Star PolyMOCs with Diverse Structures, Dynamics, and Functions by Three-Component Assembly

Abstract

We report star polymer metal-organic cage (polyMOC) materials whose structures, mechanical properties, functionalities, and dynamics can all be precisely tailored through a simple three-component assembly strategy. The star polyMOC network is composed of tetra-arm star polymers functionalized with ligands on the chain ends, small molecule ligands, and palladium ions; polyMOCs are formed via metal-ligand coordination and thermal annealing. The ratio of small molecule ligands to polymer-bound ligands determines the connectivity of the MOC junctions and the network structure. The use of large M₁₂L₂₄ MOCs enables great flexibility in tuning this ratio, which provides access to a rich spectrum of material properties including tunable moduli and relaxation dynamics.

Graphical Abstract

A three-component assembly strategy is reported that enables the creation of a versatile class of polymer metal-organic cage (polyMOC) networks with tailored microstructures, mechanical properties, functionalities, and network dynamics.

Polymer networks are versatile materials with a wide range of structures and properties suitable for industrial and academic applications.^[1–3] In a typical network, macromolecules of choice are connected to branched junctions of a particular type; the nature of these components determines the material’s properties such as stiffness, toughness, responsiveness, etc.^[2] Several strategies have been developed to tune polymer network structure in order to realize desirable properties. For example, interpenetrating networks,^[4–5] nanocomposites,^[6] and reversible and/or dynamic covalent bonds^[7] are employed to yield materials with self-healing, stimuli-responsive, and other valuable behaviors.^[7–8] In all of these cases, control over network structure and dynamics is critical. In this communication, we describe a versatile and simple strategy for controlling structure, function, and dynamics in a relatively new class of polymer networks that is based on the use of large metal-organic cages/polyhedra (MOCs).

MOCs are discrete 3D structures assembled from x metal ions and y organic ligands via coordination bonds.^[9–22] By rational design of the ligands and proper choice of the metal ions, MOCs of different M_xL_y stoichiometries, sizes, and geometries can be synthesized.^[23–29] Inspired by this structural versatility and potential applications such as mechanical enhancement, catalysis, encapsulation, sensing, etc., much effort has been recently devoted to installing MOCs into polymer networks to provide ‘polyMOC’ hybrid materials with tunable viscoelasticity and functionality.^[30–37]

One scheme for polyMOC synthesis utilizes linear polymers linked by MOCs as crosslink junctions. For example, by using Pd-pyridine-based MOCs with distinct sizes, M₁₂L₂₄ and M₂L₄, we demonstrated that the mechanical properties of a poly(ethylene glycol) (PEG)-based polyMOC gel could be readily controlled, thereby offering a new strategy for regulating mechanics in constitutionally isomeric materials.^[31] Nitschke and co-workers reported hydrogel polyMOCs crosslinked by tetrahedral MOCs that could selectively encapsulate and release small molecules.^[32] Several other recent examples of MOC-containing polymeric materials highlight the potential of these systems^{[30, 38–42]}; however, there is still a great need to understand how polyMOC microstructure translates to bulk material properties such as modulus and relaxation dynamics. To accomplish this goal, robust, modular polyMOC synthesis strategies that enable access to a wide range of structures and properties are needed.

Herein, we describe a three-component assembly approach for the modular synthesis of polyMOCs from ligand functionalized tetra-arm star polymers, small molecule ligands (SML), and Pd²⁺ ions (Figure 1). When these components are mixed in a desired ratio and annealed, a network of star polymers connected to MOCs is formed wherein each MOC possesses a mixture of polymer-bound ligands (PLs) and SMLs. By combining star polymers, which act as covalent network junctions, and large Pd₁₂L₂₄ MOCs, which serve as dynamic network junctions as well as reservoirs for SMLs, a wide spectrum of polyMOC structures, properties, and dynamics is accessible. This approach greatly expands the versatility of polyMOC chemistry, and provides a synthetic strategy that should be amenable to new classes of polyMOCs in the future.

Figure 1.

Representative scheme showing the formation of polyMOC gels composed of polymer-bound ligands (PL), small molecule ligands (SMLs) and Pd²⁺ through the subcomponent metallosupramolecular assembly of M₁₂L₂₄ MOCs. The ratio of SML/PL is denoted as n.

The star polymer used in this study (Figure 1, M_n = 18k, Đ = 1.09) was prepared by atom transfer radical polymerization (ATRP) from pentaerythritol tetrakis(2-bromoisobutyrate) initiator followed by end-group substitution with a para-bispyridyl phenol (see Supporting Information, SI, for details). ¹H NMR spectroscopy suggested greater than 95% chain end functionalization (Figure S1). The SML was an analogous para-bispyridyl methyl ether derivative (Figure 1, R = Me).

We synthesized a series of polyMOC gels where the concentration of polymer was held constant while the amounts of SML and Pd²⁺ were varied (the ratio of Pd²⁺ to pyridine groups was always 1:4). We name these polyMOC gels as Gel_n, where n is the molar ratio of SMLs to PLs (PL = equivalents of star polymer × 4) used in the synthesis. If we assume that all of the ligands are incorporated into M₁₂L₂₄ MOCs, then the average number of SMLs in each MOC is a = 24n/(n+1); the number of PLs per MOC is f = 24 - a. For example, Gel₅ is a polyMOC made by mixing 1 equivalent of the star polymer (i.e., 4 PLs) and 20 equivalents of SML (20 / 4 = 5) with a stoichiometric amount of Pd²⁺. Thus, each MOC in this network has an average of 20 SMLs (a = 20) and 4 PLs (f = 4). PolyMOC gels with n values ranging from 0 (i. e. no SML) to 13 were prepared.

Figure 2 shows the aromatic region of the magic angle spinning (MAS) ¹H NMR spectra for selected polyMOCs alongside the solution ¹H NMR spectra for the SML and the M₁₂L₂₄ MOC formed from this SML. MAS NMR spectra for other polyMOCs are provided in Figure S2. From these data, it is clear that as n increases the MOCs become less restricted; the aromatic resonances in the MAS NMR spectra become more similar to those of the free MOC in solution.

Figure 2.

Solution ¹H NMR (ω_0H/2π = 400 MHz) of SML and its corresponding M₁₂L₂₄ MOC along with the MAS NMR spectra (ω_0H/2π = 500 MHz, ω_r/2π = 12.5 kHz) of selected polyMOC Gels.

To confirm that the SMLs were integrated within the polyMOC network, we thoroughly extracted the gels with fresh DMSO-d₆ and analyzed the soluble fraction by ¹H NMR. As shown in Table S1, as the amount of SML decreased, the percentage of soluble MOCs (MOCs not connected to the network) also decreased: ~15% for Gel_11, 13% for Gel₉, 10% for Gel₅, and 4% for Gel₃. When n ≤ 2, free MOCs were not detectable. Thus, the majority of the SMLs are incorporated into these polyMOCs. When n = 13, the average MOC contains only f = ~1.7 PLs. Thus, though gels were obtained, more than 25% percent of SML as well as some PL were extracted.

To investigate how the average number of PLs connected to each MOC (f) affects the MOC mobility, we measured the transverse relaxation time, T_2, of a characteristic MOC peak (9.6 ppm) by MAS ¹H NMR. We assume that MOCs with more SMLs (i.e., higher n and smaller f) will relax more rapidly (larger T₂). The T₂ values for the as made versus extracted Gel₉ (f = 2.4) were 9.7 ± 0.6 ms and 8.5 ± 1.0 ms, respectively (Figure S3). The slight decrease in the T₂ after extraction is likely due to the removal of soluble MOCs, which should have a higher T₂ than the network-bound MOCs. For comparison, the T₂ value for as-made Gel₃ (f = 6) was 4.7 ± 0.6 ms, which confirms that the MOCs in this material are more restricted compared to Gel₉. In a control experiment, free MOCs were added to Gel₃; as expected, the measured T₂ value increased (to 6.3 ± 0.6 ms).

The structure of these polyMOCs was also characterized by small-angle X-ray scattering (SAXS)(Figure 3). For Gel₀, a broad and weak peak was observed at q = 0.065 Å⁻¹. As n increases, three peaks emerge and sharpen. The q value for the first peak ranges from 0.79 for Gel₃ to 0.90 Å⁻¹ for Gel₁₁, while the second and third peaks remain constant at q = 0.29 and 0.58 Å⁻¹. The latter two peaks agree well with the form factors of a 3.5 nm nanoparticle; they are assigned to the MOCs embedded within the network.^[37] The data suggests that as n increases, the fidelity of MOC formation also increases. The low q SAXS peak for each polyMOC is assigned as the average distance between adjacent MOCs linked by polymer chains (Figure 3, inset). This distance decreases as n, which is proportional to the MOC concentration, increases.

Figure 3.

Small-angle X-ray scattering (SAXS) profiles of polyMOC gels.

Next, we used shear oscillatory rheometry to investigate how n impacts the storage (G′) and loss (G″) moduli of these materials. First, we note that for all samples from n = 0 to 13 the G′ values were larger than the G″ values at all tested frequencies; these materials are elastic solids (Figure S4). A plot of G′ (at ω = 1 rad/s) versus n reveals a maximum at n = 3 (Figure 4a). Initially, G′ increases rapidly: from 400 ± 100 Pa for Gel₀ to 3800 ± 200 Pa for Gel₃; i.e., Gel₃ was nearly 10 times stiffer than Gel₀, which is remarkable given that the ~18 SMLs in each MOC in Gel₃ are not elastically effective. As n increases beyond 3, the polyMOCs become softer: G′ dropped to 500 ± 50 Pa for Gel₁₃. In these examples, a stoichiometric amount of Pd²⁺ was used to fully coordinate all PLs and SMLs. As expected, off-stoichiometry studies between Pd²⁺ and pyridine ligands led to decreases in G′, since the crosslinking density decreases when Pd²⁺ is in deficiency and f decreases when Pd²⁺ is in excess (See SI, Table S2).

Figure 4.

(a) Shear elastic modulus (G′) of polyMOC gels with different n values and schematics showing the proposed classes of network structure as n increases.

The observed relationship between G′ and n can be rationalized by considering how n differentially affects the network crosslink density and f. As n increases, the MOC density, and therefore the crosslink density, increases (as confirmed by SAXS). At the same time, f decreases since each MOC must contain fewer polymer-bound ligands. Since n and f are inversely related, a plot of n × (f − 2) (assuming that linear junctions, f = 2, do not contribute to elasticity) closely resembles Figure 4a (Figure S5). Notably, the swelling ratios of these materials increased with n, suggesting that the network mesh size increases despite the fact that the MOC concentration increases (Figure S6).

Based on the data above, we can divide the structure of these polyMOCs into four characteristic classes defined by n (Figure 4). In class I, n < 1 (f > 12): the MOC concentration is low and the MOCs likely have a large fraction of topological defects^[43–44] in order to pack >12 polymer chains around a single MOC (as indicated by broad peaks in NMR and SAXS). Thus, the materials are very soft. In class II, 1 ≤ n < 6, (3.4 < f ≤ 12): SMLs facilitate the formation of well-defined MOC junctions, the crosslink density increases and f is quite large, which results in large G′ values. In class III, 6 ≤ n < 11 (2 < f ≤ 3.4): the MOC concentration is high, but f is low, and the materials are soft. Finally, in class IV, n ≥ 11: the MOC stoichiometry is such that on average two or fewer polymer chains are attached to each MOC. Thus, some MOCs no longer serve as crosslink junctions, but instead as linear linkers between polymer chains or dangling ends. Cage-saturated star polymers may also be present.

Next, we sought to evaluate the stress relaxation dynamics of these polyMOC gels. Stress relaxation is a critical parameter in polymer network design that is exploited in synthetic and biological polymer networks to achieve unique time-dependent behaviors. To our knowledge, there are no reports on stress relaxation in polyMOC materials. However, Fujita and coworkers have shown that ligand exchange in analogous M₁₂L₂₄ MOCs in solution is very slow due to cooperativity effects.^[45–46] At ambient temperature, the metal-coordination bonds in these MOCs are considered to be as static as covalent bonds; they become dynamic again only when heated to >70 °C. Since the dynamic nature of supramolecular networks is related to the rates of ligand exchange and polymer chain diffusion,^{[42, 47–48]} we reasoned that stress relaxation in our polyMOCs could be dependent on n, and thus a diverse range of temperature-dependent mechanical timescales could be accessed.

Stress-relaxation studies were conducted to measure G′ under constant strain as a function of time at various temperatures for polyMOCs of varied n (Figure 5). As shown for Gel₉ in Figure 5a, at 25 °C and 40 °C the gels show little relaxation; at 55 °C and 70 °C they relaxed progressively more rapidly. Following from these observations, the gels undergo self-healing when heated above 70 °C (Figure S7). Figure 5b–d and Figure S8 compare the relaxation behavior of polyMOCs with different n values at various temperatures. Gel₁, Gel₅, and Gel₉ are chosen as they fall into the different aforementioned network classes (I, II, and III, respectively). At 40 °C, all the gels show similarly slow relaxation (Figure S8). At 55 °C, both Gel₁ and Gel₉ relaxed more rapidly than Gel₅ (Figure 5b). Fitting using the Kohlrausch model^[49] (solid lines in Figure 5b–c, see SI for details), provided the characteristic relaxation times τ: 2693 s, 11352 s, and 5804 s for Gel₁, Gel₅, and Gel₉, respectively. A similar trend was observed at 70 °C, as shown in Figure 5c–d.

Figure 5.

Stress-relaxation curves for selected polyMOC gels. (a) Gel₉ at different temperatures; open circles: raw data; solid lines: fitted data. (b,c) Comparison of Gel₁, Gel₅, and Gel₉ at 55 °C and 70 °C. (d) Comparison of the characteristic relaxation times for polyMOCs Gel₁, Gel₅, and Gel₉.

Interestingly, the trend for τ versus n follows the network classifications depicted in Figure 4 and mirrors the G’ data in Figure S5. For Gel₁ (class I), topological defects and ill-formed MOCs facilitate fast relaxation.^[45–46] For Gel₅ (class II), the MOCs are well-formed and thus the metal-ligand bonds are static. Furthermore, the crosslink density is high and polymer diffusion may be slow, leading to slowed relaxation even in the presence of SMLs. For Gel₉ (class III), f is low and the high concentration of MOCs enables fast ligand exchange and enhanced relaxation (SMLs should diffuse faster than PLs). Taken together, these results reveal that polyMOCs have dynamic covalent nature, and that their dynamics can be easily tuned through addition of free SMLs.

Finally, having shown that SMLs can readily be incorporated into MOCs to provide a range of novel network structures and dynamics, we sought to use an alternative SML to selectively install functionality into the polyMOC junctions, thus demonstrating control over structure, dynamics, and function in this system. We prepared a series of Gel₇ derivatives where the total amount of ligand was held constant, but 5%, 20%, or 60% of the methyl-ether SML was replaced with a pyrene-modified SML (Figure 6a). After swelling and extraction, the three polyMOCs had the same volume, which suggests that their network structures are similar as would be expected if the pyrene-SMLs were incorporated into the MOCs without otherwise changing the network connectivity. Fluorescence spectra of the gels confirmed that the increase in fluorescence intensity is qualitatively proportional to the amount of pyrene-SML incorporated (quantum yields were 0.08 for 5% and 60% gel, and 0.07 for 20% gel; see SI). The fluorescence can also be observed under a bench top UV lamp (Figure 6b, inset). These observations confirm that varied amounts of functional SMLs can be introduced to broaden the functional diversity of polyMOCs.

Figure 6.

(a) The chemical structure of methyl-SML and pyrene-modified SML. (b) Fluorescence emission spectra for Gel₇ containing 5%, 20%, and 60% of pyrene-modified SML. Inset: pictures of gels under bench-top UV lamp.

In conclusion, we have described a simple yet versatile approach –based on three-component assembly– that enables the synthesis of star polyMOC gels with diverse network structures, mechanical properties, dynamics, and functionality. Our strategy should translate to other potential polyMOC materials, e.g., ones based on the M₃₀L₆₀ cage recently reported by Fujita et al.^[50] Importantly, our results show that MOCs can be used to precisely tune the structure and dynamics of polymer networks.