Iron Catalyzed Synthesis and Chemical Recycling of Telechelic 1,3-Enchained Oligocyclobutanes

Abstract

Closed-loop recycling offers the opportunity to help mitigate plastic waste through reversible polymer construction and deconstruction. While examples of the chemical recycling polymers are known, few have been applied to materials derived from abundant commodity olefinic monomers that are the building blocks of ubiquitous plastic resins. Here we describe a [2+2] cycloaddition oligomerization of 1,3-butadiene to yield a previously unrealized telechelic microstructure of (1,n’-divinyl)oligocyclobutane. This material is thermally stable, has stereoregular segments arising from chain-end control, and exhibits high crystallinity even at low molecular weight. Exposure of the oligocyclobutane to vacuum in the presence of the pyridine(diimine) iron precatalyst used to synthesize it resulted in deoligomerization to generate pristine butadiene, demonstrating a rare example of closed-loop chemical recycling of an oligomeric material derived from a commodity hydrocarbon feedstock.

Single-use plastics have become ubiquitous in consumer and industrial products that enable conveniences of modern life due to their low cost, lightweight, durability, and high stability. However, these features also present a challenge for waste management, resulting in the environmental accumulation of end-of-life plastics.^1–4 With current recycling rates of approximately 10%, the fate of the majority of commodity plastics are either in landfills or the environment.^{2, 5} Additionally, current mechanical recycling practices incur substantial loss-of-value in each cycle producing materials with inferior physical properties.²

Chemical recycling offers an alternative strategy and holds promise for up-cycling waste plastics to their constituent monomers or other value-added products.^{2, 6–8} Recent efforts toward chemical recycling have focused on the introduction of new polymeric structures bonded through covalent, heteroatomic linkages that are susceptible to reversible chemical cleavage (Figure 1A).^9–15 While these examples provide valuable proof-of-concept, state-of-the-art methods still require specialized monomers that are dwarfed in scale by those used for commodity hydrocarbon resins (Figure 1B).¹⁶ Further, the attendant properties (i.e., crystallinity, glass transition temperature, stress/strain, chemical inertness, tear, gas-barrier) of these polymers often exhibit attributes that make them unsuitable for commercial application. By contrast, the six most common plastic resins produced industrially exhibit excellent material properties and are derived from monomers obtained directly from abundant petrochemical feedstocks (e.g. ethylene, propylene, styrene, xylene); five are generated exclusively from olefinic monomers.^{17, 18, 19} As a consequence, there is a need to develop chemically recyclable polymers derived from hydrocarbon feedstocks that do not require the use of specialized monomers.^{16, 19–21} Further, these materials need to exhibit physical properties comparable to conventional polyolefins. In this regard, chemically recyclable semi-crystalline polymers are particularly rare. Here we describe the realization of this goal with the synthesis, characterization, and controlled deconstruction of oligomers consisting of 1,3-linked cyclobutane repeat units (Figure 1C). These findings have identified a fundamentally unique microstructure of polybutadiene amenable to closed-loop chemical recycling.

Figure 1. Approaches to the closed-loop recycling of polymers.

a. Current chemically reversible polymers utilize covalent linkages formed from tailored monomers (ref. 11–14). b. Five of the most common plastic resins are constructed from feedstock olefins and consist of chemically recalcitrant linkages. c. Accessing a chemically recyclable, olefin derived oligomer through the [2+2] oligomerization of butadiene.

Results and discussion.

Synthesis and Characterization of the Oligocyclobutane Microstructure.

It is well-established that pyridine(diimine)-supported iron complexes promote the intermolecular [2+2]-cycloaddition between alkenes, or alkenes and dienes.^{22, 23} Mechanistic studies support metallacyclic intermediates arising from oxidative cyclization of the unsaturated hydrocarbon substrates, often giving rise to cyclobutanes with high regio- and stereoselectivities.^22–25 For diene-alkene [2+2] cycloaddition, the iron catalyst engages only one of the alkenes of the diene, suggesting that a cascade cycloaddition oligomerization of 1,3-butadiene to yield a new microstructure of oligobutadiene would be possible (Figure 2A). Dissolving the iron dinitrogen precatalyst, ((^MePDI)FeN₂)₂(μ-N₂) (^MePDI = 2,6-(2,6-Me-C₆H₃-N=CMe)₂C₅H₃N) in neat butadiene resulted in consumption of the liquid monomer after three days at 50 °C and generated a white solid. Removal of the volatiles and extraction of the residue with ethyl acetate yielded a soluble, semisolid fraction and an insoluble, hard, crystalline, light tan powder (Figure 2B).

Figure 2. A unique microstructure of polybutadiene obtained through iron-catalyzed [2+2]-cycloaddition/oligomerization.

a. Known butadiene polymers have been obtained through insertion mechanisms, whereas polycyclobutane can be obtained from cascade oxidative cyclization/oligomerization. b. Oligomerization of butadiene under elevated temperatures and neat conditions produces lower molecular weight, organic soluble (1,n’-divinyl)polycyclobutane and higher molecular weight, insoluble (1,n’-divinyl)polycyclobutane.

Characterization of the organic soluble fraction by 1D and 2D NMR spectroscopies established that the oligomer microstructure was distinct from that of known polymers resulting from 1,4- or 1,2-addition of butadiene. Instead, the spectroscopic data establish an oligomer of 1,3-linked cyclobutanes terminated by vinyl groups on both chain ends. NMR spectroscopic analysis of the crystalline fraction in 1,1,2,2-tetrachloroethane-d₂ at 120 °C revealed analogous features, differing only in the increased ratio of repeat units to chain ends (Supplementary Information, Figure 1 and Figure 2). DFT (Density Functional Theory) simulation of the NMR chemical shifts corroborated a 1,3-linked cyclobutane polymer structure. Quantitation of oligomer molecular weight, based on ¹H NMR integrations, revealed that the organic soluble material has a M_n of 486 g mol⁻¹, corresponding to an average of eight cyclobutyl repeat units, while the insoluble material is 973 g mol⁻¹, or a number average chain of seventeen cyclobutyl rings. While both of these molecular weights were too low to be analyzed by gel permeation chromatography (GPC), the mass values obtained by ¹H NMR spectroscopy were corroborated by mass spectrometry (APCI MS) (Supplementary Information, Figure 4). On multigram scale, molecular weights approximating 580 g mol⁻¹, or ten enchained cyclobutanes, were routinely obtained. Deviation from the above reaction conditions through modification of the reaction temperature resulted in reduced molecular weight oligomers, while introduction of cyclohexane solvent in the reaction mixture ceased oligomerization altogether (Supplementary Information, Table 5).

The iron-catalyzed [2+2] oligomerization of butadiene was also initiated in the presence of ethylene. Adding equimolar amounts of ethylene and 1,3-butadiene to a vessel containing ((^MePDI)FeN₂)₂(μ-N₂) at ambient temperature over two days yielded vinylcyclobutane, as previously reported,²² along with oligomeric products arising from further [2+2] cycloaddition/oligomerization of the formed vinylcyclobutane with butadiene. The resulting oligomers consisted of cyclobutane repeat units terminated by a vinyl group on only one end on the chain (Supplementary Information, Figure 6 and 7). This result indicates that tailored control of the chain end component of the oligocyclobutane is possible through choice of olefin coupling partner, representing a marked deviation from previous reports of 1,3-linked polycyclobutane structures.^26–28

The unique structure of the oligocyclobutanes prompted more detailed characterization studies. While the products of butadiene [2+2]-cycloaddition/oligomerization are intrinsically achiral (meso), each cyclobutane repeat unit may be constituted with 1,3-substituents in either a syn or anti disposition. High field NMR spectroscopy enabled the assignment and quantification of chain end diad sequences and polymer chain triad sequences. The results of comprehensive, quantitative NMR peak identification and integration yield the following statistics on sequence distribution. First, there are equal amounts of syn and anti cyclobutane rings; the same is true at the vinyl chain ends. Second, adjacent cyclobutane rings have a higher probability (~60%) of having the opposite disposition (i.e., anti-syn or syn-anti). Density functional and transition state theory (DFT/TST) calculations for [2+2]-cycloaddition of butadiene by [(^MePDI)Fe] agree with the NMR analysis, namely, the vinyl chain end cyclobutane has equal probability of being anti or syn (ΔΔ(E+ZPE)^‡ < 1 kcal/mol for reductive elimination, see Supplementary Information, Figure 21), and subsequent [2+2]-cycloadditions favor the opposite disposition (ΔΔ(E+ZPE)_TPSSh^‡ ~ 2 kcal/mol for reductive elimination, see Supplementary Information, Figure 21).

While other pyridine(diimine) iron precatalysts were also competent for the [2+2]-cycloaddition oligomerization, they afforded oligomers of similar or lower molecular weight and no perturbation of the previously observed tacticity (Supplementary Information, Table 6). Notably, in situ activation of a pyridine(diimine) iron dihalide precatalyst with magnesium butadiene generated oligocyclobutane of slightly lower molecular weight (~865 g/mol). Linear 1,4-products were also formed and likely arise from incomplete reduction of the iron dihalide precatalyst or other side reactions generating unidentified iron compounds that promote unwanted acyclic material. While the search for other complexes that are able to catalyze this transformation are ongoing, a related redox-active pyridine(dicarbene) iron dinitrogen compound^29,30 as well as a phosphine-ligated cobalt complex known to catalyze [2+2] transformations of olefins^31,32 produced no cyclic oligomer (Supplementary Information, Table 6).

Crystallographic and Thermal Properties of Oligocyclobutanes.

The crystallinity of the insoluble fraction obtained from butadiene homo-oligomerization catalyzed by ((^MePDI)FeN₂)₂(μ-N₂) was confirmed by powder X-ray diffraction studies. While unambiguous structural determination was precluded by the inability to obtain a crystal suitable for single crystal X-ray diffraction, comparison of the powder pattern to that of known hydrocarbon polymers bore no analogy, confirming a unique crystalline microstructure. Variable-temperature Wide Angle X-ray Scattering (VT-WAXS) revealed that the oligomer regained crystallinity upon cooling from 170 °C to 30 °C (Figure 3A and Supplementary Information, Table 7), with a rotator phase appearing at a temperature range of 110–130 °C. Taken together, these data suggest a temperature-dependent ordering of domains within the oligo-cyclobutane structure.

Figure 3. Select thermal data for crystalline (1,n’-divinyl)polycyclobutane.

a. Variable temperature WAXS data indicating amorphous material crystallizing upon cooling. b. DSC data demonstrating retention of melting and crystallization temperatures upon temperature cycling. Isothermal regions have been omitted for clarity; the full data set can be found in the Supplementary Information, Figure 15.

Thermal data obtained on the crystalline oligomer lends insights into the robustness of this new microstructure. Thermal Gravimetric Analysis (TGA) of the crystalline material revealed a bulk decomposition event at an onset temperature of 413 °C (Supplementary Information, Figure 14). Analysis of the volatile decomposition products by TGA-GC/MS indicates that butadiene is not evolved during bulk decomposition; thus, retro-cycloaddition is not thermally induced. The overall thermal properties of the material parallels that of 1,4-polybutadiene of approximately the same M_n, despite the drastically different morphologies of the materials. Differential Scanning Calorimetry (DSC) indicated that the crystalline domains of the cyclobutyl oligomer are recoverable upon heating from −80 °C to 250 °C and subsequently cycling from 250 °C to 30 °C twice (Figure 3B). The bimodal melting temperatures are ascribed to an initial transition of the crystalline domain to a rotator phase (onset at ~ 110 °C), with subsequent transition to completely amorphous material (onset at 135 °C), corroborating the VT-WAXS features. The melting event at 155 °C is analogous to the reported melting temperature of the 90% trans 1,4-polybutadiene congener.³³ Additional DSC analysis at variable scan rates revealed a thermal event at ~ −10 °C consistent with a glass transition (T_g) (Supplementary Information, Figure 16).

To assess the impact of vinyl end-groups on thermal properties, a representative oligomer was hydrogenated to (1,n’-diethyl)oligocyclobutane (see Methods and Supplementary Information, Section I) and its thermal properties examined. By DSC, only small (~ 5 °C) shifts in the thermal events were observed, but the major features between the unsaturated and saturated variants remained unchanged (Supplementary Information, Figure 17). Oxidation induction time testing of both the hydrogenated and native material indicated that both oligomers undergo rapid oxidation (≤10 min), with the hydrogenated material being slightly more resistant to oxidation at 200 °C (Supplementary Information, Figure 18).

Simulation of a crystal lattice representative of the observed scattering data was pursued computationally using molecular dynamics (MD) simulations. Multiple (1-n’-divinyl)cyclobutane oligomers (n = 17) were generated with sequence distributions (syn/anti diads) consistent with those observed by NMR spectroscopy. Low energy conformers obtained from the molecular mechanics simulations were used as inputs to simulate crystalline polymorphs from which a series of low-energy P2₁ symmetric unit-cells were obtained (Supplementary Information, Section V). The polymorph (P2₁; a = 29.76, b = 10.45, and c = 10.53 Å; α = 90.00°, β = 73.64°, γ = 90.00°) that best reproduced the WAXS scattering data is shown in Figure 4A and 4B. The resulting strand is comprised primarily of alternating syn and anti dispositions of cyclobutyl rings, consistent with the NMR and computational data (vide supra). Construction of a supercell of energy minimized conformers resulted in the packing distribution depicted in Figure 4C, which was able to accurately reproduce the room temperature powder diffraction data (Figure 4D). MD simulations were also used to generate an amorphous supercell generating a simulated scattering pattern consistent with experimental data obtained at 170 °C (Supplementary Information, Figure 13).

Figure 4. Molecular Dynamics simulated (1,17’-divinyl)polycyclobutane oligomer.

a. Depiction of an energy-minimized single strand oligomer. Vinyl end-groups are colored red, syn and anti colored diastereomers are shown in blue and green, respectively. b. Depiction of a plane of the P2₁ crystal used to construct the supercell for prediction of the powder diffraction data. c. Final frame of 1 ns MD equilibrated P2₁ supercell used for X-ray predictions, colored by individual oligomer chain. Hydrogens have been omitted for clarity. d. Experimental WAXS data (30 °C) overlaid with simulated scattering data from the MD equilibrated supercell (P2₁; a = 93.32 Å, b = 88.45 Å, c = 92.02 Å; α = 89.99°, β = 74.96°, γ = 89.99°).

Chemical Recycling of Cyclobutane Motifs.

Based on our previous studies on the mechanism of [2+2] cycloadditions of olefins and dienes^22,34 and DFT/TST calculations, a unified catalytic cycle for the generation of oligomeric materials is presented in Figure 5. Examination of the reaction landscape with three different functionals revealed a sequence of transformations in which an oxidative cyclization event occurs from a putative iron bis(diene) intermediate. Oxidative cyclization produces a divinylated metallacycle from which C–C bond-forming reductive elimination occurs to generate the cyclobutane ring. One of the vinyl groups of the formed 1,3-divinylcyclobutane then proceeds as a substrate in another oxidative cyclization event. Propagation by sequences of oxidative cyclization, reductive elimination, and engagement of another vinyl group of the oligocyclobutane result in propagation of the cyclobutyl repeat units along the chain.

Figure 5. Proposed catalytic cycle for the generation of oligocyclobutanes.

a. [2+2] cyclodimerization of butadiene in the initiation cycle (left) generates 1,3-divinylcyclobutane, which then proceeds as an olefin in the oligomerization cycle (right). b. DFT/TST calculation of the catalytic cycle (TPSSh, RKS, LACV3P**+//LACVP** (50 °C and 1 atm)) using three different solvent models (SM8, CPCM [conductor-like polarizable continuum model], gas phase). The calculated Gibbs free energy profile is given in the background, while the SCFE+ZPE energetics appear in the foreground. For the growing polymer chain (INT4) relative to monomer-coordinated species (R) an energy difference of 2 ± 1.2 kcal mol⁻¹ was found regardless of solvent model used, indicating an approximately thermoneutral process (see Supplementary Information, Figure 21 for full reaction profile and transition state structures). Additional stationary points for multiple spin manifolds can be found in the Supplementary Information, Figure 22.

Exploration of the above reaction sequence across three different solvent models at the TPSSh/RKS/LACV3P**+//LACVP** level also indicates that the overall reaction is close to thermoneutral (Figure 5). In this energy landscape analysis, substantial overestimation of entropic contributions excludes the use of free energy differences (ΔG); as such, energy differences were calculated using only the self-consistent field energy (SCFE) plus the zero-point energy (ZPE).³⁵ In the reaction profile in Figure 5, the Gibbs free energy profile is given in the background graph for reference, while the SCFE+ZPE profile appears in the foreground. Comparison of the energy of the growing polymer chain (INT4) and the monomer coordinated (^MePDI)Fe precatalyst (R) reveal that both species are within 2 ± 1.2 kcal mol⁻¹, supporting a near thermoneutral overall process. This is corroborated by the energy difference between the growing oligomer and butadiene, which is, is −7.2 +/− 1.6 and 6.2 +/− 1.6 kcal mol⁻¹ for Δ(E+ZPE) and ΔG, respectively (see Supplementary Information, Table 9). The former energy difference excludes entropy and internal energy contributions which are problematic within the rigid-rotor harmonic approximation. As such, the true energy difference will be higher. The computational data in Figure 5 also corroborate the NMR data identifying alternate dispositions of cyclobutyl rings within the oligomer (vide supra), as the calculations for each cyclobutyl disposition show that the stereochemical probabilities for the 1^st turn-over are equal for initial generation of a syn or anti cyclobutane, while subsequent turn-overs prefer opposite configurations. DFT/TST calculations further established C–C bond-forming reductive elimination as the turnover-limiting step (Figure 5, TS2 and TS4 and Supplementary Information, Figure 21), consistent with previous mechanistic studies on analogous intermolecular [2+2] reactions incorporating dienes.^{22, 34} Energies for the reaction channels calculated with additional functionals and spin manifolds can be found in the Supplementary Information, Figure 22.

A key motivator for pursuing the [2+2]-cycloaddition/oligomerization strategy was the potential for reversibility and chemical recycling – a prospect seemingly accessible given the calculated thermodynamics of the overall reaction. While ring-opening C–C oxidative addition of vinylcyclobutane with ((^MePDI)FeN₂)₂(μ-N₂) and cycloreversion had been demonstrated previously in stoichiometric experiments,²² the viability of this reverse reaction under catalytic conditions had yet to be demonstrated. The catalytic decomposition of such structures was posited to be potentially challenging given the strong energetic preference for coordination of butadiene to the iron catalyst. Control experiments indicated that sequestration of the butadiene formed upon retro-cycloaddition was key to achieving meaningful levels of deoligomerization. As such, activated 5 Å molecular sieves were added to a benzene-d₆ solution containing ((^MePDI)FeN₂)₂(μ-N₂) and vinylcyclobutane to sequester any liberated butadiene. After 6 days at 50 °C, reisolation of the adsorbed volatiles from the sieves and ¹H NMR analysis established 99% consumption of the vinylcyclobutane and recovery of the constituent ethylene (75% yield) and butadiene (94% yield; Figure 6A). Likewise, catalytic retro-[2+2]-cycloaddition of 3-vinyl-1,1’-dicyclobutane, isolated from the [2+2] cycloaddition/oligomerization of ethylene and butadiene (Supplementary Information, Section I), was reverted to the starting hydrocarbons. Under the same conditions, the deoligomerization of 3-vinyl-1,1’-dicyclobutane regenerated 57% of total ethylene and 99% of total butadiene (Figure 6B).

Figure 6. Catalytic chemical recycling of cyclobutane structures.

A. Quantitative decomposition of vinylcyclobutane affording 1 equivalent of butadiene and 1 equivalent of ethylene. B. Quantitative decomposition of 3-vinyl-1,1’-bicyclobutane affording 2 equivalents of butadiene and one equivalent of ethylene. C. Partial deoligomerization of (1,5’-divinyl)polycyclobutane with butadiene as the sole recovered monomer.

Extension to catalytic chemical recycling of (1,n’-divinyl)oligocyclobutane was demonstrated on the organic soluble material with a number-average length of 5 cyclobutyl rings (M_n = 324 g mol⁻¹), obtained from an oligomerization reaction run for 24 hours. In benzene-d₆ solution under static vacuum, 34% of butadiene was recovered after heating at 50 °C for 6 days (Figure 6C). The success of the deoligomerization was dependent on the volume of reaction vessel used, likely a result of product inhibition of butadiene on the activity of the catalyst. Indeed, examination of a catalytic deoligomerization by ¹H NMR spectroscopy indicated that butadiene coordination to the iron catalyst inhibits the retro-cycloaddition process (Figure S20). Extensions to the deoligomerization of an oligomer consisting of an average of ten enchained cyclobutane rings using the above conditions resulted in 5% recovery of butadiene monomer (Supplementary Information, Figure 19). Crosslinking of the oligomer chains does not appear responsible for the low recovery of monomer as a gelation test indicated only minimal amounts (~ 2% by weight, Supplementary Information, Section VIII) of insoluble residue are present in bulk material. More plausibly limitations in experimental setup likely contribute to the relatively low amount of recovered monomer; further, solubilizing the higher molecular weight oligomer must be balanced with the temperatures at which the catalyst can operate without decomposing in solution. Current efforts are devoted to more sophisticated engineering to improve recovery. Nevertheless, the ability to recover pristine butadiene clearly establishes this hydrocarbon oligomer is amenable to chemical recycling using the same catalyst used for its synthesis.

Conclusions.

In summary, sequential [2+2]-cycloaddition of feedstock olefins mediated by pyridine(diimine)-ligated iron catalysts produced oligomeric chains consisting of cyclobutyl rings linked in the 1,3 positions. These unprecedented architectures were chemically deoligomerized back to pristine monomer and thus constitute a rare class of semi-crystalline, chemically recyclable hydrocarbon oligomers derived from inexpensive, ubiquitous feedstock olefins. The telechelic nature of these oligomers also open new prospects for crosslinking and post-synthetic modification procedures for the tailored synthesis of new hydrocarbon-based materials. Current efforts are directed toward preparation of more stereoregular (e.g. iso- and syndiotactic) material and products with higher molecular weight, as well as functionalization of the vinylic chain ends.

Methods:

General Considerations:

All air- and moisture-sensitive manipulations were carried out using vacuum line, Schlenk and cannula techniques or in an MBraun inert atmosphere (nitrogen) dry box unless otherwise noted. All glassware was stored in a pre-heated oven prior to use. The solvents used for air- and moisture-sensitive manipulations were dried and deoxygenated using literature procedures.³⁶ Butadiene and ethylene were purchased in reagent grade from Aldrich. Butadiene was stored over calcium hydride in a thick-walled glass vessel for at least 24 hours and degassed before use. Ethylene was stored over activated 4 Å molecular sieves for at least 24 hours before use. The following compounds were prepared according to literature procedures: vinylcyclobutane,²² Mg(C₄H₆)·2THF,³⁷ (^MePDI)FeCl₂,³⁸ ((^MePDI)FeN₂)₂(μ-N₂),³⁹ ((^Me(Et)PDI)FeN₂)₂(μ-N₂),³⁹ (^iPrPDI)Fe(N₂)₂,⁴⁰ (^(TB)PDI)Fe(N₂)₂,²⁴ (^iPrPDAI)Fe(C₄H₆),⁴¹ (^MePDI)Fe(C₄H₆),⁴² (^MesCNC)Fe(N₂)₂.⁴³

¹H and ¹³C NMR spectra were recorded on Bruker NanoBay 300, Varian iNova 400, or Bruker Avance III 500 spectrometers operating at 300.13 MHz, 399.80/100.54 MHz, and 500.46/125.86 MHz, respectively. All ¹H and ¹³C NMR chemical shifts are reported in ppm relative to SiMe₄ using the ¹H (chloroform-d: 7.26 ppm; benzene-d₆: 7.16 ppm, 1,1,2,2-tetrachloroethane-d₂: 6.00 ppm) and ¹³C (chloroform-d: 77.16 ppm; benzene-d₆: 128.06 ppm, 1,1,2,2-tetrachloroethane-d₂: 73.78m ppm) chemical shifts of the solvent as a standard. ¹H NMR data for diamagnetic compounds are reported as follows: chemical shift, multiplicity (s = singlet, d = doublet, t = triplet, q = quartet, p = pentet, br = broad, m = multiplet, app = apparent, obsc = obscured), coupling constants (Hz), integration, assignment. ¹³C NMR data for diamagnetic compounds are reported as follows: chemical shift, number of protons attached to carbon (e.g. CH₂), assignment.

GC analyses were performed using a Shimadzu GC-2010 gas chromatograph equipped with a Shimadzu AOC-20s autosampler and a Shimadzu SHRXI-5MS capillary column (15 m × 250 μm). The instrument was set to an injection volume of 1 μL, an inlet split ratio of 20:1, and inlet and detector temperatures of 250 °C and 275 °C, respectively. UHP-grade S3 helium was used as carrier gas with a flow rate of 1.96 mL/min. APCI-MS analyses were performed on an Advion ExpressIon APCI MS instrument with an ASAP injection probe. MS scans were taken in positive ion mode with a scan time of 500 ms, scan delay of 100 μs, and a mass detection range of 100 to 1200 Da.

Variable temperature wide-angle scattering (WAXS) data were obtained on a SAXSLAB Ganesha 300XL X-ray Scattering Instrument. This instrument contains a motorized vacuum-compatible Pilatus 300K 2D detector from Dectris that can be moved from 90–1400mm inside the chamber. Temperature data were obtained in a Linkam HFSX350 stage by suspending a small amount of sample between two 6um sheets of mica in a small sandwich cell. The scattering images were masked of any single-crystal diffraction peaks from the mica before data reduction. WAXS sample-to-detector distance was calibrated with a LaB₆ standard.

TGA/GCMS analyses were performed on a TGA-8000 Thermogravimetric Analyzer, Clarus 680 Gas Chromatography, Clarus SQ 8 T Mass Spectrometry hyphenated system equipped with a PerkinElmer Elite 5MS column (30 m× 250 μm). GC injections were triggered at a TGA temperature of 450 °C. The GC inlet temperature was set to 250 °C, and the temperature method utilized was as follows: hold at 50 °C for 1 min, ramp to 250 °C at 20 °C/min, hold at 250 °C for 4 min (total time 15 min). EI+ MS scans were taken every 0.2 s for 15 min with an interscan delay of 0.1 s. The mass range analyzed was from 32 to 300 Da. DSC analyses were performed on a PerkinElmer double-furnace automated multi-sample DSC-8500 Differential Scanning Calorimeter (DSC).

Computational modelling was performed to facilitate analysis of experimental NMR and WAXS data. For the computations reported in our manuscript, NMR analysis utilized the B3LYP-D3 function, while reaction energies were calculated with hybrid B3LYP-D3, meta-GGA M06-L, and meta-hybrid TPSSh functionals. We found the latter functional to be most consistent with experiment. Further TPSSh was selected as the DFT functional of choice for four reasons: 1) it is the best performing functional to correctly model the energy difference between both spin states for a wide range of metals and oxidation states, ranging from electron configurations d⁴ to d⁷ (Cr(II), Mn(III), Mn(II), Fe(III), Fe(II), and Co(I)).⁴⁴ In this study TPSSh outperformed OPBE, M06-L, B3LYP*, CB3LYP, B3LYP, and M06. 2) TPSSh outperformed B3LYP and BP86 in a benchmark study of 80 transition metal complexes.⁴⁵ 3) TPSSh has been shown to be suitable for [2+2] cycloaddition reactions, outperforming B3LYP and other functionals.⁴⁶ 4) TPSSh can accurately predict Mossbauer parameters.⁴⁷ Full computational details are provided in the Supplementary Information (Section II for NMR, Section V for scattering data). Other details for DFT simulations performed for reaction energies are provided in Section IX.

Preparation of (1,n’-divinyl)oligocyclobutane:

A 50 mL thick walled glass vessel containing a stir bar was charged with 12 mg (0.0125 mmol) of ((^MePDI)FeN₂)₂(μ-N₂) in a nitrogen-filled glovebox. The flask was sealed, brought out of the glove box, and degassed on a high vacuum line. Butadiene (12.5 mmol) was added to the reaction vessel using a calibrated gas bulb. The flask was sealed, warmed to room temperature, then added to an oil bath at 50 °C for 72 hours. Complete consumption of liquid butadiene and deposition of a white solid on the walls of the flask was observed. After 72 hours, the reaction was quenched by vacuum transferring the volatiles into a flask containing 800 – 1000 μL of chloroform-d for analysis by ¹H NMR, in which only residual butadiene was observed. The nonvolatiles were extracted with approximately 10 mL of ethyl acetate, producing a light tan solution and off-white precipitate. The ethyl acetate soluble fraction was decanted away from the insoluble material and passed through a pipet plug containing ~ 3 cm of silica, eluting with additional ethyl acetate. Both fractions were evaporated to dryness. The ethyl acetate soluble material was washed with approximately 5 mL of methanol and dried under vacuum to yield 0.043 g (7% yield based on starting butadiene consumed) of (1,n’-divinyl)oligocyclobutane (M_n ~ 486 g mol⁻¹, 8 total cyclobutyls) as a white semisolid. Additionally, 0.299 g (47% yield based on starting butadiene consumed) of ethyl acetate insoluble (1,n’-divinyl)oligocyclobutane (M_n ~ 973 g mol⁻¹, 17 total cyclobutyls) was isolated as a light tan crystalline powder.

For analysis of the ethyl acetate insoluble material by NMR, approximately 3 mL of o,o-dichlorobenzene was added to approximately 25 mg of the light tan powder. The heterogeneous mixture was heated with stirring to 120 °C in a vial, and dissolution of the solid was observed. The hot solution was added dropwise into a vial containing 5 mL of vigorously stirring methanol at ambient temperature. The resultant white precipitate was isolated by vacuum filtration and suspended in 600 μL of 1,1,2,2-tetrachloroethane-d₂. The suspension was heated with stirring to 120 °C, after which the contents were transferred to an NMR tube and analyzed by ¹H and ¹³C NMR at a probe temperature of 120 °C. Characterization data can be found in the Supplementary Information, Figures 1–3.

Preparation of (1,n’-diethyl)oligocyclobutane:

A glass-lined autoclave (600 mL) equipped with PTFE coated magnetic stir bar was charged with (1,n’-divinyl)oligocyclobutane (M_n ~ 581 g/mol, ~10 cyclobutyls, 1.00 g), cyclohexane (100 mL) and PtO₂ (0.01 g, 2.5 mol %). The autoclave was then sealed and pressurized with H₂ (100 PSIG). It was then heated to 100 °C where it was allowed to react for 20 h. The reactor was then cooled to ~80 °C where it was vented and opened. The reaction mixture, a homogenous, pale, yellow solution was filtered hot to remove PtO₂. The filtrate was then concentrated to a solid under reduced pressure to afford (1,n’-diethyl)oligocyclobutane as a colorless powder. Yield: 0.75 g, 75 %. The material was characterized by ¹H and ¹³C{¹H} NMR (TCE-d₂; 120°C) spectroscopy and ATR spectroscopy to confirm consumption of vinyl groups. Characterization data can be found in the Supplementary Information, Figure 5.

Preparation of 3-vinyl-1,1’-dicyclobutane:

A 200 mL thick walled glass vessel containing a stir bar was charged with 93 mg (0.2 mmol) of ((^MePDI)FeN₂)₂(μ-N₂) in a nitrogen-filled glovebox. The flask was sealed, brought out of the glove box, and degassed on a high vacuum line. Butadiene (20 mmol) was added to the reaction vessel using a calibrated gas bulb. Without thawing the flask, ethylene (20 mmol) was added to the reaction vessel using a calibrated gas bulb. The flask was thawed and stirred at ambient temperature for 48 hours. Deposition of a liquid on the bottom of the flask was observed. After 48 hours, the volatiles were vacuum transferred at ambient temperature into another flask. Analysis of this material by ¹H NMR indicated the formation of vinylcyclobutane (1.00 g, 61% isolated yield), consistent with a previous literature report.²² 3-vinyl-1,1’-dicyclobutane was isolated from the remaining residue by vacuum transfer while gently heating the bulk material with a heat gun, and was obtained in 15% isolated yield (0.400 mg). Characterization data can be found in the Supplementary Information, Figure 7.

General Procedure for Chemical Recycling of Cyclobutane Structures:

To a 50 mL thick walled glass vessel in the glove box was added a stir bar and 8 mg of 5 Å molecular sieves. 0.0264 mmol of substrate in 700 μL of benzene-d₆ was added to the vessel using a syringe. A stock solution containing 5 mg (0.0066 mmol) of ((^MePDI)FeN₂)₂(μ-N₂) in 400 μL of benzene-d₆ was freshly prepared, of which 99 μL (0.00132 mmol, 0.05 equiv.) was added to the reaction vessel. The vessel was immediately sealed, frozen in liquid nitrogen, and evacuated on a high vacuum line. The flask was then thawed, added to an oil bath at 50 °C, and stirred vigorously for 6 days. After this time, all volatile contents of the reaction flask were vacuum transferred to a J Young tube containing 50–150 μL of a 0.0416 M solution of 1,3,5-trimethoxybenzene in benzene-d₆, and the volatiles were analyzed by ¹H NMR. Yields of butadiene and ethylene are reported relative to the 1,3,5-trimethoxybenzene internal standard, using the peak appearing at 4.93 ppm for quantification of butadiene. The resultant NMR spectra can be found in the Supplementary Information, Figure 19.