Isoselective Lactide Ring Opening Polymerisation using [2]Rotaxane Catalysts

Abstract Polylactide (PLA) is a fully biodegradable and recyclable plastic, produced from a bio‐derived monomer: it is a circular economy plastic. Its properties depend upon its stereochemistry and isotactic PLA shows superior thermal‐mechanical performances. Here, a new means to control tacticity by exploiting rotaxane conformational dynamism is described. Dynamic achiral [2]rotaxanes can show high isoselectivity (Pi=0.8, 298 K) without requiring any chiral additives and enchain by a chain end control mechanism. The organocatalytic dynamic stereoselectivity is likely applicable to other small‐molecule and polymerization catalyses.

Polylactide (PLA) is ac ommercially available bio-derived thermoplastic;its biocompatibility and biodegradability have enabled substitution of petro-polymers in medical, packaging and fibre applications. [1] PLAst hermal-mechanical properties depend upon its microstructure:a tactic PLA is amorphous whilst isotactic stereoblock PLA is semi-crystalline. Stereoblock PLA shows an even higher melting temperature (T m )t han isotactic PLLA. [2] All tacticities of PLA are produced by the ring opening polymerization (ROP) of lactide.Acurrent challenge is to deliver stereoblock PLA from racemic lactide (rac-LA) using selective catalysis. [3] Organocatalysts are attractive due to the high polymerization control exhibited and thioureas, [4] amidines, [5] phosphazenes [6] or N-heterocyclic carbenes (NHCs) also show high rates. [7] Very few organocatalysts are isoselective (rac-LA) and the most successful are chiral. [8] Fore xample, binaphthol-derived phosphoric acid (k D /k L = 28), [9] chiral synthetic prolines (P i = 0.87-0.96) [10] and ac hiral thiourea (P i = 0.82) [11] operate by enantiomorphic site control. Fundamentally this mechanism cannot maintain both high rates and conversions:once the preferred enantiomer is consumed the remaining enantiomer reacts very slowly.T he alternative chain end control (CEC) mechanism could overcome this limitation but is under-developed. It has been successfully applied to various metal complexes [12] and there are intriguing hints it may operate for af ew organocatalysts. [6a, 7] Because CEC relies on in operando stereochemical interactions,i ti s much harder to rationally improve selectivity.W ea re interested in understanding how catalyst dynamic processes, for example,l igand fluxionality,m ay enhance stereoselectivity. [12] We previously reported high isoselectivity yttrium catalysts and correlated ligand fluxionality with stereocontrol. [12] Okuda and co-workers showed that fluxional enantiomeric scandium catalysts were heteroselective. [13] Schaper and co-workers showed fluxional copper catalysts were isoselective. [14] So far,r elated examples of dynamic organocatalysts are unknown.
We reasoned that mechanically interlocked molecules such as rotaxanes could allow for dynamic stereocontrol as macrocycle and axle components can be stimulated to move independently.S uch macrocycle/axle dynamic behavior underpins many of their applications,f or example,i nm olecular recognition/sensing, [15] molecular machines, [16] switchable catalysis [17] and artificial ribosome mimics. [18] To test their potential for rac-LA ROP, rotaxanes 1, 2 and 3 were targeted ( Figure 1, SI for synthesis). Each comprises ac rown ether macrocycle and an axle,t he latter containing both an ammonium and at hiourea or triazole group.I nt his form, the macrocycle resides over the protonated ammonium group secured by strong intramolecular hydrogen bonding (Figure S3, S10, S14). [19] After deprotonation, the amine interacts weakly with the macrocycle allowing its free movement along the axle.T his protonation-dependent dynamic behavior also highlights their potential to be switched "on" for catalysis.
Rotaxanes 1-3 were all inactive for rac-LA ROP, even with added benzyl alcohol, consistent with previous reports that thioureas alone cannot activate lactide to attack by alcohols. [4a] Neutral rotaxanes,d eprotonated in situ by addition of base,w ere active in presence of benzyl alcohol (Table 1, entries 1, 7a nd 8). Therates differ, with 1 reaching 80 %after 4d,whilst 2 and 3 achieved high conversions within 12 h. All catalysts showed high polymerization control and produced PLA with predictable molar masses (M n ), linear evolution of M n with conversion, and narrow dispersity ( < 1.20) (Figure 2A,S 31-S34). End-group analysis,u sing ESI mass spectrometry,s howed as ingle set of peaks corresponding to benzyl ester end-capped chains ( Figure S28-S30). Tr ansesterification side reactions were not observed for 1 and 3,a se videnced by the 144 peak separations,w hilst peak separations of 72 for 2,indicates transesterification. It is noted that neutral rotaxane 2 is the first example of atriazolecontaining organocatalyst for lactide ROP.
Three components are required for control:r otaxane, base and alcohol. Polymerizations conducted without rotaxane were rapid but uncontrolled yielding atactic PLA (Table 1, entry 2). Reactions without alcohol were inactive (Table 1, entry 4). Under the appropriate conditions, 1 produced highly isotactic PLA (P i = 0.81;F igure 2B,S 35). Replacing the thiourea group of 1 with the weaker hydrogen bond-donating triazole in 2 significantly reduces stereocontrol (P i = 0.66;F igure S39). Thiourea accessibility also appears to be important, as 3 shows lower stereocontrol (P i = 0.73, Figure S40). In all cases,a nalysis of defect tetrad resonances,inthe methine region of the 1 H{ 1 H} NMR spectra, suggests al ikely dominant CEC mechanism (see Figure S35, SI). [20,21] Despite as mall extent of epimerization occurring during the reaction ( Figure S27), the highest stereocontrol, achieved by 1,m atches values for chiral organocatalysts, [8] urea/alkoxide systems [22] or organometallic catalysts. [23] Perhaps more interesting is the finding that the rotaxane structures,a nd dynamism, are inherent to stereocontrol. Under identical conditions,t he macrocycle alone yields only atactic PLA, while the protonated thiourea axle is inactive (SI). Thep ossibility that isoselectivity results from only thiourea steric congestion can also be ruled out because the equivalent acycliccatalyst 4 shows lower selectivity (P i = 0.69, Figure S41). Furthermore,i ncreased thiourea steric hindrance,i n3,r educes stereoselectivity.T hese observations point towards an alternative stereocontrol.
TheNMR spectra of neutral rotaxanes were compared to understand catalyst structure in more detail. Addition of equimolar base to 1,results in upfield shifts to resonances H 3 and H 4 ,w hich are immediately adjacent to the ammonium  [24] [d] Determined from the 1 H{ 1 H} NMR by integration of normalized methine tetrads and Bernoullians tatistics( Figure S35-S41, SI). [25]  group (Figure 3). Such shifts indicate formation of an eutral amine group.A tt he same time,t hiourea resonances H a and H b shift downfield, consistent with enhanced thiourea-macrocycle interactions.Nosignificant changes were observed after the addition of benzyl alcohol ( Figure S44). Significant broadening of all other resonances indicates the macrocycle is moving along the axle slower than the NMR timescale. Adding base to 2,a lso deprotonates the ammonium group ( Figure S45). Ther emaining signals are sharp suggesting the rate of macrocycle movement along the axle is faster than for 1.T he ROESY NMR spectrum of (neutral) 2 provides unequivocal evidence of macrocycle movement between amine and triazole sites,e vident from the numerous crosspeaks between the macrocycle protons and those arising from the aromatic stopper units on both ends of the axle (Figure S49). Intense ROESY signals correlate with the macrocycle being preferentially located at the amine site,likely due to its greater hydrogen bonding acidity compared to the triazole.In contrast, neutral 1 would be expected to show favorable interactions with both thiourea and amine sites.I ts ROESY NMR reveals stronger through-space interactions between the macrocycle and the thiourea ( Figure S47). Accordingly, stronger macrocycle-thiourea hydrogen-bonding may slow macrocycle translation along the axle and broaden the NMR spectrum. The 1 HNMR spectrum for neutral 3 also indicates the axle amine group was deprotonated ( Figure S45). ROESY NMR could not confirm the macrocycle site occupancyd ue to the complex, overlapping aromatic resonances ( Figure S50). Nonetheless,t he bulky stopper group may be expected to hinder thiourea-macrocycle interactions and favour macrocycle occupation of the amine site.
Control experiments support the neutral rotaxane being the true catalyst (Scheme 1A). Neither potassium hexafluorophosphate,t he amine by-product (N(SiMe 3 ) 2 H) nor benzyl alcohol initiate polymerization, either individually or in combinations (Section S3.2, SI). Thef ree axles of 1 and 2 were also inactive.Astrong base is necessary for rotaxane deprotonation:r eplacing KN(SiMe 3 ) 2 with DABCO (a weaker base) results only in limited ROP( 7% after 120 hours). To rule out formation of rotaxane-potassium complexes,the 1 HNMR spectra of the potassium and sodium salts of 1 were compared. Both spectra showed identical chemical shifts for all resonances ( Figure S51). As K + /Na + coordination would be expected to give different chemical shifts,the active species are unlikely to be Group 1complexes. Overall, the data indicate that the catalyst is an eutral rotaxane and DOSY NMR indicates it is monomeric under the conditions of catalysis ( Figure S48).
To rationalize the differing catalytic performances,apolymerization pathway is proposed inspired by the two-component thiourea/amine organocatalysis literature (Scheme 1 B). [4a,b] Neutral 1/2 feature thiourea/triazole groups which activate the lactide monomer and amine groups to activate benzyl alcohol. This mode of initiator/ monomer activation is only possible if the macrocycle is free to move along the axle. Polymerization is initiated, and propagated, by an activated monomer mechanism. Importantly,f or these rotaxanes,t he basicity of the axle secondary amine is considerably enhanced by interaction with the crown ether macrocycle component compared to the free neutral axle,asthe stability of the crown ether-R 2 NH 2 + conjugate acid upon protonation strongly favours its formation thermodynamically. [19c, 26] Accordingly, the activated benzyl alcohol attacks the coordinated lactide forming an a-hydroxyl-w-ester which is subsequently (re)activated and attacks another lactide.I nl ine with this,t he relative rates correlate with rotaxane amine basicity,a nd hence capacity for alcohol activation. Because the rotaxanes also feature other macrocycle coordination sites,w hen these sites are strongly coordinating the amine basicity is reduced. [27] For 2 and 3,m acrocycles occupy the amine site, increasing its basicity and accelerating rates (Scheme 1B).
Only 1 shows high isoselectivity and operates by an unusual chain end control mechanism. To investigate further, the rotaxane-lactide binding constants were determined by 1 HNMR titrations. 1 shows ah igher binding constant (K a % 400 m À1 )t han 2 (K a < 1m À1 )( Section S4.2, SI). [28] From the aforementioned ROESY studies,i ta lso shows significantly slower macrocycle shuttling. It is tentatively proposed that both the strong binding and slow shuttling allow for the preferential orientation of the growing polymer chain and lactide resulting in the isoselectivity.
In conclusion, these are the first rotaxane catalysts for lactide ROPand they show isoselectivity.The polymerization selectivity correlates with macrocycle-axle translocation rates and with sterically-accessible,s trong monomer coordination sites.T he conformational dynamism appears to be responsible for both rates and stereoselectivity and provides an ew means by which to control tacticity.T his new type of stereocontrol is expected to be applicable to transesterifications,t ransamidations and other ring-opening polymerizations of epoxides,cyclic carbonates and lactones.