Polymeric frustrated Lewis pairs in CO2/cyclic ether coupling catalysis

Frustrated Lewis pairs (FLPs) are now ubiquitous as metal-free catalysts in an array of different chemical transformations. In this paper we show that this reactivity can be transferred to a polymeric system, offering advantageous opportunities at the interface between catalysis and stimuli-responsive materials. Formation of cyclic carbonates from cyclic ethers using CO2 as a C1 feedstock continues to be dominated by metal-based systems. When paired with a suitable nucleophile, discrete aryl or alkyl boranes have shown significant promise as metal-free Lewis acidic alternatives, although catalyst reuse remains illusive. Herein, we leverage the reactivity of FLPs in a polymeric system to promote CO2/cyclic ether coupling catalysis that can be tuned for the desired epoxide or oxetane substrate. Moreover, these macromolecular FLPs can be reused across multiple reaction cycles, further increasing their appeal over analogous small molecule systems.


Introduction
Valorisation of CO 2 as a renewable carbon feedstock is desirable in the pursuit of a sustainable, carbon-neutral society. Current efforts to use this typically unreactive substrate rely either on designer gas-capture systems and subsequent reactions, or combining CO 2 directly with suitably reactive substrates. 1 Epoxides are one such reactive substrate that has been extensively explored, where insertion of CO 2 yields the corresponding cyclic or polycarbonate products. 2 While the majority of early research efforts in this area have employed expensive or nonabundant metal-based systems, the use of metal-free catalysts to facilitate these transformations has expanded in recent years. 3 Frustrated Lewis pairs (FLPs) are ubiquitous as metal-free catalysts for a myriad of chemical transformations. [4][5][6][7] The frustration induced by sterically hindered Lewis acid (LA) and base (LB) centres permits cooperative action of, and catalysis with, a multitude of substrates including carbon dioxide, dihydrogen and cyclic ethers. [8][9][10][11][12][13][14][15] In 2017, our group demonstrated that FLP reactivity is maintained when incorporated into a polymeric system, revealing a new class of stimuli-responsive materials that exploit FLP-mediated small molecule activation. 16,17 Inspired by these systems, Yan et al. later reported similar macromolecular FLPs capable of activating CO 2 and catalysing amine formylation. 18 Polymeric FLPs have also been reported in C-H functionalisation, amination and hydrogenation catalysis with demonstrated potential for catalyst recovery and reuse. [19][20][21][22] Recently, we reported the successful crosslinking of highly Lewis acidic styrenic copolymers with a corresponding Lewis basic copolymer, via ring-opening of cyclic ether substrates. 23,24 Given these results, we wondered whether our systems would enable effective catalytic insertion of CO 2 into the ring-opened cyclic ether substrates. Aryl and alkyl boranes have previously been applied successfully in the formation of both cyclic and polycarbonates when paired with a phosphonium or ammonium salt partner. [25][26][27][28][29][30][31] More recently, superbasic phosphazenes have also proved active under mild conditions. 32 The use of a phosphine LB in these reactions is however, to our knowledge, previously unreported. Herein, we report the rst use of both conventional small molecule and polymeric FLPs to catalyse the insertion of CO 2 into cyclic ether substrates with high selectivity towards the cyclic product (Fig. 1).
networks, suggesting that the addition of CO 2 promotes decomposition aer the rst turnover. Decomposition occurs at both moderate and elevated temperatures and mimics reversible reactions commonly associated with polymeric alkyl boranes, 33 therefore preventing further catalytic testing using B2.
Aryl-boron-containing polymers have a higher thermal stability. 21 This was proven when pairing a polymeric LB with B1 and B3. While the capacity of B3 to ring-open cyclic ethers was already established, 24 we anticipated B1 would also facilitate this reaction based on previous small-molecule systems. 15,27 Previous studies into small molecule borane catalysed formation of cyclic carbonates reported a reactivity decrease when using stronger LAs as a result of adduct stability. 27 Indeed, the same trend can be observed within our polymeric systems when considering estimations of Lewis acidity by the Gutmann-Beckett method. 34,35 Reactions employing B3 (acceptor number, AN ¼ 84.4) were notably slower than those with the less Lewis acidic B1 (AN ¼ 67.6) under the same optimised conditions ( Fig. 2 and S5 †). Given its superior performance, we selected B1 as the Lewis acidic component for the rest of this study.
With catalytic conversions using B1 remaining low, optimisation of the system was essential. We hypothesised that catalyst decomposition was occurring, as extended reaction times did not increase product formation. Analysis of the reaction mixture by 31 P NMR spectroscopy revealed the presence of a new oxidised phosphine species, a commonly encountered off-cycle product during phosphine-promoted catalysis. 36 As direct CO 2 activation is not observed with this FLP system under the mild conditions used in catalysis, the mechanism likely proceeds via ring-opening of the epoxide substrate, followed by CO 2 insertion as reported with other aryl borane-containing systems. 27 Triphenylphosphine oxidation was not observed in the absence of epoxide, meaning that an on-cycle decomposition process involving PO must be occurring. Deoxygenation could potentially form propene, however analysis of the crude reaction mixture by 1 H NMR spectroscopy revealed no obvious side products, potentially due to their gaseous nature.
As we previously noted that ring-opening of styrene oxide (SO) by poly(FLP) systems was rapid, 23 we hoped to use this substrate to better understand this catalytic system. Under the same conditions, only 19% conversion from SO to styrene carbonate was observed. However, the system showed increased phosphine oxidation and the concomitant formation of styrene as conrmed by the presence of vinyl peaks at 5.3 ppm and 5.8 ppm in 1 H NMR spectra ( Fig. S7 and S9 †). Inspired by this nding, we re-attempted the reaction with PO under milder conditions and preventing gaseous product release, and indeed observed propene in situ (Fig. S8 †).
Evidence of alkene formation prompted us to consider the possible routes to epoxide deoxygenation that would result in phosphine oxidation. Previous work had shown ring-opening of epoxides by tertiary phosphines results in Wittig-like reactivity, 37 while heating an FLP activated N 2 O complex releases N 2 gas, forming a P]O-B linked species. 38 Ring-opening of epis-uldes, the sulfuric analogue of epoxides, also leads to formation of similar linkages. 15 It is thus hypothesised that, at elevated temperatures, CO 2 insertion is competitive with alkene   elimination and phosphine oxidation, gradually leading to catalyst decomposition (Fig. 3).
Catalyst activity could thus be renewed through the addition of disiloxane reducing agents 36,[39][40][41][42] used to regenerate the phosphine. Indeed, we found that addition of 1,3-diphenyldisiloxane (DPDS) to our PO/B1/P1/CO 2 reaction mixture allowed for >99% conversion of PO (Fig. S10 †). Although DPDS addition increased catalytic performance, the mix of desired carbonate products and alkene side products obtained when using SO both rendered the reaction less green and made product separation more difficult.
From this understanding, we sought to design a system to preclude this decomposition pathway. In N 2 O systems, elimination of N 2 occurred when the phosphorus and boron components were oriented in the less energetically favourable cis conformation. 38 If the same conformational requirements are present, as previous small molecule studies suggested the cis conguration is preferred for PO activation by a BPh 3 / PtBu 2 Me FLP pair, 15 a bulkier polyphosphine may be preferred, especially as PPh 3 has a lower cone angle than PtBu 2 Me. 43 This doesn't preclude decomposition from the trans congurations as attack of the phosphine copolymer on SO occurs at the more hindered carbon of the epoxide ring (i.e. adjacent to the phenyl ring) as assigned by 31 P-13 C coupling constants (44.5 Hz, Fig. 4A). With this mode of attack, rotation from conformations I to II (Fig. 3) affords the undesired conformation, with decomposition encouraged by the enthalpic driving force of conjugated product formation (styrene).
The switch to our previously reported mesityl-substitute P2 (ref. 16) would increase both steric bulk (favouring the trans attack and precluding rotation) and increase Lewis basicity to enhance its reactivity. The impact of this increased steric bulk from the mesityl groups was readily observable using small molecule model compounds, with a 7 : 1 preference for ringopening at the less hindered carbon observed for SO (Fig. 4B). Under the same reaction conditions, the P2/B1 catalyst combination led to >99% conversion of SO to styrene carbonate. Notably, no vinyl peaks were observed in the reaction mixture by 1 H NMR spectroscopy (Fig. S13 †), with minimal phosphine oxidation observed by 31 P NMR spectroscopy. Taken together, the mesityl groups and polymer backbone bulk mitigate unproductive alkene elimination.

Substrate scope
With an optimised system in hand, we investigated the substrate scope and overall reactivity of the P2/B1 system. Conversion of PO to propylene carbonate reaches >99% conversion in just 7 hours, dramatically outperforming the less Lewis basic P1. Realising that CO 2 /epoxide catalysis with small molecule FLPs is also unreported, we compared our macromolecular catalysts to their small molecule equivalents, BPh 3 and PMes 2 Ph. These conventional FLPs, under the same reaction conditions, gave >99% conversion of PO within just 6 hours (Fig. S27 †), the rst example of conventional FLP use in CO 2 insertion to epoxides. Use of a B/P system therefore signicantly outperforms previously reported organocatalytic systems which typically require higher catalyst loadings. 3,44 Use of a phosphine LB also allows for milder conditions than ammonium salt systems, 29 while maintaining the same selectivity, although phosphonium systems are substantially more active. 27 A variety of substrates were screened for P2/B1 using similar reaction conditions (Fig. 5). All terminal epoxide substrates were quantitatively converted in under 24 hours, although their rates differed markedly depending on sterics and electronics. 3-Chloropropylene oxide (CPO) outperformed all other tested substrates, consistent with trends reported for other systems. 27,45 While the electron withdrawing chlorine promotes rapid turnover, the sterically similar methyl group (butylene oxide, BO) shows decreased conversion. The role of steric bulk was also apparent for internal epoxides. While ring-opening of cyclohexene oxide (CHO) is possible using both P2/B1 and our previously reported poly(FLP) systems, 23,24 productive conversion into product is not observed. Previous attempts to use conventional FLPs in the copolymerisation of CHO and carbonyl sulde were also unsuccessful, 26 although borane/ phosphonium salts do form polycarbonates. 27,28 To check if this was due to the strain induced during CO 2 insertion preventing cyclohexyl rearrangements, cis-2,3-butylene oxide, CBO, was trialled. With no conversion for this derivative, it is likely down to steric clashes during ring-closure that prevent product formation.

Reactivity with oxetane substrates
Notably, oxetane (OX) can also be successfully ring-opened by P2/B1. Conversion to trimethylene carbonate (TMC) was slower due to the lower ring strain of OX relative to epoxide substrates, 46,47 as well as the relatively high strength of the intermediate crosslinked networks formed as reported in our previous work. 23 Attempts to expand the OX substrate scope with both 3-bromooxetane and 3,3-dimethyloxetane were unsuccessful. Although a change in chemical shi was observed in 11 B NMR spectra, indicating LA coordination, no change was observed by 31 P NMR spectroscopy for the LB counterpart, suggesting steric clashes prevent nucleophilic attack. Selectivity for cyclic 6-membered carbonates over their polymeric counterparts is oen cited as problematic when using a CO 2 insertion synthetic route. 48 With epoxide substrates, high selectivity for the cyclic product is achieved at all temperatures screened. However, if the high temperatures optimised for epoxide ring opening are used (120 C) with oxetane, only 61% preference for the cyclic product TMC (Table 1, entry 1) is observed. Lowering the temperature to 100 C gives a signicant improvement in selectivity, although longer reaction times are required to achieve high conversions (Table 1, entry 2 and 3). This additionally supports the formation of the polycarbonate product directly, rather than via TMC polymerisation. Interestingly, while better leaving groups have been proposed as routes to improve cyclic product selectivity in small molecule catalysts, 29 the opposite trend is observed herein where the better P1 leaving group gave less selectivity (Table 1, entry 4).

Catalyst recyclability
Polymer-supported catalysts have also shown potential for facile reuse or incorporation into ow reactors as an enabler for green manufacturing. 49,50 Previous polymeric Lewis bases have been synthesised in pursuit of multiple use cycles but require more forcing conditions and suffer from limited substrate scopes. 51,52 We thus decided to explore the catalytic performance of P2/B1 across multiple reaction cycles (Fig. 6) using a precipitationrecovery strategy (Section 6, ESI †), comparing it to the spiking of the reaction mixture with fresh substrate. This more rigorous recovery procedure is a true stress test of catalyst reuse using P2/B1.
As can be seen in Fig. 6, and for either reuse methodology, catalytic performance decreases signicantly for P2/B1 aer the rst reaction cycle. Smaller rate decreases are observed over subsequent cycles. Gradual phosphine oxidation can be observed (Fig. S40 †), although attempted regeneration using DPDS did not substantially increase reaction rates, as the   Fig. S31-S34).

Fig. 6
Conversions obtained in consecutive reaction cycles for CO 2 insertion to PO, as calculated by 1 H NMR spectroscopy, using 1,3,5trimethoxybenzene as an internal standard. Reaction conditions as stated in Fig. 5. The data shown in green was obtained by adding fresh substrate at the end of each cycle (i.e. without purification and catalyst recovery). The recovered P2/B1 catalysts remain active after 4 cycles, achieving >99% conversion with extended reaction times (cycle 5).
additional steric bulk from the mesityl groups could prohibit silane attack. 39 It is important to note that this is a reduction in rate only, as extending the reaction time gives quantitative conversion (Fig. 6, cycle 5 and S39 †). It is important to highlight that these macromolecular catalysts are generated in situ, with the addition of substrate triggering gelation. This means that the accessibility of the catalyst active site is a key factor in reuse efficiency. Postreaction, we observed that the catalyst resting state was present in its gelated, cross-linked form, supported by 31 P and 11 B NMR spectroscopy, and GPC measurements of hydrodynamic volume (Fig. S43 and S44 †). Precipitation thus forms a superstructure that does not dissolve, but instead swells (Fig. S42 †), creating a physical constraint on substrate access to catalytic sites. 20 In our case, crosslinking occurs at the active components, limiting polymer chain segmental motions in the vicinity of the FLP and likely suppressing the rate of epoxide capture and CO 2 insertion. This hypothesis is consistent with the observation that a signicant decrease in activity is only observed aer the rst cycle. Future work in the team directed at incorporating these metal-free catalysts into ow reactors is an important step in overcoming these practical constraints.

Conclusions
In summary, we report the rst use of both conventional and polymeric FLPs in the formation of carbonates from cyclic ethers using CO 2 , with high selectivity for the cyclic carbonate product. The synthesised macromolecular catalysts displayed good reactivity for a variety of terminal epoxide substrates and are also active catalysts for the formation of trimethylene carbonate from oxetane and CO 2 . Post-reaction, the poly(FLPs) used can be easily recovered and reused, with gradual decreases in catalyst efficiency attributed to partial phosphine oxidation as well as increased crosslinking. Further tuning of both the catalytic system and reaction conditions is expected to offer a more diverse substrate scope with higher selectivity for oxetane substrates. Further development of intermolecular poly(FLPs) for catalysis is expected to instead utilise the effects seen from crosslinking to enhance reactivity with other substrates via compartmentalisation.

Author contributions
Thomas A. R. Horton and Meng Wang: investigation, methodology and manuscript writing. Michael P. Shaver: conceptualisation, manuscript writingreview and editing, supervision and funding acquisition.

Conflicts of interest
There are no conicts to declare.