Simple Zn(ii) complexes for the production and degradation of polyesters

Nine new complexes based on thioether appended iminophenolate (ONS) ligands have been prepared and fully characterized in solution by NMR spectroscopy. Solid-state structures were also obtained for seven complexes. In solution, all complexes were monomeric. The complexes were highly active for the polymerization of purified rac-lactide ([M] : [Zn] : [BnOH] = 10 000 : 1 : 30 at 180 °C) reaching TOF values up to 250 000 h−1. The kinetics of the polymerization have been probed by in situ Raman spectroscopy. The rate of reaction was dramatically reduced using technical grade rac-lactide with increased initiator loading. To move towards a circular economy, it is vital that catalysts are developed to facilitate chemical recycling of commodity and emerging polymeric materials. In this vein, the complexes have been assessed for their ability to break down poly(lactic acid) and poly(ethylene terephthalate). The results from both the polymerization and degradation reactions are discussed in terms of ligand functionality.


Introduction
The widespread use of polymers has revolutionised almost every aspect of modern life. However, the proliferation of these materials has come with serious ecological implications. [1][2][3] Reducing our reliance on the fossil fuel feedstocks which are necessary for producing nearly all commercial plastics is imperative due to the catastrophic environmental impact of extracting and processing these nite resources. 4 The use of renewable feedstocks to create new polymers or to replace hydrocarbon components of existing products is therefore one of the most important scientic challenges of the 21 st century. [5][6][7][8] Poly(lactic acid) (PLA) is an important bio-renewable polymer that is sourced from starch-rich materials and is amenable to enzymatic degradation or chemical recycling. 9,10 It is established as a packaging material and has also found use in the agricultural industry and the bio-medical industry where biocompatible materials are required. 9,11-14 PLA is typically produced through the ring-opening polymerization (ROP) of lactide initiated by a metal complex. Industrially, Sn(Oct) 2 is used but there are toxicity issues associated with tin residues and so the current focus is to achieve industrially relevant activity with environmentally benign metal initiators. Poly(Llactide) dominates the PLA market due to the ease of L-lactic acid biosynthesis. However, the stereoselective ROP of rac-lactide can improve the material properties of the polymer and thus it is oen studied in academic research. A diverse range of metals has been applied to lactide ROP including Mg(II), 15,16 group IV, [17][18][19][20][21][22][23][24] Fe(II/III), [25][26][27][28][29][30][31][32] Al(III), [33][34][35][36][37][38][39][40][41][42][43][44][45][46][47][48][49] and In(III). [50][51][52] Initiators based on zinc(II) have consistently shown high activity under solvent-free conditions and, in some cases, stereocontrol is observed. Coates and co-workers published a bdiiminate zinc complex that produced heterotactic PLA (P r ¼ 0.94) at ambient conditions. 53 The most isoselective zinc initiator was reported by Ma and co-workers (P r ¼ 0.08) using an aminophenolate complex in toluene at À20 C. 54 However, the low temperature required for high selectivity reduced the activity (TOF ¼ 117 h À1 ) and high stereoselectivity was not maintained under solvent-free conditions (P r ¼ 0. 19-0.20). A dizinc bis(imino)diphenylamido initiator reported by Williams and co-workers remains the most active solution-based zinc initiator for lactide ROP (TOF ¼ 60 000 h À1 ). Using high temperature (130-180 C), solvent-free, industrial conditions, Jones et al. reported simple ethylenediamine monophenolate complexes that polymerised L-lactide with TOF values in excess of 100 000 h À1 . 55 Further work by the same group demonstrated the introduction of a propyl linker which increased the activity at 180 C with L-lactide giving high conversion aer just 1 minute at low initiator loading (10 000 : 1 : 33). 56 The most active zinc initiator to date for lactide ROP was reported by Herres-Pawlis and co-workers using a bisguanidine complex which signicantly outperformed Sn(Oct) 2 and gave highly crystalline PLLA whilst also being active for technical grade raclactide. 57 Further work using similar complexes from Pellecchia and co-workers further demonstrated their effectiveness at industrial conditions whilst also achieving lactide copolymerisation with 3-caprolactone. 58 Most oil-based commodity plastics do not readily decompose in the environment and end up in landll or accumulate in the oceans where plastic microparticles are devastating to marine ecosystems. 59 Mechanical recycling can be effective in some cases but leads to material downcycling over time and this limits the number of potential cycles. 60 Products must ultimately be repurposed and so this is a short-term solution to retaining value in the polymer economy. Conversely, chemical recycling to monomer or value-added products offers a much more long-term solution to value retention either through depolymerisation to virgin monomer or degradation to valueadded products. 61 Both offer an intrinsic economic incentive to industry either through a potentially innite number of monomer to polymer cycles or through the upgrading of waste polymer to value-added and industrially relevant chemicals in the case of degradation. 62 This approach has huge potential to offer a more sustainable approach to plastic usage and is crucial to attaining a circular economy for polymers.
Simple metal salts have been reported for PLA methanolysis and can give good methyl lactate yields (Y Me-LA ¼ 87% for FeCl 3 ) but require high temperatures (130-180 C). [63][64][65] The rst example of a zinc complex for PLA methanolysis was reported by Avilés et al. with a dizinc NHC complex capable of producing methyl lactate over 24 hours at ambient temperature. 66 More recently, Jones and co-workers have reported several homoleptic zinc monophenolate complexes that are active for the degradation of PLA into alkyl lactates at mild conditions. 56,67-71 An imine-monophenolate complex with a propylene diamine linker was able to achieve a high yield of methyl lactate (Y Me-La ¼ 89%) aer just 30 minutes at 50 C. 56 This far exceeds the activity shown by the ethylenediamine-based analogues which required several hours to reach high conversion at comparable temperatures. 70 The most effective propylenediamine catalyst was tested for the degradation of post-consumer PLA products into ethyl lactate. 67 Although activity was lower than with the clear PLA cup used for the methanolysis experiments, good yields of ethyl lactate were observed and the variation in activity was attributed to additives and the ease of dissolution of each product. More recent work has demonstrated an ethylene diamine-based ligand with a catam (N-H) moiety that far exceeds the activity of the direct imine-based counterpart, achieving 85% Me-LA yields aer 30 minutes at 50 C. 70,72 The aforementioned ligand is a building block of the tetradentate catalen system, zinc complexes of which were also active for PLA degradation to methyl lactate at moderate yields aer 8 hours at 80 C. 71 Recent examples have also achieved PLA alcoholysis at ambient conditions and with no solvent, in some cases demonstrating full PLA conversion aer 1 hour. 73,74 Organocatalysts, such as DMAP, 75 TBD, 76 TMC 77 and ionic liquids, [78][79][80] have also been reported but typically require high catalyst loading and can be limited by issues relating to corrosivity, toxicity and high cost.
Herein, we report a series of nine Schiff base ligands each bearing a thioether motif where the phenolate substituents, thioether substituent and linking group have been varied systematically (Scheme 1). The homoleptic zinc(II) complexes were applied to lactide ROP under solvent-free conditions as well as the methanolysis of commercial PLLA and the glycolysis of PET.

Results and discussion
Synthesis and characterisation of complexes Ligands 1-9H were prepared via an imine condensation reaction between the relevant amine and salicylaldehyde derivative, with purity and structure conrmed through 1 H NMR, 13  Solid state structures were obtained for seven complexes of which ve, Zn(1, 2, 4, 5, 7) 2 , were shown to be four-coordinate with no Zn-S bonding (Table 1). Of all complexes, Zn(1) 2 (R 1 ¼ t Bu, R 2 ¼ CH 3 , X ¼ -C 2 H 4 -) was closest to tetrahedral geometry ðs  Fig. 1), deviates further from the tetrahedral ideal ðs  81 and thus it is reasonable to assume that there is no bonding interaction. The geometry of Zn(4) 2 deviates further from tetrahedral ðs 0 4 ¼ 0:64Þ; probably due to the rigid phenylene linker constraining chelation. This is also true for Zn(5) 2 where two crystallographic structures were present in the sample, both with a relatively low geometry index {Zn(5) 2 A, s The average distance between zinc and sulphur atoms is considerably shorter than with the ethylene-bridged analogues, with values below the sum of van der Waals radii {Zn(4) 2 , 3.1 A; Zn(5) 2 A, 3.1 A; Zn(5) 2 B, 3.2 A} suggesting that there could be a degree of bonding interaction. Again, this is likely a consequence of the more rigid structure forcing the sulphur to approach the metal centre more closely. Modication of the thioether substituent to CF 3 {Zn(7) 2 : Fig. 1) caused the geometry to become more tetrahedral ðs 0 4 ¼ 0:68Þ; and caused the average Zn-S distance to increase to 3.4 A. This reects a combination of the modication in electron density at the potential sulphur donor atom and the steric difference between the substituents. The solid-state structure of Zn(6) 2 indicates a ve-coordinate geometry with one sulphur atom formally bonded to zinc (Fig. 1). The geometry index for a ve-coordinate complex, s 5 , was calculated as 0.02, indicating an almost perfectly square pyramidal geometry. Considering the similarity of the Zn-S bond lengths {Zn(1)-S(1) ¼ 2.8032(18) A, Zn(1)-S(2) ¼ 2.9 A}, an octahedral geometry is a more feasible description of the structure. This is supported by Bader's quantum theory of atoms in molecules (QTAIM) analysis, in which a bond critical point (BCP) was located between Zn(1) and S(2) with r ¼ 0.145 e A À3 (cf. BCP from Zn(1) to S(1) has r ¼ 0.185 e A À3 ). Assuming an octahedral geometry, the ligands wrap in a mer-mer orientation with the thioether and phenoxy donors of the respective ligands adopting a pseudo axial position The complex based on ligand 9H gave a dimeric structure of the form Zn 2 (9) 4 with two ve coordinate zinc centres (Fig. 2). Zn(1) tended slightly towards trigonal bipyramidal (s 5 ¼ 0.54) and Zn(2) was much closer to square pyramidal (s 5 ¼ 0.33). There are a range of Zn-S interatomic distances with the closest (Zn-S ¼ 3.4 A) situated trans to the apical Zn(2)-N(4) bond. As with Zn(6) 2 , some degree of Zn-S interaction could be inuencing the geometry at Zn(2). DOSY NMR was used to assess whether the dimeric structure is maintained in solution through comparison with Zn(4) 2 and Zn(7) 2 . One species was observed for each complex. Comparison between the diffusion constant of the complex and the solvent (Table SI2 †) was used to calculate the hydrodynamic radii of the complexes. All were sufficiently similar to indicate a consistent structure {rZn (4) Based on an approximate measurement of the complex diameter from the solid-state structures, it is likely that the monomeric homoleptic complex is present exclusively in solution.
keeping with other zinc monophenolate complexes under these conditions. 56,69,71 When the lactide to initiator ratio was increased, ([LA] : [I] : [BnOH] ¼ 3000 : 1 : 10) the activity differences between the complexes became more apparent (Table 3). Ethylene bridge complexes, Zn(1-3) 2 , took much longer to reach high conversion (t ¼ 30-60 min) than the phenylene bridge equivalents. The chlorinated analogues were the least active in each series whilst also showing a slight increase in heterotacticity compared to the other initiators. This reects an alteration of the coordination sphere, facilitated by the chloride substituents, that increases steric crowding of the zinc centre. The octahedral structure of Zn(6) 2 is evidence of this effect, which could also apply to Zn(3) 2 . There is no signicant difference in activity between the -CH 3 and -CF 3 substituted complexes, except in the case of Zn(9) 2 , where the CF 3 group appears to partially negate the drop in activity that is associated with the chloride substituents. Phenylene bridged complexes, where R 1 ¼ H or t Bu [Zn (4,5,7,8)], were clearly the most active initiators converting sufficient lactide aer less than 5 minutes to stop stirring. Reasonable molecular weight control was maintained throughout and there was a narrowing of dispersity, (Đ ¼ 1.08-1.48) that is oen associated with increasing the monomer to initiator ratio. The differences in activity between the complexes show some correlation with the measured Zn-S interatomic distance (Table 1) in a volcano-style relationship (Fig. SI47 †). Ethylene bridge complexes, Zn(1, 2) 2 , where the Zn-S distance is large (4.4-4.5 A), and there is presumably no interaction, are comparatively slow. Conversely, the direct phenylene-bridged analogues {Zn(4, 5, 7) 2 } have a smaller average Zn-S distance (3.1-3.4 A) and are the most active for lactide polymerisation. Zn(9) 2 has an average Zn-S distance of 3.9 A, intermediate to the two aforementioned groups of complexes, and this is reected in the activity. The zinc centre in Zn(6) 2 is considerably closer to the two sulphur atoms (Zn-S ¼ 2.8-2.9 A). The activity is similar to the ethylene-bridged    complexes. This trend can be rationalised by considering the coordination of lactide to the saturated zinc centre. When the Zn-S distance is small, the complexes tend to distort further from tetrahedral geometry (e.g. Zn(1) 2 : s 0 4 ¼ 0:84, TOF ¼ 2080 h À1 ; Zn(1) 2 : s 0 4 ¼ 0:48=0:55, TOF ¼ 124 200 h À1 ). The displacement of weakly bound sulphur by lactide is easier than coordination to an undistorted tetrahedral centre. This explains the increase in activity up to a Zn-S distance of around 3 A, aer which the sulphur becomes more difficult to displace and the activity drops sharply.
The polymer produced by Zn(3) 2 at this loading was suitable for MALDI-ToF analysis (Fig. SI43 †). The repeat unit of the main series is 144 g mol À1 , and the spectrum is centred around 4457.9 g mol À1 ; similar to that measured by GPC and to the theoretical molecular weight (M n GPC ¼ 4750 g mol À1 , M n calc. ¼ 5850 g mol À1 ). At low molecular weight there is evidence of ionisation by incidental potassium and a small amount of transesteried PLA.
The most active initiators, Zn(4, 5, 7, 8, 9) 2 , were tested for rac-lactide polymerisation at very low catalyst loading  (Table 4). 82 Under these conditions, Zn (5) and Zn(7) 2 (R 1 ¼ t Bu, R 2 ¼ CF 3 , X ¼ C 6 H 4 ) were clearly the most active initiators achieving 68% and 51% conversion in 3 and 9 minutes respectively. These were selected for further kinetic analysis. With the exception of Zn(5) 2 , conversion was limited to around 50%, suggesting some initiator decomposition is taking place. The molecular weight control was maintained at this loading and the dispersities remained consistent (Đ ¼ 1.20-1.44). To further probe the effectiveness of these complexes at industrial conditions, the most active complexes, Zn(4, 5, 7, 8, 9) 2 , were tested at 180 C ( Table 5). All the initiators tested were sufficiently robust to show activity at this temperature and were consistently more active than at 130 C (t ¼ 2-12 min, conversion ¼ 56-83%), due both to the increase in temperature and the decreased viscosity allowing for better mixing at higher conversions. As might be expected at elevated temperatures, the dispersities were broader (Đ ¼ 1.45-1.68) although reasonable molecular weight control was maintained. Zn(5) 2 and Zn (7) 2 were also tested for the polymerisation of L-lactide, the most commonly used industrial monomer. There was a small reduction in activity with L-lactide for both initiators, although the activity was still high. The reaction was very well controlled with comparatively narrow dispersities (Zn(5) 2 , Đ ¼ 1.16; Zn(7) 2 , Đ ¼ 1.15) and excellent molecular weight control (Zn(5) 2 , M n ¼ 32 050 g mol À1 , M n (theo.) ¼ 35 750 g mol À1 ; Zn(7) 2 , M n ¼ 28 050 g mol À1 , M n (theo.) ¼ 27 050 g mol À1 ). DSC analysis showed that the melting point of PLLA produced by Zn(5) 2 and Zn(7) 2 was 162 C and 167 C respectively (Fig. SI45 and SI46 †). This suggests that there is limited epimerisation and the presence of a single methine peak in the 1 H{ 1 H} NMR spectrum supports this conclusion. Despite the yellow colouring of the initiators, white PLA was produced at the highest ratios employed and this is desirable from a commercial standpoint (Fig. SI44 †). The performance of Zn(5) 2 and Zn (7)  [BnOH] ¼ 10 000 : 1 : 33) taking 3 minutes to achieve a higher conversion of 94%. 55 In order to probe the polymerisation mechanism and the role of the co-initiator, a stoichiometric reaction was carried out. A 1 : 1 solution of rac-lactide and Zn(7) 2 in CDCl 3 was heated to 50 C for 90 min and monitored by NMR spectroscopy (Fig. SI48 †). Resonances similar to those of opened lactide increased signicantly upon addition of BnOH to 60% in ten minutes at room temperature and 94% conversion aer a further 90 min at 50 C. This shows that a co-initiator is required for high activity but that a degree of activity might be observed in its absence. This could be explained through initiation by impurities in rac-lactide or through a coordinationinsertion mechanism facilitated by a dissociated ligand as observed by McKeown et al. with zinc ONN complexes. 55 Some evidence of ligand dissociation was observed in the methyl region of the 1 H NMR spectrum where two minor resonances corresponding to t-butyl groups steadily increased throughout the reaction (Fig. SI49 †). Furthermore, a new signal was present in the 19 F NMR spectrum at the end of the reaction (Fig. SI50 †). DOSY NMR analysis of the nal reaction mixture showed that the ring-opened lactide had a distinctly different diffusion constant to the metal complex (Fig. SI51 †). This suggests an activated monomer mechanism wherein the components of the reaction are never fully bonded to the metal and the co-initiator is exogenous. Therefore, the best explanation for the conversion observed prior to BnOH addition is through initiation by impurities in the monomer. The observation of additional resonances in the 1 H and 19 F NMR spectra is evidence of complex decomposition and could account for the limited conversion achieved by these systems.

Polyester degradation
Zn(1-9) 2 were investigated for PLA methanolysis at 80 C in THF. The product, Me-LA, can be converted to lactide for the production of virgin PLA and is also an emerging green solvent. 62 Furthermore, Me-LA can replace lactic acid in many transformations so has potential as a platform chemical. 85 Waste polymer from a commercial source (0.25 g, PLLA cup, M n ¼ 45 510 g mol À1 ) and catalyst (20 mg, 8 wt%) were added in a glovebox and subsequently dissolved in THF. When the polymer was fully solubilised, methanol was added to initiate the degradation reaction. This reaction has been shown to proceed via a two-step mechanism with PLA initially broken down into oligomers before being converted to the product. 70 Analysis of the methine region (ca. d ¼ 4.2-5.2 ppm) of the 1 H NMR spectrum gives three key parameters: internal methine conversion (X int ), Me-LA yield (Y Me-LA ) and the selectivity to Me-LA (S Me-LA ).
All complexes were active for the methanolysis of PLA (Table  6). Of the ethylene bridge complexes, Zn(1-3) 2 , the highest activity was recorded for Zn(2) 2 (X int ¼ 87%) and this complex produced the most Me-LA of all the catalysts that were tested (S Me-LA ¼ 72%, Y Me-LA ¼ 63%). This contrasts with the polymerisation studies where Zn(2) 2 signicantly underperformed most of the phenylene-bridged analogues. This can be rationalised by considering the equilibrium between oligomer and methyl lactate where the backward reaction (k À2 ) will be slower with a less effective polymerisation catalyst thus driving the reaction towards the product. The most active catalysts were Zn(4) 2 (R 1 ¼ t Bu, R 2 ¼ CH 3 , X ¼ C 6 H 4 ) and Zn(4) 2 (R 1 ¼ H, R 2 ¼ CH 3 , X ¼ C 6 H 4 ) which gave internal methine conversions of 94% and 96% respectively. The latter also gave high selectivity and yield of Me-LA (Y Me-LA ¼ 59%, S Me-LA ¼ 62%). As observed during polymerisation, Zn(6) 2 was the slowest in this series (R 1 ¼ Cl, R 2 ¼ CH 3 , X ¼ C 6 H 4 ) presumably resulting from steric crowding from the pseudo-octahedral structure. Interestingly, the introduction of a CF 3 group at the R 2 position resulted in a sharp decline in activity, particularly for the non-halogenated complexes Zn(7-8) 2 , both of which converted 39% of internal methine aer 8 hours with correspondingly low Me-LA yields. However, Zn(9) 2 (R 1 ¼ Cl, R 2 ¼ CF 3 , X ¼ C 6 H 4 ) achieved good activity and selectivity (X int ¼ 81%, S Me-LA ¼ 50%, Y Me-LA ¼ 41%) under the same conditions.
The most active catalysts [Zn (2,4,5,9) 2 were tested at 50 C for 18 hours (Table 7). There was a reduction in activity and selectivity compared with the 80 C reactions and this is consistent with literature examples. 69,86 Performance particularly decreased for Zn(4) 2 (X int ¼ 57%, S Me-LA ¼ 21%, Y Me-LA ¼ Table 5 Solvent-free polymerisation of rac-lactide at 180 C using Zn (4,5,7,8,9)   12%) and Zn (5) There was an indication that Zn(4-5) 2 are sensitive to residual moisture in the THF and this could lead to deactivation over extended reaction times. Due to the lack of labile alkyl groups around the metal, it is likely that this hydrolytic sensitivity comes from the active species or a reaction intermediate. Zn (2) 2 remained the most selective catalyst and gave reasonable conversion of internal methine (X int ¼ 77%). Kinetic analysis of Zn(2, 4, 5, 9) 2 was attempted through taking aliquots every hour for 4-6 hours with a nal point measured aer 8 hours. Based on previous literature, pseudo rst-order kinetics of PLA consumption were assumed. 70 Zn(4) 2 and Zn(5) 2 were not amenable to sampling and the reactivity was quenched.
To expand the scope of polyester degradation, the most active PLA degradation catalysts were applied to the glycolysis of PET to bis(2-hydroxyethyl) terephthalate (BHET); a useful product that can be polymerised to make virgin PET or can be used to make unsaturated polymer resins. 87,88 PET is one of the most prevalent commodity plastics accounting for 9% of global plastic demand in 2015. 89 Although widely mechanically recycled, it is still a major contributor to plastic pollution and is not degradable in the environment. 90 Zn [2,4,5,9] 2 were all active for the glycolysis of PET, with full dissolution observed aer 1.5-4 hours for a standard carbonated drinks bottle (Table 8). Full degradation of PET was assumed at this point and BHET was isolated through recrystallisation from water. Isolated yields between 42% and 55% were observed, and this is consistent with previous literature examples using zinc catalysts. 71,72 The most active catalyst {Zn(2) 2 : t ¼ 1.5 h, Y BHET ¼ 0.16 g (48%) was tested with thin-lm PET; a proxy for manufacturing waste. As expected, the reaction time was signicantly reduced (t ¼ 0.75 h) and a percentage yield of 51% was attained, again in keeping with previous results by Payne et al. 71,72 In the reaction with Zn(2) 2 , colourless BHET was obtained aer recrystallisation and this is important when considering the quality of PET if it were repolymerized (Fig. SI62 †). 91 Conclusions A series of nine ONS zinc complexes were prepared and characterized through NMR spectroscopy and single-crystal X-ray diffraction. The rapid polymerization of rac-lactide was demonstrated under industrially relevant, solvent-free conditions. Up to 83% conversion was achieved in 2 minutes at 180 C with TOF values up to 250 000 h À1 . Crystalline PLLA was also produced rapidly with minimal epimerization. Good control of molecular weight and dispersity (Đ ¼ 1.08-1.88) was maintained throughout. Raman kinetic experiments showed that the activity is signicantly reduced when using technical grade raclactide even at relatively high catalyst loading. For the optimised preparation with Zn(7) 2 , an apparent rate constant k app ¼ 2.23 Â 10 À3 s À1 was measured. All complexes were active for the degradation of PLA to methyl lactate under mild conditions. Two waste sources of PET were also degraded to colourless BHET with full PET consumption observed aer 0.75-4 hours (Scheme 1).

Experimental section
All experimental details are provided in the ESI † with the original data.

Conflicts of interest
There are no conicts to declare. Table 7 Degradation of PLLA cup to Me-LA using Zn (2,4,5,9)