Starting materials were purchased from Sigma-Aldrich and are of the highest available purities. Complexes Na[W₂(µ-Cl)₃Cl₄(THF)₂]·(THF)₃ (1) [25], (ⁿBu₄N)₃[W₂(µ-Cl)₃Cl₆] (2) [26] and (ⁿBu₄N)[W₂(μ-Cl)₃Cl₅(THF)] (3) [27] were prepared according to literature procedures. NBE was dissolved in the solvent used in the reaction, dried by stirring with CaH₂ under argon and was distilled under vacuum prior to use. NBD was passed through an Al₂O₃ column. THF and diethyl ether were distilled over Na/Ph₂CO, toluene and hexane over Na, CH₂Cl₂ over CaH₂, and methanol over sodium methoxide. Benzaldehyde was purified by distillation under reduced pressure. All solvents were distilled in an inert atmosphere, and were degassed by three freeze-pump-thaw cycles, with the exception of methanol, which was degassed by bubbling nitrogen or argon for 0.5 h. All operations were performed under a pure dinitrogen or argon atmosphere, using Schlenk techniques on an inert gas/vacuum manifold or in a drybox (O₂, H₂O < 1 ppm).

NMR spectra were recorded on a Varian Unity Plus 300 spectrometer. In all cases, chemical shifts are reported in ppm relative to the deuterated solvent resonances. Size exclusion chromatography (SEC) experiments were carried out with a modular instrument consisting of a Waters model 600 pump, a Waters model U6K sample injector, a Waters model 410 differential refractometer and a set of 4 μ-Styragel columns with a continuous porosity range of 10⁶–10³ Å. The columns were housed in an oven thermostated at 40 °C. THF was the carrier solvent at a flow rate of 1 mL/min. The instrument was calibrated with PS standards covering the molecular weight range of 4000–900,000.

A typical procedure is described as follows. The substrate (e.g., NBE, 423 mg, 4.5 mmol) was added to a solution of complex (e.g., 1:9.0 mg, 0.009 mmol) in a solvent (2.0 mL). The mixture was allowed to react for a given time (see Table 1), after which it was concentrated to half volume and treated with excess methanol to precipitate the polymeric products. The resulting solids were filtered and washed repeatedly with methanol. They were redissolved in THF and the above procedure was repeated. The products were dried in vacuo.

Complex 1 was dissolved in THF-d⁸ (8.0 mg, 0.008 mmol) and the green solution was transferred in an NMR tube. The appropriate amount of the substrate (NBD: 15.1 mg, 16 μL, 0.16 mmol; NBE: 150.6 mg, 1.60 mmol) was added using a microliter syringe.

The stereochemistry of the polymers obtained (PNBE, PNBE–COOME, PVNBE and the soluble part of PNBD) was determined by ¹H and ¹³C NMR [28,29,30,31,32].

Complex 1 has been isolated and structurally characterized [25]. In the solid state it contains a tungsten–tungsten triple bond, bearing three halide bridges and two THF ligands, one in each tungsten atom, in a cis arrangement along the dimetal axis. Such species in solution, in the presence of donor ligands or coordinating solvents, may exist in an equilibrium between the highly symmetric confacial (D_3h, fasbo) structure 1 and the edge-sharing (D_2h, esbo) biooctahedral one 1' (Scheme S1 in the Supporting Information) [33]. It is air sensitive (oxygen, moisture), but in the solid state, it is stable at room temperature for a few hours. It is soluble in THF, CH₃CN and dme, less soluble in CH₂Cl₂, and insoluble in toluene, Et₂O and CS₂. The complex has been repeatedly recrystallized and checked carefully for purity (Vis-UV) before use. Polymerization reactions were carried out at room temperature, for a given time. The results are summarized in Table 1.

The polymerization reactions of NBE were carried out in different media and were found to be solvent dependent. They do not proceed in strongly coordinating or chelating solvents (CH₃CN, dimethoxyethane (dme)), indicating the inhibition of substrate coordination. They are slow in THF affording small yields (entry 1), while in CH₂Cl₂ fast gelation is observed and the polymer has low polydispersity index (M_w/M_n, entry 2). Suspension of 1 in toluene gives moderate yield of PNBE (entry 5), while suspensions in Et₂O and CS₂ (entries 6 and 7) give high yields of the polymer. After the end of the heterogeneous reactions, the catalyst can be separated and reused (at least four times) with fresh amounts of monomer and solvent, without significant loss of reactivity. The supernatant solution remains colorless during the reaction. The cis-stereoselectivity is high (80%–86%) and is not affected by the reaction conditions.

All polymers are soluble in common organic solvents (CHCl₃, CH₂Cl₂, THF). Their geometric structure was determined by ¹H and ¹³C NMR spectra (Figure S1) [28]. The relative proportions of double-bond pair sequences, represented as trans-cis (tc), trans-trans (tt), cis-cis (cc) and cis-trans (ct) units, were determined from the four methine carbon (C^1,4) signals of the ¹³C NMR spectrum of PNBE at δ_C 43.67 (tc), 43.44 (tt), 38.88 (cc) and 38.67 ppm (ct). The fraction of cis double bonds (σ_c = 0.85) estimated from this ¹³C NMR spectrum is in good agreement with that obtained from the ¹H NMR spectrum (σ_c = 0.86 from the integrals of the signals at δ_H 2.73 (HC^1,4cis-PNBE) and 2.37 ppm (HC^1,4trans-PNBE). The reactivity ratios r_c=cc/ct = 7.4, r_t=tt/tc = 1.3 and r_cr_t = 9.6 were calculated from the heights of the relevant signals in the ¹³C NMR spectra.

Comparison of 1 (as per reactivity, yield, M_w, and cis-selectivity) with other representative catalytic systems revealed that it is much more reactive than the Masuda type W(VI) halides, resembling more to the bicomponent ones (e.g., WCl₆/AlEtCl₂, SnMe₄ etc.) and to the high-valent well-defined Schrock carbenes. Its cis-specificity is much more enhanced than that of the aforementioned systems, but it is comparable to that of the low-valent (Katz) and high-valent design catalysts (Schrock, Basset).

Substrates having strongly coordinating substituent groups (–COOH, –OH, –CN) deactivate 1, while those with softer ones (–COOMe, –CH=CH₂) are quantitatively polymerized, albeit with lower rates than NBE (Table 1, entries 11 and 12). The polymers are white, elastomeric, soluble in organic solvents and they have been characterized from their ¹H and ¹³C NMR spectra. The broadness of the peak of olefinic protons (5.26–5.37 ppm) of NBE–COOMe (because of overlapping signals of endo-, exo-isomers, possibly presence of HH, HT and TT units) and the complexity of the ¹³C NMR spectra prevents the estimation of the cis-content [29,30]. In the case of PVNBE, the peak at 5.30 ppm suggests that the ring-strained C=C bond is cleaved, while the vinylic one is left intact. The overlapping signals of the olefinic protons of the polymeric chain with the vinylic ones prevent the stereoregular assignment for the same reasons as above [30]. NBE–COOMe has been polymerized by a variety of catalytic systems (Mo, W, Re, Ru, Os, Ir) with yields varying from 20% to quantitative and molecular weights higher than 800,000 [29]. In contrast, there is only one report for the polymerization of VNBE, namely with the [(OC)₄W(µ-Cl)₃W(GeCl₃)(CO)₃] (4) complex under controlled conditions, giving high yields of PVNBE along with small amounts of side products (~10%) [31]. Both of these monomers are of interest, as their polymers can be easily functionalized by the addition of new side groups.

Finally, it must be noted that the isostructural and isoelectronic perhalocomplex (ⁿBu₄N)₃[W₂(µ-Cl)₃Cl₆] (2) [26] under the same conditions was found inactive towards the ROMP of NBE, whereas the monosubstituted W₂⁷⁺ derivative (ⁿBu₄N)[W₂(μ-Cl)₃Cl₅(THF)] (3) [27] provided very small yield (~10%).

The reactions of NBD with 1 homogeneously or heterogeneously (bulk or suspension in toluene) occur very fast and in quantitative yields (Table 1, entries 13–16). In a solvent-free regime, the polymer can be detached from the surface of the catalyst, which can be reused for several times. The PNBD formed is insoluble in most common organic solvents, indicating extensive cross-linking of the cyclopentene rings. At lower concentrations (1/NBD: 1/20) a partly soluble polymer (CHCl₃) is obtained. The microstructure of the soluble PNBD has been analyzed [32] (Figure S2). From the integral ratio of signal due to olefinic protons δ_H = 5.40 (trans double bond) and δ_H = 5.60 ppm (cis double bond), the value of the cis double bond content was calculated to be ca. 63%. A cis double bond content of ca. 68% was calculated from the ¹³C NMR spectrum. For this purpose, the well-resolved peaks of the methine carbon (C^1,4) at δ_C 48.58 (trans double bond) and 43.95 ppm (cis double bond) were integrated.

Mechanistically, the metal-olefin transformation to active carbenes or metallocyclobutanes in “ill-defined” catalytic systems is an intriguing problem, and several mechanistic pathways concerning mostly mononuclear complexes have been proposed [2,3,34,35,36,37,38,39,40]. In the simple case of NBE, the rearrangement is believed to be a net 1,2-H-shift along the metal coordinated C=C bond, based on experimental evidence [36,37,38,39,40,41]. The higher unsaturation of NBD makes its transformations far more complicated than it had been considered and remain elusive.

In relation to this work, worth noting in some detail are the results reported by Szymańska-Buzar et al. on the mechanism of ROMP of NBE and NBD with a series of heterobimetallic M–M' (M = Mo, W; M' = Sn, Ge) complexes, e.g., [MCl(SnCl)₃(CO)₃(NCMe)₂] (5, M = W; 5', M = Mo) and [(OC)₄Mo(µ-Cl)₃Mo(SnCl₃)(CO)₃] (6) [35]. In the case of 5/NBD, the first step of the reaction has been unequivocally established with the isolation and characterization (spectroscopic, structural) of the [WCl(SnCl₃)(CO)₃(η⁴-NBD)] (5a) adduct, in which the two C=C bonds of the NBD ligand are nonequivalently bonded to the tungsten atom. This compound has also been proven to be an efficient precatalyst for the ROMP of NBD, along with formation of minor amounts of exo-2-chloronorborn-5-ene and exo-3-chloronortricyclane [42]. Monitoring (¹H and ¹³C NMR) the reactions (5 or 5a in CDCl₃ or CDCl₃/CD₂Cl₂, room temperature) does not reveal the carbenic protons of the active species that trigger polymerization (attributed to their very small concentration). Exploratory density functional theory (DFT) calculations on the mechanism of the spontaneous η⁴-NBD to carbene transformation have been carried out for complexes [MoCl(SnCl₃)(CO)₂(η⁴-NBD)] 5'a [42], [MoCl(SnCl₃)(CO)₂(NCMe)₂(η⁴-NBD)] (7) [42], [W(CO)₃I₂(η⁴-NBD)] (8) [43] (forty pathways overall have been considered), but the activation energies were found too high. More recent calculations for [W(CO)I(η²-dppm)(η⁴-NBD)] (9) taking into account extra important intermediate steps [44], which lower the activation energies by half, favor the mechanism first hypothesized by Handzlik et al. [42] (oxidative coupling to form metallacyclopentane, followed by a 1,4-H-shift between the two α-bonded carbons across the face of the metal atom (Scheme S2).

In the case of NBE (4, 5a, 6), the corresponding η²-NBE adducts are not isolable, but can be detected by ¹H NMR. Apart from polymer formation, exo-2-chloronorbornane and 2,2'-binorbornylidene (bi-NBE) were detected. The formation of bi-NBE provides direct evidence of η²-NBE transformation to a tungstanorbornylidene species (1,2-hydride shift) capable to initiate either the ROMP of NBE, or a carbene–carbene coupling reaction, giving a new olefin (bi-NBE) [45,46,47].

The small initiation efficiency of 1 (it is calculated that only ~10% of 1 is activated by NBE) coupled to the sensitivity of the active species, undermined our efforts to isolate adducts of 1 with NBD and NBE or any other catalytically active yield-related intermediates. Similar was the fate of attempts to establish whether the active species are bimetallic or 1 serves as a precursor to mononuclear compounds (by ESI-MS). Nevertheless, valuable information has been obtained by monitoring in situ the reactions of 1 with the substrates by ¹H and ¹³C NMR. Solvent of choice was tetrahydrofuran (THF) because of the rapidity of the polymerizations in halogenated solvents (CHCl₃, CH₂Cl₂, PNBD formation even at low ratio (1/NBD: 1/2), gelation at higher ones) and the insufficient solubility of 1 in the above solvents at lower temperatures.

In the case of NBD, as soon as it was added to a catalyst solution (1/NBD 1/20, THF-d⁸; Figures S3 and S4) the peaks at 4.30 and 1.98 ppm, due to the coordinated THF molecules, instantly diminish and gradually disappear, indicating their fast displacement by the substrate, but no peaks due to adduct formation could be detected. Instead, at t = 0–1 h, multiple broad unresolved high field peaks in the region 9.80–13.80 ppm, characteristic of the propagating tungsten alkylidene PNBD chains [2,48,49] appear simultaneously with peaks at 6.18, 5.90, 3.78 and 2.92 ppm, and one at 3.92 ppm, which could be assigned to the formation of exo-2-chloronorborn-5-ene and exo-3-chloronortricyclane, respectively [42]. As the reaction progresses, the intensity of the above peaks increases, along with parallel and increasing formation of PNBD. Corroborative evidence for the alkylidene ligand(s) by ¹³C–NMR could not be obtained, presumably as a consequence of their minute concentration. The multiplicity of the signals could be attributed either to formation of several active species deriving from the 1/NBD interaction, to fasbo/esbo equilibria and/or concurrent metathetical reactions. Their broad unresolved nature may be due to paramagnetism of the active species [50], restricted rotation of the alkylidene ligand(s) and/or the existence of several species with nominally similar structural features and slightly differing environments, possibly isomeric alkylidene THF adducts of the type {“W_x”(=CHP)(THF)_y} (P = polymer chain; x, y = 1 or 2). Quenching the reaction mixture with benzaldehyde (1/NBD/PhCHO: 1/20/100) causes the disappearance of the high-field peaks.

Interestingly, detection of the 1/NBD interaction has been possible at stoichiometric catalyst to substrate ratio (Figure S5). Under these conditions, ligand (THF) displacement is slower and accompanied with observation of small peaks at 5.44, 5.20 and 4.62 ppm (ratio ≈ 2:2:1) indicative of coordinated olefinic protons (absorptions are at higher field than those of free NBD by 1.28, 1.50 and 2.10 ppm, respectively). When substitution is nearly complete (t = 4 h), the first two nearly disappear, while the third increases and becomes dominant. This peak may be assigned to the equivalently bound olefinic protons of an η⁴-NBD ligand, either bridging the two tungsten atoms (1a) or chelating at one of them (1'a, Scheme 2). In support of formation of 1a is the considerable insight on alkyne activation at bimetallic metal–metal bonded Group 6 centers [33,51,52], and in this particular case, the formation of [W₂(µ-Cl)₃Cl₄(η²-C₂R₂)₂]⁻ and dimetallacarbotetrahedrane complexes [W₂(µ-Cl)₂Cl₆(µ-C₂R₂)]²⁻ (R = Me, Ph) deriving from the reaction of 1 with internal alkynes [53]; also, a limited number of representative alkene and diene interactions [54,55,56,57,58,59,60].

Evidently, our findings are remarkably similar to those reported by Szymańska-Buzar et al. on the polymerization of NBE and NBD by 5, 5' and 6[35,42,45,46,47]. However, the NBD-to-carbene transformation in this system is more facile than those occurring in 5a, and 6/NBD, as this is reflected in the stabilities of the relevant adducts (4a stable and isolable; 1a or 1'a unstable) and the polymerization reaction rates, respectively (5 or 5a/NBD/CHCl₃ or CHCl₃/CH₂Cl₂ slow; 1/NBD/CH₂Cl₂ vigorous).

At this stage, our understanding of the system does not allow for the drawing of solid conclusions. However, taking into account the results reported, those obtained in this work, as well as the characteristics of 1 (available coordination sites, W₂⁶⁺ to W₂⁸⁺ facile interconversions [26], favoring bimetallacycle formation [44], the synergistic effect of the W₂⁶⁺ core that may substantially lower the required activation energies) [44,61], and providing that no other events during the reaction course induce rupture of 1 (e.g., disproportionation, excessive substrate binding), tentatively a bimetallic mechanism seems plausible and appears as a particularly attractive possibility (Scheme 2).

According to this scheme, there is no apparent reason that species 1b/1c should not retain their bimetallic integrity, since they would contain a tungsten–tungsten double bond of formal order of two (W = W)⁸⁺ supported by halide bridges. Such bridged W(IV)–W(IV) double bonds are strong ([W₂(µ-Cl)₂Cl₈)]_n, (2.688(2) Å) [62], [K(18-crown-6)_1.5][W₂(µ-H)(µ-O)(OC^tBu)₆] (2.445(1) Å) [63], [W₂(κ-O₂C^tBu)₄(µ-MeCCMe)₂)] (2.4888(2) Å) [52], [W₂(κ-O₂C^tBu)₄(µ-PhCCMe)₂)] (2.4925(2)Å) [52], [W₂(OCH₂^tBu)₈]_n [56]. The 1b/1c transformation occurring by a-H-abstraction induced by the vicinal tungsten atom, followed by ring opening (C–H bond formation), is a possibility worthy of further investigation. Subsequently, propagation may proceed further via a two-center “windscreen” mechanism. Such an hypothesis is not unprecedented and has been previously proposed to explain the ROMP of NBE along an edge of the trinuclear [Os₃(µ-H)₂(CO)₁₀] cluster [64]. Alternatively, the possibility of 1c being bridging cannot be excluded; in such a case, the reaction may occur via a bimetallic mechanism, as encountered in alkyne and cycloolefin polymerization with the [W₂(CO)₁₀(µ-CMe₂)] [65] and [Cp₂Ru₂(CO)₂(μ-CO)(μ-CH(Me)] [66] bridging carbenes.

In the case of NBE, monitoring of the reactions, apart from the difficulties encountered in CH₂Cl₂, is hampered by additional ones, such as very slow reaction rates even at high 1/NBE (1/200) ratio and the masking of NBE and PNBE at the fingerprint regions of the ¹H NMR spectra, which makes results interpretation problematic. Competition between THF and substrate for ligation is evident, with the THF ligand displacement being very slow. Coordination of NBE could not be detected, but during the course of the reaction (t = 0–20 h) appearance of broad high-field peaks at the region 12.80–9.75 ppm (Figure S6), which may be assigned to the tungsten alkylidene PNBE propagating chains and of peaks at 3.84, 2.38, 2.30, 2.0–1.9 ppm (Figure S7), typical of exo-2-chloronorbornane [46,67], are observed. At longer times (t = 40 h) the characteristic peaks of PNBE at 5.25 and 5.10 ppm appear, whereas the intensity of the alkylidenic proton signals decay, indicating their decomposition. In this case, the evidence concerning the first step(s) of the reaction and the mechanism overall is much less clear. The process may operate either by the coupling of two η²-NBE ligands in a [W₂Cl₇(η²-NBE)₂]⁻ complex, in a manner analogous to that proposed for NBD, or with formation of a tungstenonorbornylidene species, deriving from the [W₂Cl₇(THF)(η²-NBE)]⁻ anion [45]. This bimetallic mechanism may also explain the high cis-efficiency of the reaction, in terms of an enhanced “wall type” [68] steric hindrance along the one side of the dimetal axis. This would oblige the incoming monomer molecule (NBE) to approach the metal center minimizing steric repulsions that is favorable for the formation of cis- rather than trans-junctions in the polymer chain. Such an effect has been considered responsible for the formation of an all-cis-PNBE polymer from the in situ reaction of WCl₆/PSLi/NBE (PS = polystyrene) [69] and the dramatic stereoselectivity changes of the [Mo₂(NCMe)₈](BF₄)₄/EtAlCl₂ ((Mo ⁴ Mo)⁴⁺) system homogeneously (30% cis) and heterogeneously (SiO₂, 70% cis) [14].