Introduction

The palladium-catalyzed substitution reaction of allylic electrophiles has found numerous applications in synthetic organic chemistry1,2,3. Since the late 1980s, a number of polymerization methods leveraging this reactivity have been developed by Saegusa, Okada, Endo, Nomura, Leibfarth, and others, affording unsaturated carbon-chain and heterochain polymers via either a step- or chain-growth mechanism (Fig. 1a)4,5,6,7,8,9,10,11.

Fig. 1: Catalytic polymerization via propargyl/allenyl palladium species: challenge and strategy.
figure 1

a π-Allyl Pd catalysis in polymer synthesis; b Multipathway-trapping of propargylic/allenylic electrophiles challenges their implementation in polymerization, especially via chain-growth; c Working hypothesis; d This work: realization of Pd-catalyzed controlled chain-growth polymerization of VDCP.

Full size image

Despite of the structural similarity, analogous catalysis with propargylic/allenylic electrophiles remains underexplored in polymer science. Arguably, this is due to the dramatically increased complexity brought by an additional π-bond in the intermediate (Fig. 1b)12. In π-allyl catalysis, allylation is always the sole outcome. In contrast, nucleophilic trapping in propargylic/allenic systems can take place via three distinct and interconvertible Pd(II) species, namely the σ-allenyl, π-propargyl, and σ-propargyl palladium complexes, leading to allenylation (type A), alkenylation (type B), and propargylation (type C), respectively13,14,15. The reaction is thus subjected to subtle changes in the substrate-catalyst combination, and substantially less predictable and efficient16,17. To date, only a few attempts have been disclosed by Koizumi, where alkene-based polymers were synthesized in a step-growth fashion, mostly via type B trapping18,19,20.

A chain-growth method still remains unknown. While type B trapping is the most common and efficient, in particular for soft nucleophiles21,22, it virtually precludes a chain-growth reaction as it is followed by either a second nucleophilic attack23 or β-hydride elimination24. To overcome this problem, we became interested in a special allenylic electrophile, that is vinylidenecyclopropane 1,1-dicarboxylate (VDCP). It can be easily synthesized from commodity chemicals, butyne-1,4-diol and malonate in two steps. VDCP can oxidatively add to Pd(0) as reported by Shi and coworkers25,26,27,28. The resulting amphiphilic adduct drew our attention (Fig. 1c). The interaction between the enolate and the palladium center should lead to an unusually strong bias favoring the σ-allenyl complex ΙΙ over its π-propargyl isomer Ι. This might allow one to escape from the frustrating type B trapping if a facile exit is available via ΙΙ. With this consideration, a chain-growth process might become possible involving the terminal-selective nucleophilic attack of ΙΙ by a malonate chain-end, a typical soft nucleophile.

Here we report the Pd-catalyzed controlled ring-opening polymerization of VDCP (Fig. 1d). This reaction features high molecular weight (MW) (Mn up to 94.2 kg/mol), narrow dispersity (Đ ~ 1.1), ultrafast kinetics (full conversion in minutes), and excellent chemoselectivity. A broad scope of end-groups can be incorporated with high fidelity, such as terminal alkenes/alkynes, methacrylates, and even aryl halides, allowing the access to advanced macromolecular architectures. Leveraging the exceptional reactivity of VDCP, we demonstrated gradient pseudo-diblock copolymerization with vinylcyclopropanes (VCP). Experimental and computational evidence was found in support of propargyl/allenyl palladium intermediates.

Results and discussion

We initiated our study by evaluating the polymerization of dimethyl 2-vinylidenecyclopropane-1,1-dicarboxylate (M1) at 40 °C, in the presence of catalytic quantities of Pd2(dba)3 (2 mol% [Pd]) and a bidentate phosphine ligand, DPEPhos (Table 1). At the end of the reaction, trifluoroacetic acid was added to quench the enolates. The presence of an initiator (1a) and a small amount of sodium tert-butoxide base was necessary for a successful polymerization (entry 1). While leaving out the base led to only trace P1, polymerization without 1a took place in an uncontrolled manner, presumably via a sluggish self-initiation process (entries 2–4). No polymerization occurred in the absence of palladium (entry 5). Alternative tert-butoxide bases gave comparable results without observable counterion effect, whereas an organic base, 1,8-diazabicyclo(5.4.0)undec-7-ene (DBU), yielded only oligomers (entries 6–8). Acetonitrile was found a viable alternative solvent among others screened (entries 9–11).

Table 1 Selected Optimizationsa
Full size table

We next evaluated the ligand bite angle effect. It was found that BINAP (93°) provided similar results with DPEPhos (108°) (entry 12). Yet, further reducing the angle to ~78° (dppe) or leave out the phosphine led to only low oligomers (entries 13–14). This is consistent with literature findings that small bite angle ligands favor type B trapping, which frustrates the polymerization16. A monophosphine, i.e., PPh3, promoted the polymerization in a slightly lower efficiency, whereas electron-rich phosphines completely shut down the polymerization (entries 15–16). The polymerization took place smoothly at ambient temperature (entry 17). Finally, varying the [Pd] loading had a minimal impact on the MW, and a narrower dispersity can be obtained at 0.5 mol% [Pd] (entry 18, Mn = 9.2 kg/mol, Đ = 1.08). In fact, full conversion was already achieved within 5 min under these conditions (entry 19).

Given the fast polymerization rate, kinetic profiling was performed under diluted conditions (Fig. 2a, b). The living characteristics the polymerization was evidenced by a linear increase of MW with respect to conversion. The controllability of the reaction was tested by varying the monomer-to-initiator ratio (Fig. 2c). The observed MW well matched the predicted values at [M1]:[1a] up to 200:1 (Mn = 32.2 kg/mol, Đ = 1.07). Going beyond this number led to MW that were lower than expected along with slightly higher dispersity (Mn = 94.2 kg/mol, Đ = 1.27 at 1000:1).

Fig. 2: Controlled polymerization of M1.
figure 2

a, b Kinetic study ([M1]:[1a]:[Pd]:[NaOtBu] = 100:1:0.5:5, [M1] = 0.2 M). c MW control: Mn,Theo = MW(1a) + MW(M1) × conversion × ([M1]/[1a]) d, e Assigned 1H NMR (CDCl3, 400 MHz) spectrum and MALDI-TOF analysis of P1 (Table 1, entry 18).

Full size image

The 1H and 13C NMR spectra of a sample of P1 (Table 1, entry 18) suggest complete ring-opening to give an alkyne-based backbone (Fig. 2d, Figs. S5, S6). The correct peak spacing was identified in matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectroscopy (Fig. 2e). A single set of peaks was found, which is consistent with 1a incorporated at the α-end and protonation at the ω-end. This was corroborated by the 1H NMR end-group analysis. The MW of P1 estimated from NMR integrals, GPC, and mass spectrometry were in good agreement.

Having established the polymer structure, we set out to evaluate the scope of initiators with [M1]:[1] = 50:1 and 0.5% [Pd] (Fig. 3a). 2-Alkylated malonates all demonstrated high-efficiency initiation by giving predetermined MW and well-defined end-group signals on 1H NMR. Useful handles for postpolymerization-transformations can thus be incorporated at the chain-end, including allyls (1b), terminal propargyls (1c), and methacrylates (1d), which highlights the chemoselectivity of the palladium catalysis. Even an aryl bromide (1e) remained intact, indicating a highly selective oxidative addition process. The parent malonate (1f) and methylenes with a similar acidity (1g) are also competent initiators, providing chain growth in both directions29,30. We were able to expand this chemistry to initiators based on a sulfonamide (1h) and a 1,3-diketone (1i), however, the initiation efficiency for the former was moderate. A brief survey of monomer scope was next conducted (Fig. 3b). Notably, we challenged our reaction with bulky monomers that failed in the radical polymerization that we previously developed (M5-6)31. Decent polymerization results can still be obtained using the palladium-catalyzed protocol.

Fig. 3: Scope of the method.
figure 3

a, b Scope of initiators/VDCP monomers ([M]:[1]:[Pd]:[NaOtBu] = 100:2:0.5:5, [M] = 1.0 M). c, d Chain-extension and grafting-from polymerization. e Copolymerization of M1 and M7: kinetics and DOSY analysis (CDCl3, 400 MHz).

Full size image

To further demonstrate the synthetic utility, chain extension, and graft polymerization were performed. Using isolated P1 as a macroinitiator, the synthesis of a diblock copolymer was supported by GPC and diffusion-ordered spectroscopy (DOSY), albeit a small shoulder peak in the GPC trace was observable (Fig. 3c, Fig. S47). Polymers with advanced topologies are important targets for synthesis32. To this end, we show that living radical polymerization of methacrylate 1e and subsequent Pd-catalyzed polymerization successfully produced a graft polymer bearing P1 as side-arms (Fig. 3d).

Without any modification, the same catalyst system enabled a copolymerization reaction of VDCP with its vinyl analog, vinylcyclopropane dicarboxylates (VCP) (Fig. 3e). Kinetic profiling revealed a remarkable reactivity difference between the two monomer classes. VDCP M1 was almost fully consumed within 5 min, at which point the conversion of VCP M7 was merely 8%. Therefore, a gradient copolymer P1-grad-P7 with a pseudo-diblock sequence was formed upon full conversion, which is supported by comparing its NMR spectra with the corresponding homopolymers (Fig. S61). This result again underscores the exceptionally high reactivity of VDCP in palladium-catalyzed polymerization.

A plausible mechanistic hypothesis is depicted in Fig. 4a. The key steps involve the oxidative addition of Pd(0) with VDCP to generate the σ-allenyl complex ΙΙ, and the subsequent nucleophilic attack at the terminal carbon. The propagating chain-end would be most likely a malonate anion rather than an organopalladium33. This is consistent with the end-group structures found and the fact that the MW is independent on the [Pd] loading (Table 1). Besides literature small molecule reactions25,26,27,28, facile ring-opening of VDCP mediated by Pd(0) is supported by trace side product 2 formed in parallel with the polymerization, which was not detected in the absence of palladium. Either leaving out the base (Table 1, entries 2–3) or using electron-rich monophosphines, i.e., PCy3 (Table 1, entry 16), strongly promoted this ring-expansion pathway. This might be attributable to the rapid isomerization of ΙΙ to a η3-palladium enolate with such ligands34. In a stoichiometric experiment, Pd2(dba)3 was first mixed with DPEPhos to afford a red solution (Fig. 4b). An immediate color change occurred upon the addition of M1, which was also noticed in catalytic reactions (Figure S3). An aliquot of this mixture was analyzed by high-resolution mass spectrometry, and molecular ion peaks matching ΙΙ were identified35.

Fig. 4: Mechanistic studies.
figure 4

a Mechanistic hypothesis of the polymerization. b Mass spectrometry detection of the putative oxidative addition adduct. c DFT studies at the SMD(1,4-dioxane)/M06-L-D3/def2-TZVP-SDD//B3LYP-D3(BJ)/6-31 G(d,p)-SDD level. The bond lengths are given in Å.

Full size image

To further shed light on the key trapping step, density function theory (DFT) calculations at the SMD(1,4-dioxane)/M06-L-D3/def2-TZVP-SDD//B3LYP-D3(BJ)/6-31 G(d,p)-SDD level was employed. Indeed, the σ-allenyl palladium (INT2) is substantially more stable than its π-isomer (INT1), and the kinetic barrier for the palladium dancing is very low (TS1, 4.4 kcal/mol). We were able to identify the nucleophilic attack transition state for each complex, respectively. Type B trapping at the central carbon was found kinetically less facile (TS3, 6.9 kcal/mol) than the palladium dancing. Therefore, INT1 likely isomerizes to INT2 before π-propargyl trapping. Subsequently, INT2 undergoes nucleophilic attack by the initiator or polymer living chain end to yield an anionic chain end, enabling the chain growth process.

In conclusion, we present an example of controlled chain-growth polymerization via propargyl/allenyl palladium intermediates. The mechanistic key is a unique allenylic electrophile motif with high reactivity, VDCP, which enables selective trapping via the σ-allenyl palladium complex rather than the more common π-propargyl pathway. Using malonates as the initiators, we demonstrated the synthesis of high MW alkyne-based polymers with high precision, including block, gradient, and graft copolymers. The excellent chemoselectivity and ultrafast kinetics associated with this new method could open up new opportunities for building challenging macromolecular architectures.

Methods

General procedure for polymerization

An oven-dried 10 mL re-sealable screw-cap vial equipped with a Teflon-coated magnetic stir bar was charged with Pd2(dba)3 (1.0 mg, 0.25 mol%, 0.001 mmol), DPEPhos (1.6 mg, 0.75 mol%, 0.003 mmol) and NaOtBu (2.0 mg, 5.0 mol%, 0.02 mmol). The reaction vessel was evacuated and backfilled with nitrogen (this sequence was repeated a total of three times). The reaction vessel was taken into a nitrogen-filled glovebox. 1a (1.2 mg, 2.0 mol%, 0.008 mmol) in anhydrous dioxane (0.40 mL) was added to the reaction vessel. The tube was sealed and taken out of the glovebox. To the tube was added M (0.40 mmol, 1.0 equiv.) via syringe. The reaction mixture was vigorously stirred at r.t. till full conversion of M. The crude reaction mixture was quenched by the addition of a few drops of trifluoroacetic acid, dissolved in a minimal amount of THF, and precipitated by dropwise addition to vigorously stirred MeOH (or a mixture of MeOH/H2O). The precipitate was collected by centrifugation and then reprecipitated from MeOH. The obtained material was dried in vacuo at 40 °C overnight to afford the polymer, which was analyzed by gel permeation chromatography to determine its molecular weight and dispersity. The conversion of the monomer was determined by 1H NMR analysis of the crude reaction mixture.

Synthesis of M1-M6

A round bottom flask equipped with a magnetic stir bar was charged with cesium carbonate (2.5 equiv.) and THF (200–500 mL). A solution of bistosylate (1.0 equiv.) and malonate (1.0 equiv.) in THF (100 mL) was added to the flask dropwise using a constant pressure dropping funnel. The resulting slurry was stirred at 70 °C for 17–21 h. After being cooled to r.t., the reaction mixture was filtered through a plug of celite and eluted with ethyl acetate. The filtrate was dried over anhydrous Na2SO4 and concentrated in vacuo. The residue was purified by silica gel column chromatography to afford M1–M6.

Synthesis of M7

A round bottom flask equipped with a magnetic stir bar was charged with cesium carbonate (2.5 equiv.) and 1,4-dibromo-2-butene (1.0 equiv.), The reaction vessel was evacuated and backfilled with nitrogen (this sequence was repeated a total of three times). Dibenzyl malonate (1.0 equiv.) and THF (60 mL) were added to the flask via syringe. The resulting slurry was stirred at 60 °C for 17 h. After being cooled to r.t., the reaction mixture was filtered through a plug of celite and eluted with ethyl acetate. The filtrate was dried over anhydrous Na2SO4 and concentrated in vacuo. The residue was purified by silica gel column chromatography to afford M7.

Synthesis of 1e

A round bottom flask equipped with a magnetic stir bar was charged with sodium hydride (1.2 equiv.) and THF (50 mL). At 0 °C, add dimethyl malonate, and stir for thirty minutes. Then, add 1-bromo-2-bromomethyl-benzene and continue stirring at room temperature until the reaction is complete. Quench with saturated NH4Cl and extract with ethyl acetate. The filtrate was dried over anhydrous Na2SO4 and concentrated in vacuo. The residue was purified by silica gel column chromatography to afford 1e.

Synthesis of 1h

Tosyl chloride (1.1 equiv.) was slowly added as a solution in 20 mL dry THF to a solution of propargylamine (1.0 equiv.) and diisopropylethylamine 1.1 equiv.) in 100 mL dry THF at r.t., and the resulting mixture was stirred overnight. The reaction mixture was washed with saturated aqueous NH4Cl and brine. The organic phase was dried over Na2SO4 and concentrated. The residue was purified by silica gel column chromatography to afford 1 h.