Carbodiimide Ring-Opening Metathesis Polymerization

Controlled incorporation of nitrogen into macromolecular skeletons is a long-standing challenge whose resolution would enable the preparation of soft materials with the scalability of man-made plastics and functionality of Nature’s proteins. Nylons and polyurethanes notwithstanding, nitrogen-rich polymer backbones remain scarce, and their synthesis typically lacks precision. Here we report a strategy that begins to address this limitation founded on a mechanistic discovery: ring-opening metathesis polymerization (ROMP) of carbodiimides followed by carbodiimide derivatization. An iridium guanidinate complex was found to initiate and catalyze ROMP of N-aryl and N-alkyl cyclic carbodiimides. Nucleophilic addition to the resulting polycarbodiimides enabled the preparation of polyureas, polythioureas, and polyguanidinates with varied architectures. This work advances the foundations of metathesis chemistry and opens the door to systematic investigations of structure-folding-property relationships in nitrogen-rich macromolecules.

A remarkable feature of polymers produced in Nature is the abundance of heteroatoms in their backbones. Proteins, whose backbones are one-third nitrogen, are an excellent example of this general fact ( Figure 1A). Backbone nitrogen abundance is pivotal for protein folding, which in turn determines its enzymatic, signaling, and structural functions. 1,2 In contrast, the backbones of synthetic polymers are, by and large, composed of carbon and have disordered 3D structures. Nitrogen-rich backbones are precedented among synthetic polymers but are relatively scarce: some key examples are polyamides, 3 polyurethanes, 4 polyethylenimine, 5 polycarbodiimides, 6−8 and polyphosphazenes 9 ( Figure 1B). Furthermore, most of them remain challenging to prepare with high structural precision. New strategies toward the synthesis of nitrogen-rich polymer backbones that could lend themselves to greater precision would unlock untapped functional potential of synthetic polymers. Here we demonstrate such a strategy founded on the discovery of carbodiimide ring-opening metathesis polymerization (CDI ROMP).
Carbodiimides (CDIs) stand out as some of the most versatile nitrogen-containing functional groups. A notable feature of CDIs is that through nucleophilic addition, they can be converted quantitatively into a broad range of other nitrogencontaining functional groups: for example, ureas, thioureas, guanidines, and a long list of heterocycles. 10−12 This feature would be particularly valuable in a macromolecular context: numerous classes of nitrogen-rich polymer backbones could be derived in one step from main-chain poly(carbodiimide)s (polyCDIs). Indeed, existing polyCDIs have already demon-strated industrial utility as antihydrolysis stabilizers in foams and cross-linkers for fiber reinforcement and insulating coatings. 10,13 Yet, the full potential of polyCDIs and their derivatives remains unrealized due to the challenges posed by their synthesis.
PolyCDIs are currently synthesized via decarboxylative stepgrowth polycondensation of diisocyanates ( Figure 1C). 10,15 This method has been state-of-the-art since 1962 despite the fact that it suffers from cross-linking side-reactions, as well as the limitations inherent in step-growth methods: high molecular weight dispersity (Đ ∼ 2), difficulty to attain high-molecular weight, and poor control over polymer architecture. Chain growth methods for polyCDI synthesis are unknown at present. Note that although chain growth coordination/insertion polymerization of carbodiimides has been thoroughly investigated by the Novak group, the resulting polymers no longer have carbodiimide functional groups and are more akin to poly(guanidine)s. 6−8 Given the unsaturated nature of CDIs and inspired by the farreaching impacts of ROMP of cyclic alkenes 21,22 and allenes, 23,24 we envisioned that CDI ROMP could, likewise, prove highly impactful: it could grant entry to entirely new compositions and architectures of polyCDIs with previously inaccessible levels of precision and molecular weights. This precision could in principle, also be inherited by the derivatives of polyCDIs obtained through postpolymerization modification. Catalytic CDI metathesis has a handful of precedents in literature, though these examples are exclusively of the cross-metathesis variety ( Figure 1D): transition metal (TM) imido, iminocarbene, and guanidinate complexes as well as iminophosphoranes have been established as capable catalysts in this context. 16−20 Notably, these examples present a mechanistic dichotomy of the metathesis process: the imido/iminocarbene complexes and iminophosphoranes mediate [2 + 2] cycloaddition/elimination, while the iridium guanidinate complex 1 is believed to undergo insertion/elimination with carbodiimides. 16−20 Incidentally, the former generally requires temperatures ≥100°C to proceed, while the latter is rapid even at ∼20°C. 16−20 We envisioned that the high cross-metathesis activity of 1 under mild conditions would enable it to be coopted for ROMP of cyclic CDIs ( Figure  1E). 16 Two known cyclic CDIs�M1 25 and M2 26,27 ( Figure 2A)� representative of N,N-diaryl and N,N-dialkyl CDIs were selected as monomers. On the basis of computations of homodesmotic reactions ( Figure S1 of the Supporting Information, SI), we predicted both of these CDIs to have moderate ring-strain energies�9.5 and 8.1 kcal/mol for M1 and M2, respectively� and sufficient overall free energies of ring-opening (−2.3 and −1.9 kcal/mol, respectively) to drive ROMP. 28 On the basis of 1 H nuclear magnetic resonance (NMR) spectroscopy, both were found to polymerize rapidly and nearly quantitatively at 23°C and [monomer] = 0.5 M in the presence of 1 in a range of organic solvents, including ethers, haloalkanes, and aromatics  Table S1). Meanwhile, Đ increases from ∼1.2 to 2 with conversion ( Figures  2D and S4), which suggests that chain transfer is operative in the polymerization of M1. Some chain transfer was anticipated due to the presence of linear CDI units in the growing polymer chain in addition to di-p-tolyl CDI (L1) generated at the initiation stage (vide infra) from 1 ( Figure S10). To validate this hypothesis, polyM1 was prepared and treated in situ with additional L1 (five equivalents relative to 1) before termination, which resulted in an ∼80% reduction in M n over 16 h at 23°C ( Figure S11). Nonetheless, though the polymerization of M1 is not living due to chain transfer, it exhibits good molecular weight control, matching closely with theoretical values up to experimental DPs of at least 750. Notably, though L1 generated during initiation could lead to eventual halving of the polymer M n , crude 1 H NMR spectra of the polymerization of M1 reveal that L1 is only partially consumed during the time scale of polymerization, which explains the close agreement of the theoretical and experimental DPs ( Figure S12).
In contrast to M1, experimental M n -s of M2 deviate negatively from the theoretical ones ( Figures 2D and S5). The negative deviation is consistent with chain transfer and/or establishment of ring−chain equilibrium earlier in the polymerization. 29 The same trend is observed for experimental vs theoretical DPs of polyM2 obtained by varying the [M2]: [1] ratio and taking monomer conversion into account (Figures 2E, S8, and S9; Table S2). In addition to previous experiments displaying clear chain transfer with L1 in M1 polymerization, this hypothesis is also supported by the growth of characteristic p-tolyl-CDI polymer end-group resonances and the disappearance of L1 ones in the crude 1 H NMR spectra during the first 2 h (∼71% conversion) of the polymerization of M2 (Figures S13−S15). Furthermore, for [M2]:[1] = 500:1 and 1000:1, low-molecular weight species are observed in the differential refractive index (dRI) GPC trace of crude polyM2, integrating to 19 and 25% of the sample, respectively ( Figure S8). Nonetheless, despite the extensive chain transfer/backbiting, the DP of polyM2 could be controlled by tuning the [M2]: [1] up to 500:1, and DPs of up to ∼400 can be obtained experimentally ( Figure 2E).

■ CDI ROMP MECHANISM
At the outset, we hypothesized that CDI ROMP proceeds via an insertion-elimination mechanism analogous to the one described by Bergman and Holland ( Figures 1D, 3A, and 3B). 16 In our proposed mechanism, initiation begins with the insertion of a CDI monomer into the iridium−nitrogen bond of 1, converting the four-membered iridocycle into a six-membered one. Next, L1 is eliminated from the six-membered iridocycle to form a four-membered one, which now has the monomer incorporated; at this point, initiation is complete. Another round of insertion and elimination leads to the ring-opening of the first inserted monomer. Propagation proceeds through further monomer insertion and elimination/ring-opening, which we believed at the outset would be driven by the release of ring strain energy. All the steps described above are expected to be reversible. Insertion of linear CDIs in growing polymers via backbiting or intermolecular cross-metathesis leads to chain transfer ( Figure S10). Our kinetic studies described above are consistent with this mechanism.
We turned to theoretical analysis to gain further insights into the insertion and elimination processes and catalyst selectivity for cyclic versus linear CDIs. Density functional theory (DFT) 32 was utilized to model the metathesis mechanism in both the case of M1 (Figures 3A and 3B) and a linear analogue di-o-tolyl CDI (L2, Figures S16) with complex 1. Specifically, we used the PBE-D2 functional 33−35 with a split basis set: 6-311+G(d,p) for N, 36,37 6-311G(d,p) for C and H, 37 and LANL2DZ with included effective core potential (ECP) 38 augmented with one fpolarization function (0.938) for Ir 39 and CPCM solvation model for THF. 40,41 These computations shed light on the mechanism of insertion: the transition state (TS) for a concerted mechanism could not be located; meanwhile, initial coordination of the CDI to Ir allowed us to locate the insertion TS (TS1) with an overall activation barrier of 12.9 kcal/mol. Notably, insertion is  Figure S16). This contrasting behavior suggests that the ring strain present in M1 is largely alleviated upon insertion due to the change in the N− C−N bond angle from 170°to 130°.
Insertion of M1 was found to form the tetra-coordinate iridobicyclic intermediate 3, as opposed to the tricoordinate intermediate 3* (Figure S17) previously postulated by Holland and Bergman; 3* is a viable structure in principle, but it was computed as both a higher-energy intermediate and one that is formed via a higher-barrier insertion ( Figure S17). 16 Completing initiation, elimination of L1 via TS2 (Figures 3A and 3B), followed by dissociation of L1, was computed to have a relatively low barrier (13.9 kcal/mol) and to be 4.9 kcal/mol "uphill" from 3, though still overall 6 kcal/mol "downhill" from the start of initiation.
Propagation begins with coordination/insertion of another M1 into 5, forming irido-bicylic intermediate 7 through TS3. Elimination/ring-opening of M1 is expected to occur directly from 7 to coordinated intermediate 8 via TS4. Both TS3 and TS4 are expected to be viable transition states because they have been located and their energies computed at lower levels of theory (see "Computational Methodology" in SI). All in all, during propagation, 2.5 kcal/mol are released, which is remarkably close to the calculated ΔG for the homodesmotic reaction of M1 (−2.3 kcal/mol, Figure S1). The overall weakly exergonic propagation is consistent with experimentally observed establishment of ring−chain equilibria at high monomer conversions.
Lastly, computations support the hypothesis of chain transfer via competitive insertion between linear and cyclic CDIs, with a ΔΔG ‡ of 5.8 kcal/mol in favor of L2 (Figures 3A, 3B, and S16). Importantly, while the use of L2 is informative of chain transfer with monomeric linear CDIs, it does not fully capture the steric bulk of polyCDIs. Therefore, conclusions drawn about chain transfer behavior with polyCDIs�either via backbiting or to other polymer chains ( Figure S10B, S10C)�from this model should be viewed in this context.

■ POSTPOLYMERIZATION MODIFICATION
Postpolymerization modification of polyM1 enabled the synthesis of analogous polyguanidines, polythioureas, and polyureas following modified procedures previously reported for other polyCDIs or small molecule CDIs ( Figure 4A). In the presence of n-butylamine, complete conversion of polyM1 to the corresponding polyguanidine polyM1-N was observed after 7 min at 23°C (Figure 4A), 43 judging from changes in the 1 H and 13 C NMR and attenuated total reflection Fourier transform infrared (ATR-FTIR) spectra ( Figures S18−S20). Quantitative conversion of polyM1 into bottlebrush polymer polyM1-BB was also achieved after a 1-day reaction of polyM1 at 40°C with monoamine-terminated polydimethylsiloxane (PDMS) with M n = 2 kg/mol ( Figures S18, S21, and S22). 43 The bottlebrush architecture of polyM1-BB was confirmed by atomic force microscopy (AFM) ( Figure 4B).
PolyM1 was also quantitatively converted to the corresponding polythiourea polyM1-S by treating an N,N-dimethylformamide (DMF)/THF solution of the former with H 2 S (1 M in THF) for 1 h, as validated by 1 H and 13 C NMR and ATR-FTIR spectroscopy (Figures S18, S23, and S24). 44 Lastly, treatment of polyM1 in a solution of 25 mM LiBr in DMF with 5 M HCl (aq) furnished polyurea polyM1-O after 1 h at 40°C. 44 Quantitative conversion in this case was confirmed by solid-state 13 C NMR and ATR-FTIR spectroscopy (Figures S18 and S25). Attempts to determine the molecular weight of the modified polymers by GPC-MALS proved unsuccessful either due to the interactions of the polymers with the column stationary phase, 45−47 intramolecular hydrogen-bonding interactions/aggregation, 48 poor solubility in the mobile phase, or a combination thereof, leading to distorted elution profiles ( Figure S26) or inability to collect a spectrum altogether.

■ THERMAL PROPERTIES
Derivatization of polyM1 led to substantial changes in the materials' thermal properties as measured by thermogravimetric analysis (TGA) and differential scanning calorimetry (DSC) ( Figure 4C and 4D). With the exception of polyM1-BB, all derivatives of polyM1 displayed reduced temperatures at which 5% mass loss was observed, with polyM1-S being the least thermally stable of the series ( Figure 4C). Interestingly, in contrast to polyM1-O, polyM1-N, and polyM1-BB, both polyM1 and polyM1-S displayed incomplete (∼90%) ultimate mass loss which suggests that these two polymers undergo carbonization at high temperatures. The mechanistic factors responsible for this dichotomy are unclear at present.
DSC revealed a surprising similarity in glass transition temperatures (T g -s) for polyM1 and polyM1-N: 36°C vs 37°C , respectively ( Figure 4D). However, polyM1-S and polyM1-O have broader glass transitions centered at higher temperatures of 72 and 67°C; further changes in the DSC profile of polyM1-S at higher temperatures are believed to be due to polymer degradation, based on TGA ( Figure 4C). PolyM1-BB does not display any thermal transitions within the accessible temperature range of the instrument, and no melting transitions are observed for any of the derivatives.
In contrast to polyM1 and its derivatives, polyM2 retains ≥95% of its mass until 321°C after which it loses all of the mass in three stages ( Figure S27). The corresponding DSC data reveals a small exotherm at 75°C, and a large exotherm at ∼150°C , which is attributed to CDI cycloaddition events ( Figure  S28). Upon a second heating, the exotherm is no longer present, no new thermal transitions are observed at temperatures below that initial exotherm, and a new T g is observed at 137°C.

■ CONCLUSIONS
In this report, we demonstrate rapid, clean, and efficient CDI ROMP of both N,N-diaryl and N,N-dialkyl cyclic CDIs using an iridium guanidinate catalyst. Experimental and computational mechanistic analysis are consistent with an insertion/elimination metathesis mechanism, as well as the presence of competitive chain transfer reactions. Nonetheless, this method can produce polyCDIs with DPs up to 750 in a controlled manner. Furthermore, the N,N-diaryl polyCDIs can be transformed quantitatively into polyureas and polythioureas, as well as both linear and bottlebrush polyguanidines. Lastly, thermal stability and phase transitions of all of these materials have been measured and found to vary considerably depending on the identity of the nitrogen-containing functional groups. Thus, we lay the foundation for the development of controlled synthesis of a broad array of existing and novel nitrogen-rich polymer backbones, which will enable a systematic evaluation of structure-folding/morphology-property relationships in these valuable classes of materials. ■ ASSOCIATED CONTENT
Materials and methods, experimental procedures, supplementary Figures S1−S28 referenced in the main text, Tables S1 and S2, GPC characterization, spectral characterization, thermal property characterization, and crystallographic data (PDF) Computational data (PDF) Crystallographic data for 1 (CIF)