Bein30} The complexity of nucleation in gels renders these systems more difficult to characterize, therefore, our discussion will focus on studies of sols (i.e., pathway in Figure 3b) that constitute the basis of our understanding of zeolite nucleation.

Experimental and computational studies of zeolite nucleation

One of the most widely studied zeolites is silicalite-1 (siliceous MFI framework type), which is prepared from clear growth solutions that enable facile in situ characterization by scattering and microscopy techniques. Amorphous precursors of silicalite-1 are the primary particles (Figure 4a–b) that range in size from 1–6 nm, depending on the pH of the growth solution,^{Reference Davis, Drews, Ramanan, He, Dong, Schnablegger, Katsoulakis, Kokkoli, McCormick, Penn and Tsapatsis31–Reference Kumar, Davis, Ramanan, Penn and Tsapatsis33} and have a core–shell arrangement with a silica-rich core and an OSDA-rich shell^{Reference Rimer, Vlachos and Lobo24,Reference Fedeyko, Rimer, Lobo and Vlachos32,Reference Rimer, Lobo and Vlachos34} with no evidence of the silicalite-1 structure. Recently, Auerbach and Monson used a reaction ensemble Monte Carlo method to model the self-assembly of silica in the presence of OSDA (Figure 4c), which provided insights on the ring-size distribution and OSDA location.^{Reference Chien, Auerbach and Monson35} Their results are in agreement with the conclusions drawn from small-angle x-ray scattering and small-angle neutron scattering experiments by Rimer, Lobo, and Vlachos in that they indicate the absence of the silicalite-1 crystal structure within the nanoparticles.^{Reference Kragten, Fedeyko, Sawant, Rimer, Vlachos, Lobo and Tsapatsis36,Reference Rimer, Trofymluk, Navrotsky, Lobo and Vlachos37} It appears that the silica nanoparticles are a metastable form of the silica-OSDA assembly with slow dynamics of evolution toward the silicalite-1 structure.

Figure 4.

Nucleation of zeolite silicalite-1 (MFI framework type). Cryogenic transmission electron microscope (cryo-TEM) images^{Reference Davis, Drews, Ramanan, He, Dong, Schnablegger, Katsoulakis, Kokkoli, McCormick, Penn and Tsapatsis31} of (a) fresh primary particles and (b) after 180 days of aging at room temperature reveal structural evolution. (c) Snapshots of particles with an ordered arrangement of silica and organic structure-directing agent (OSDA) from reaction ensemble Monte Carlo simulations. Color code: Si (yellow), oxide (blue), bridging oxygen (red), hydroxyl (gray), and OSDA (purple). Reproduced with permission from Reference Reference Chien, Auerbach and Monson35. © 2015 American Chemical Society. (d–e) Cryo-TEM images^{Reference Kumar, Wang, Penn and Tsapatsis38} of samples extracted at periodic times reveal the amorphous-to-crystal transformation of ca. 20-nm aggregates of precursors: (d) 200 days and (e) 220 days at room temperature. Insets are selected-area electron diffraction patterns.

Cryogenic transmission electron microscopy (cryo-TEM) was used to track the evolution of precursors during crystallization at room temperature, and revealed a sequence of events (depicted in steps 1 and 2 of Figure 3b) that started with the aggregation of precursors (Figure 4d).^{Reference Kumar, Wang, Penn and Tsapatsis38} The experiments were performed at room temperature, where the disparity of time scales of silica assembly (seconds or less) with the very slow emergence of aggregates (months) were highlighted. Selected-area electron diffraction and high-resolution TEM imaging revealed these aggregates to be amorphous; however, increased synthesis time resulted in an amorphous-to-crystalline transition that was marked by the appearance of lattice fringes in cryo-TEM images (Figure 4e).^{Reference Davis, Drews, Ramanan, He, Dong, Schnablegger, Katsoulakis, Kokkoli, McCormick, Penn and Tsapatsis31} After a relatively short time (compared to the time it took for aggregation to be detected), the aggregates became fully crystalline with a size and shape comparable to those of the original aggregate. Void spaces were detected in the nuclei, which correspond to interstitial spaces within the aggregates.

It was also observed that nuclei are predominantly single crystals, where domains (i.e., regions corresponding to original precursors in the aggregate) share the same crystallographic orientation. Misaligned domains are occasionally observed. Such domains are detected in bulk crystals of zeolite ZSM-5 (MFI framework type) as intergrowths,^{Reference Karwacki, Kox, de Winter, Drury, Meeldijk, Stavitski, Schmidt, Mertens, Cubillas, John, Chan, Kahn, Bare, Anderson, Kornatowski and Weckhuysen39} and they have also been documented for other nanocrystalline materials that grow through attachment of crystalline^{Reference Penn and Banfield40} or amorphous particles. In some cases, lattice strain at these defect sites may induce temporal reorientation, similar to Ostwald ripening of misaligned particles observed during iron oxyhydroxide crystallization by oriented attachment.^{Reference Li, Nielsen, Lee, Frandsen, Banfield and De Yoreo41}

The two-step mechanism of nucleation involves (1) the assembly of precursors and (2) their transformation into nuclei. As previously described, the first step appears to be well understood, but the second step is invariably complex and likely consists of numerous kinetic barriers, many of which may be energetically indistinguishable. Specifically, these processes involve bond breakage and formation, exchange of (alumino)silicate species between the precursor and growth solution, and the assembly of silicate monomers and fragments, including composite building units (CBUs) that correspond to repeating structures (e.g., cages or chains) of the crystalline framework.

Many computational studies in the literature examine the polymerization of (alumino)silicate clusters and the formation of rings and CBUs.^{Reference Chien, Auerbach and Monson35,Reference Malani, Auerbach and Monson42–Reference Yang, Mora-Fonz and Catlow46} A common hypothesis in the literature is that zeolite crystals grow by the direct attachment of CBUs;^{Reference Park, Lee, Zheng, Xiao, Nicholas, Lewis and Hong47} however, there is no direct evidence to support this mechanism. Much of the experimental data used to support these theories are derived from spectroscopic measurements (i.e., Fourier transform infrared, nuclear magnetic resonance, Raman, mass spectrometry) showing progressive changes in the connectivity of (alumino)silicate species during the induction period.^{Reference Lesthaeghe, Vansteenkiste, Verstraelen, Ghysels, Kirschhock, Martens, Van Speybroeck and Waroquier48–Reference Follens, Aerts, Haouas, Caremans, Loppinet, Goderis, Vermant, Taulelle, Martens and Kirschhock50} The temporal appearance or disappearance of rings, CBUs, and other oligomeric species provides evidence that structural changes occur. Linking these observations to the molecular processes of nucleation is a more difficult task. From the experimental standpoint, we should seek conditions where nucleation is not rare and fast (i.e., model systems where nucleation happens simultaneously throughout the crystallizing sols), such that changes in solution and nanoparticle structures (determined by the aforementioned techniques) can be correlated with the onset of nucleation. From the modeling standpoint, bridging approximate methods capable of describing silica polymerization and network formation^{Reference Jin, Auerbach and Monson51} with more accurate descriptions of silica speciation in the presence of OSDA^{Reference Caratzoulas, Vlachos and Tsapatsis52} could provide quantitative predictions of nucleation kinetics.

Polymorphism and intercrystalline conversions

Navrotsky and co-workers performed calorimetry studies and demonstrated that the enthalpies of formation ΔH _f for microporous zeolites^{Reference Navrotsky, Trofymluk and Levchenko53} and MOFs^{Reference Wu and Navrotsky54} with pore diameters less than 2 nm exhibit an approximately linear relationship with framework molar volume V; however, there is very little difference in ΔH _f among the different framework types (i.e., values differ by less than 10 kJ/mol). This similarity in enthalpy can result in the formation of polymorphs, which is commonly observed in MOF and zeolite synthesis. For example, there are multiple structures that can form during the preparation of zeolitic imidazolate frameworks (ZIFs) from solutions of zinc and imidazole (IM). As shown in Figure 5a, four different crystal structures—ZIF-2, ZIF-3, ZIF-6, and ZIF-10—can be prepared with identical composition, Zn(IM)₂, as a result of subtle changes in the synthesis conditions, such as temperature and solvent.^{Reference Park, Ni, Cote, Choi, Huang, Uribe-Romo, Chae, O’Keeffe and Yaghi55} This alludes to the important role of kinetics in nucleation.

Figure 5.

Examples of polymorphism for open framework materials. (a) Polymorphs of zeolitic imidazolate frameworks (ZIFs) with Zn(IM)₂ composition (IM is imidazole). Black dots indicate Zn, blue lines are IM, and yellow spheres highlight open voids (cages) within the pores. The kinetic pathways leading to the nucleation and growth of these four structures is altered by subtle changes in synthesis parameters. The three-letter abbreviations in parentheses refer to the framework structure. Adapted with permission from Reference Reference Park, Ni, Cote, Choi, Huang, Uribe-Romo, Chae, O’Keeffe and Yaghi55. © 2006 National Academy of Sciences. (b) Example of intercrystalline conversions leading to the formation of MOF-5 (metal–organic framework), which is composed of 1,4-benzenedicarboxylate (BDC) linkers. Dincǎ and co-workers showed that electrochemically induced growth led to the formation of two intermediates prior to the transformation to MOF-5 product. Adapted with permission from Reference Reference Li and Dincǎ57. © 2015 American Chemical Society. (c) The thermodynamic driving force for intercrystalline transformations in zeolite synthesis is governed by the progressive transition from initially metastable structures to those that are more thermodynamically stable (i.e., dense) structures. Rimer and co-workers reported the FAU-to-GIS-to-ANA transition that occurs with increased synthesis temperature or time.^{Reference Maldonado, Oleksiak, Chinta and Rimer27} Note: FAU, faujasite; GIS, gismondine; ANA, analcime; V, molar volume.

Notably, there are multiple kinetic barriers that must be overcome during the amorphous-to-crystalline transition in the two-step nucleation mechanism leading to the formation of secondary building units. A multitude of parameters can be manipulated in zeolite and MOF synthesis to alter the nucleation pathway. Establishing deterministic models or heuristic guidelines to control nucleation pathways a priori is one of the more challenging objectives in this field of research.

A common phenomenon during the crystallization of open framework materials illustrates the Ostwald–Lussac rule of stages (or Ostwald step rule),^{Reference Navrotsky56} which states that the first crystal to form is a metastable structure that undergoes a series of intercrystalline transformations leading to structures that are more thermodynamically stable. In the context of Navrotsky’s calorimetry experiments, the trajectory of these transformations is toward denser framework structures (i.e., lower V). For example, the synthesis of MOF-5 under cathodic bias (Figure 5b) occurs in three stages involving the initial formation of two layered structures prior to the formation of the final MOF-5 crystal.^{Reference Li and Dincǎ57} Intercrystalline transformations in zeolite synthesis are more commonly observed in OSDA-free growth solutions where alkali metals are used as inorganic structure-directing agents. An example of the Ostwald step rule for Na⁺ zeolites is shown in Figure 5c involving the transition from faujasite (FAU framework type), a common metastable zeolite, to gismondine (GIS), and then analcime (ANA) with increasing synthesis temperature.^{Reference Maldonado, Oleksiak, Chinta and Rimer27} During these transformations, the initial formation of one zeolite is followed by nucleation of a second zeolite, which progressively grows in favor of a decreasing population of the initial (more metastable) structure.

These examples suggest that nucleation of certain zeolites is favored by the formation/presence of more metastable zeolites. The driving force for intercrystalline transformation is not well understood. Some transformations occur between zeolite structures with similar building units;^{Reference Oleksiak and Rimer58} however, the three zeolite structures in Figure 5c do not share a common CBU. One possible explanation is an epitaxial relationship among zeolite structures wherein the crystal interface of the newly formed zeolite preferentially nucleates on a crystallographic face of the preexisting zeolite surface. Alternatively, the temporal change in growth solution composition during the crystallization of the metastable zeolite creates an environment that could potentially promote nucleation of the new phase. In many respects, intercrystalline transformation is analogous to seeded growth—a topic that is garnering increased attention in zeolite synthesis.^{Reference Xie, Zhang, Yang, Liu, Ren, Zhang, Meng, Yilmaz, Muller and Xiao59,Reference Itabashi, Kamimura, Iyoki, Shimojima and Okubo60} Developing a fundamental understanding of these processes could lead to improved methods of controlling nucleation pathways.^{Reference Eliášová, Opanasenko, Wheatley, Shamzhy, Mazur, Nachtigall, Roth, Morris and Čejka61}

Conclusions and outlook

We have presented an overview of the potential nucleation pathways of open framework crystals, focusing specifically on three classes of materials: zeolites, MOFs, and COFs. The topic of nucleation has received little attention in the MOF and COF literature; this presents an opportunity for future advancements in these areas of research. It is reasonable to believe that crystallization from growth solutions composed of small organic molecules and metal ions involves classical monomer-by-monomer addition; however, there exists a possibility that precursors (e.g., clusters or oligomers) form during the induction period and facilitate nucleation. The number of crystalline materials that are either known or postulated to grow by nonclassical routes involving crystallization by particle attachment (CPA)^{Reference De Yoreo, Gilbert, Sommerdijk, Penn, Whitelam, Joester, Zhang, Rimer, Navrotsky, Banfield, Wallace, Michel, Meldrum, Colfen and Dove19} is continually expanding. It is entirely feasible that MOF and COF crystals could be added to this list of materials.

Knowledge of zeolite growth by CPA dates back to the 1990s, when primary particles were first identified in silicalite-1 syntheses.^{Reference Burkett and Davis62,Reference Schoeman, Sterte and Otterstedt63} Difficulties imposed by zeolite synthesis conditions, however, render in situ characterization of nucleation and crystal growth challenging. To this end, observing zeolite nucleation at the length and time scales necessary to reconcile mechanistic uncertainties with molecular-level resolution remains elusive, but characterization and modeling methodologies have reached a level of maturity that, if applied to an appropriate system, could soon provide the first molecular-level quantitative description of zeolite nucleation kinetics. The vast number of hypothetical open framework structures that have yet to be synthetically realized point to a design space for materials discovery that is virtually limitless, provided the technical barriers for characterization can be overcome and knowledge gaps are sufficiently addressed. Indeed, the voids in our understanding of nucleation of open framework materials still loom large.