Sustainable Synthesis and Precise Characterisation of Bio-based Star Polycaprolactone Synthesised with a Metal Catalyst and with Lipase

Bio-based building blocks and sustainable synthesis pathways were used to synthesise star-shaped polymers composed of a D -sorbitol core and polycaprolactone arms by ring opening polymerisation (ROP). The use of volatile organic solvents was avoided and less energy intense reaction conditions were achieved by performing the ROP in the bulk or in a green solvent, supercritical CO 2 (scCO 2 ). Two catalysts were tested: conventional tin(II) 2-ethylhexanoate (Sn(Oct) 2 ) which is a Food and Drug Administration (FDA) approved metal catalyst and an enzyme, Novozym 435 (Lipase B from Candida Antarctica immobilised on a solid support). The influence of the reaction medium and of the nature of the catalyst on the molecular weight, the dispersity and the architecture of the PCL stars was investigated. The star polymers were characterised by 1 H and 31 P nuclear magnetic resonance ( 1 H and 31 P NMR) spectroscopy, size exclusion chromatography - multi-angle light scattering (SEC-MALS) and matrix-assisted laser desorption and ionisation-time of flight (MALDI-TOF) mass spectrometry. The use of scCO 2 enabled the reduction of the reaction temperature of Sn(Oct) 2 catalysed star D -sorbitol-polycaprolactone polymerisations from 140 to 95 °C. In addition, star polymers were successfully synthesised by enzyme catalysis in the bulk or in scCO 2 at 60 °C; lower temperatures that could provide significant energy savings on a commercial scale. The catalyst was shown to have a pronounced influence on the architecture of the PCL stars. Regular star polymers were obtained in the presence of Sn(Oct) 2 whereas Novozym 435 gave access to miktoarm-type star PCL. Finally, the influence of the number and length of the arms on the thermal properties of the star polymers was investigated by differential scanning calorimetry (DSC).


Introduction
The accumulation of non-biodegradable plastics on earth has resulted in an increasing need for biodegradable polymers, ideally produced by green technologies. In addition, the simultaneously occurring depletion of fossil resources has promoted the utilisation of naturally occurring building blocks. 1 At the same time, there has been a tremendous development in the synthesis of complex macromolecular architectures in order to design polymers tailored to specific functions and applications. In particular, star polymers have attracted attention, because of their unique rheological properties, lower hydrodynamic volume than their linear analogues, and their very interesting thermal, topological, and biological properties that are inaccessible from their linear counterparts. 2,3,4,5 Star polymers have at least three arms of often similar lengths, fused on a central core and based on the nature and uniformity of their arms, they are classified into regular or miktoarm types. 3,6 The special properties of such star-like structures result from the higher number of functional end groups and their highly branched nature, which causes, for example, a lower intrinsic viscosity, by reducing the amount of chain entanglements compared to linear polymers of similar molecular weight, and can lead to a reduced melting point. 2,3,47,8 A prominent example of star polymers for commercial applications are Lubrizol's Asteric TM viscosity modifiers. In this case, star poly(alkyl methacrylates) with 9-13 arms are used as additives for lubricants to decrease viscosity at lower temperatures while at the same time ensuring a higher viscosity index, i.e. reducing the variation in viscosity with changing temperature. 2 Star polymers can be synthesised by an arm first approach, in which arms are synthesised and then attached to the core, or by a core first approach. In the core first approach, the arms are polymerised from multifunctional initiators using, for example, ring opening polymerisation (ROP). Poly(ε-caprolactone) (PCL) is a particularly interesting polymer obtained by ROP since it is inexpensive, biocompatible and possesses a controlled degradability. In the future, the monomer ε-caprolactone (ε-CL) could potentially be sourced from renewable resources. 9,10,11 For instance, ε-CL could be obtained from lignocellulosic biomass derived 5hydroxymethylfurfural. 12,13 Additionally, ε-CL is an intermediate in the oxidation of cyclohexanol to adipic acid by an Acinetobacter sp. strain SE19 enzyme, resulting in a potential for biotechnological sourcing of ε-CL. 14 Typically, star PCL are synthesised from polyhydroxy alcohols (such as glycerol, 15,16 trimethylolpropane, 17,18,19 pentaerythritol, 15,16,18,19 di-trimethylolpropane, 17 D-sorbitol, 20 dipentaerythritol, 17 and maltitol 17 ) as initiators, which constitute the core of the stars. Multifunctional polymeric macroinitiators such as star poly(ethylene glycol) 21 or hyperbranched polymers such as poly(2-hydroxyethyl methacrylate) 22 have also been employed to form star block copolymers with PCL. Sorbitol is of particular interest as a core for star PCL, since it is an inexpensive, highly valuable polyol commercially derived from renewable feedstocks such as corn-starch, cellulose, cassava, wheat, etc. 23,24 Tin(II) 2-ethylhexanoate (Sn(Oct) 2 ) is commonly used as a catalyst 20,25 for the ROP of ε-CL.
Although Sn(Oct) 2 is a well-known Food and Drug Administration (FDA) approved metal catalyst, there are increasing concerns regarding the presence of toxic metal catalysts in the final products. This is of particular importance for polymers in skin care or cosmetic products. In previous work, we have shown that extraction using supercritical CO 2 (scCO 2 ) can be used to reduce the amount of catalyst residue in polymer products. 26 However, an alternative may be to avoid the use of a toxic metal catalyst and shift to more environmentally friendly alternatives such as enzymes. 27,28 Lipase is the most important class of biocatalyst for ROP because of its abundance in nature, high specificity, regio-and enantioselectivity and low environmental impact. 27 Uppenberg and co-workers discovered in 1994 that this selectivity was due to the 'open' conformation of the catalytic triad Ser-His-Asp in the active site of CaLB, which restricts the accessibility of the active site for certain substrates. 29 In particular, Novozym 435 (Lipase B from Candida Antarctica (CaLB) immobilised on a macroporous acrylic resin) has shown extraordinary reactivity towards a wide range of cyclic monomers, high catalytic activity, good solvent resistance and has been broadly studied as a biocatalyst for polyester synthesis. 28 Enzymatic ROP (eROP) of ε-CL by Novozym 435 has been studied extensively in the bulk and in various organic solvents to give linear PCL. 30,31 However, an obstacle in enzyme catalysis is the instability of the catalysts. At temperatures higher than 100 °C, which is typically employed in ROP of ε-CL in the bulk, denaturation of the protein structure occurs. 32 The optimum temperature for CaLB catalysis for enzymatic polyesters synthesis is thus between 60 and 70 °C. 33,34 Another challenge is the sensitivity of enzymes towards certain solvents, such as polar solvents, which interfere with the hydrogen bonding in the protein structure, thereby affecting the catalytic activity of the enzyme. 35 For example, Novozym 435 displays relatively high selectivity towards primary alcohols compared to secondary alcohols.
However, the presence of polar solvent affects the selectivity of Novozym 435. For instance, Xue et al. achieved three-arm star PCL polymers using glycerol as the initiator in presence of 1,4-dioxane at 70 °C confirming initiation from a secondary hydroxy group. 36 Other factors such as catalyst loading and the immobilisation process on the support can also influence the catalytic activity and regioselectivity. 37,38,39 Regarding the enantioselectivity of Novozym 435, Gross and co-workers have investigated the incorporation of alditols, such as D-sorbitol, D-galactitol, and D-mannitol into copolyesters with adipic acid and 1,8-octanediol by polycondensation and observed the highest molecular weights as well as the highest degree of branching for the polyesters with D-mannitol as the copolymer, indicating a preference of Novozym 435 towards all R-configuration of this alditol. Accordingly, incorporation of D-sorbitol (SRRR) gave higher molecular weights and a higher degree of branching than the incorporation of D-galacitol. 40 A potential alternative to the volatile organic solvents or reactions in the bulk at high temperatures is the use of non-toxic supercritical CO 2 (scCO 2 ). 41 The gas-like diffusivity of scCO 2 into polymer matrices provides a unique opportunity to liquefy polymers at temperatures below their glass transition temperature (T g ) and melting point (T m ). This not only lowers the viscosity of the reaction medium dramatically but also allows for a more sustainable low-temperature processing of polymers; both factors lead to significant process energy savings. 42,43 ScCO 2 is a well-established and versatile solvent for polymerisation, 44 polymer modification, 46,47 polymer blending, 43 purification, 43 and extraction, 43 and has been successfully employed as a foaming agent, 48 and for polymer particle formation. 43,49 The liquefying ability of scCO 2 has previously been exploited for the synthesis of a range of polymeric materials with controlled molecular weights and narrow dispersities using enzyme catalyst at near ambient temperature, which ensures the stability of the biocatalysts. 50,51,45,52 For example, scCO 2 was previously employed as a solvent for the synthesis of linear PCL by ROP of ε-CL using Sn(Oct) 2 or Novozym 435 as a catalyst. 33,50,53,54 In this work, we studied several polyols, including 1,6-hexanediol, glycerol, pentaerythritol and triglycerol, as initiators for the synthesis of star polyol-PCL by ring opening polymerisation. Eventually, we investigated the synthesis of complex star PCL architectures using a renewable polyol core, D-sorbitol, as the initiator, aiming at establishing an environmentally benign process to make biodegradable star D-sorbitol-PCL under mild conditions. A mild synthetic procedure is of particular importance when using temperature sensitive polyols such as D-sorbitol as the initiator since it is well known to dehydrate into isosorbide at elevated temperatures. 55 Thus, we first investigated the use of scCO 2 as a solvent for the Sn(Oct) 2 catalysed star PCL synthesis at low temperatures (95 °C), taking advantage of the plasticisation effects of scCO 2 . The results were compared to the synthesis in the bulk at 140 °C. Moreover, we reported on the synthesis of star PCL with a D-sorbitol core using Novozym 435 as the catalyst for the ROP in scCO 2 and in the bulk. Finally, we compared the architecture of the PCL stars synthesised with both catalysts. We also investigated the effect of the number and the variable length of the arms of the star PCL on the thermal properties of the polymers.  After polymerisation, the reaction mixture was dissolved in THF and the enzyme beads were separated by filtration. Subsequently, the excess of solvent was evaporated, and the crude polymer was purified by precipitation into cold ethanol from THF and dried under vacuum. The stainless steel high-pressure autoclave reactor (20 mL) used for the synthesis in scCO 2 has been described previously. 51 Star D-sorbitol-PCL was synthesised in scCO 2 using Sn(Oct) 2 as catalyst following the same procedure as given above for the synthesis in the bulk but at 95 °C. D-sorbitol (93.8 mg, 0.52 mmol) and 10 mol% of Sn(Oct) 2 relative to D-sorbitol (20.8 mg, 0.05 mmol) were loaded into the base of the autoclave and degassed with CO 2 at 3 bar to ensure removal of air.

Results and discussion
Ring-opening polymerisation (ROP) of ε-caprolactone (ε-CL) was carried out in the presence of D-sorbitol as the initiator and using two different catalytic systems (Scheme 1): a metalbased catalyst (Sn(Oct) 2 ) and an enzymatic catalyst (Novozym 435). The influence of both catalysts on the architecture of star D-sorbitol-PCL synthesised in the bulk or in scCO 2 was investigated, thereby avoiding organic solvents as the reaction medium.

Synthesis of star D-sorbitol-PCL via Sn(Oct) 2 catalysed ROP in the bulk and in scCO2
Before investigating the star PCL polymers initiated by D-sorbitol, less complex polymer architectures were synthesised from simpler alcohols such as 1,6-hexanediol, glycerol, pentaerythritol and triglycerol in the bulk (Scheme S1). These syntheses provided the first overview of suitable polymerisation conditions and insights on how the structure of alcohol in combination with Sn(Oct) 2 as a catalyst and the temperature might influence the rate of polymerisation (ESI Table S1).
First linear 1,6-hexanediol was used as an initiator for the synthesis of PCL with a targeted molecular weight of 1000 g mol -1 at 95 °C. Full conversion was observed within 1 h (Table   S1, Entry 1). However, when the amount of initiator was lowered to target a higher molecular weight of 6000 g mol -1 the monomer consumption slowed down after 84% conversion, after 6.5 h (Table S1, Entry 2). At first, this phenomenon was attributed to lower mass transfer due to lower mobility of chains at higher viscosity. As the temperature was increased to 140 °C higher monomer conversions of 97-99% were observed regardless of the initiator (Table S1, Entry 2 vs. Entries 3-7). Nevertheless, the polymerisation rate successively slowed down as the number of hydroxy groups on the initiator increased and the reaction time required to reach >98% conversion increased (Table S1,   Furthermore, at a lower temperature (95 °C) a progressively faster rate of monomer consumption was witnessed as the concentration of Sn(Oct) 2 catalyst was raised 5 fold from 10 mol% (Table 1, Entry 1) to 50 mol% (Table 1, Entry 2). Although Sn(Oct) 2 is FDA approved, it is clear that producing polyesters uncontaminated with possible toxic metallic residues would be desirable for biomedical applications. 58,54 In addition, it is also difficult to completely remove the metal catalyst from the product even after several precipitations.
Thus, increasing the catalyst content is highly undesirable. propagation stage. Additionally, the use of scCO 2 also enhanced the propagation rate in comparison to the bulk reaction (Table 1, Entry 1 and Entry 2 vs. Entry 4) ( Figure 1).  The monomer ε-CL is solubilised by scCO 2 (95 °C, 240 bar) and the resulting mixture plasticises D-sorbitol. 26 Thus, the coordination-complexation between the tin catalyst and alkoxides of D-sorbitol and ε-CL can take place faster in scCO 2 than in the bulk thereby reducing the induction period ( Figure 1). The reduction of the induction period and the higher rate of propagation in scCO 2 as compared to the bulk at 95 °C can be attributed to the unique 'gas-like' mass transfer properties of scCO 2 . 59 Regardless of the reaction temperature or induction periods, the molecular weight of the star D-sorbitol-PCL increased linearly with monomer conversion both in the bulk and in scCO 2 , indicating a well-controlled polymerisation ( Figure S1 -S2). The resulting star D-sorbitol-PCL polymers were analysed by SEC-MALS ( Figure 2) using the dn/dc value of 0.072 mL g -with the targeted (M n targ = 6000 g mol -1 , Table 1). In addition, polymers with narrow dispersities (Đ = 1.03 -1.12) were obtained in both the bulk and scCO 2 (

Synthesis of star D-sorbitol-PCL via Novozym 435 catalysed ROP in the bulk and in scCO 2
Lipases exhibit higher catalytic activity at lower temperatures. 61 Hence, enzymatic ringopening polymerisations (eROP) of ε-CL initiated by D-sorbitol were conducted at 60 °C with varying enzyme concentrations (3 wt% and 10 wt% relative to monomer, Table 2    Decreasing the supported enzyme loading from 10 to 3 wt% in scCO 2 resulted in a decrease in molecular weight and increase in reaction time to reach 96 % conversion ( Table 2 Entry 4 vs. Entry 3, Figure S4 and Figure S6). The same trend was observed in the bulk ( Table 2, Entry 2 vs. Entry 1). However, for 3 wt% enzyme loading in the bulk, the monomer to polymer conversion appeared to be incomplete (87%, Table 2, Entry 1) and plateaued between 13-18 h ( Figure S3). By contrast, the use of scCO 2 with 3 wt% enzyme resulted in a higher molecular weight (M n SEC-MALS = 8800 g mol -1 vs. M n SEC-MALS = 5650 g mol -1 ) and slightly higher conversion (96 %, 24 h vs. 87% 18 h) ( Table 2, Entry 3 vs. Entry 1).
The absolute average molecular weight and dispersity of the star polymers obtained by enzymatic catalysis were determined by SEC-MALS ( Figure 3).  Table 1 and Table 2) and m corresponds to the average number of arms.

Characterisation of star polymer architectures
The influence of the catalyst and the reaction medium on the architecture of the star Dsorbitol-PCL polymers was studied by 1 (Table 1, Entry 3 and Figure 4(A)) are 9.7 and 1100 g mol -1 , respectively. In general, the M n H-NMR (arm) of star D-sorbitol-PCL synthesised using Sn(Oct) 2 (1100-1290 g mol -1 , Table 1) were lower than those obtained with Novozym 435 (2700-3400 g mol -1 , Table 2).
Since   Such low dispersities suggest a uniform sample of star-PCL, where the number of arms is the same. The length of the arms of one individual star polymer may be the same or of different length, but with a uniform architecture throughout the sample.
These higher dispersities could be caused by transesterification side reactions occurring during eROP. Macrocycles formed by back-biting (intramolecular transesterification) or linear PCL initiated by water (mainly originating from the enzyme) can also be present in small amounts in the final star polymers. 50,65 In addition, since Novozym 435 is regioselective, 66 initiation from the primary hydroxy groups of D-sorbitol is expected to be faster than initiation from the secondary hydroxy groups, implying that there will be arms with different lengths (i.e., shorter arms from the secondary alcohols of D-sorbitol) and leading to star PCL with a miktoarm star architecture.

Mark-Houwink plot for star-branched polymer
The structure of the star polymers can be confirmed by SEC analysis with an RI and a viscosity detector using the Mark-Houwink relationship. The slope (α) of the double-  Table 1) which indicates a more branched structure and thus a larger number of arms.
MALDI-TOF analyses were performed in order to further investigate the synthesised star polymers.     25 OH] m K +

Quantifying the number of arms by a phosphitylation/ 31 P NMR method
The quantitative investigation of the number of arms (N arms ( 31 P)), the number average molecular weight of the arms (M n P-NMR (arm)), the molecular weight of the star polymer (M n P-NMR (star polymer)) and the percentage of linear PCL chains present in star D-sorbitol-PCL was performed by the phosphitylation method developed by Spyros et al. 72 (Table 3).
The N arms ( 31 P) for enzyme catalysed star PCL was evaluated accounting for the water initiated linear PCL chains (HOOC-PCL-OH) (Equation (2) (2) a Average number of arms quantified following Eq. (2) and the corresponding inverse gated decoupled 31 P NMR spectra ( Figure S10 and Figure S11) The M n P-NMR (arm) determined by 31 P NMR (Table 3 and Table 4) are in good agreement with the results from M n H-NMR (arm) analysis (Table 1 and Table 2). An average number of arms of 5 is determined by both the 31 P and 1 H NMR analyses in addition to the narrow dispersities observed by SEC-MALS (Ð = 1.03-1.12, Table 1) and the uniform distribution  Figure 8 (B)). It is important to state that detailed analysis of the phosphitylation process ( Figure S8 (B)) does show that the process is unlikely to be fully quantitative on D-sorbitol and that there are some minor impurities arising from other species that are revealed in the 2° peaks. These will, of course, contribute to the 31 P NMR spectra of the star polymers ( Figure   7) and introduce some error into our calculations of corroborate the differences observed in the products arising from tin catalysis and from the enzyme.

DSC (in N 2 )
DSC was employed to investigate the g mol -1 , Table S1 (Table 2). As expected, the polymers were semi olyol-PCL decreases with an increasing number of arms formation of crystalline domains in PCL ( Figure S12 agreement with a previous report by Endo et al. 19 who observed that a when compared to star PCL of similar molecular weight ) of the regular star D-sorbitol-PCL synthesised using is lower than that of the star D-sorbitol-PCL synthesised by eROP in scCO 2 (N arms ( 31 P) = 3.2, T m = 41 °C), which in turn is lower than the T m of the hexanediol-PCL synthesised using Sn(Oct) 2 in the bulk (N arms ( 31 P) = 2, T m = 45 °C) (Table S3).
Multiple melting peaks have also been reported in the literature for both linear poly(L-lactic acid) 74 and also for star-shaped poly(L-lactic) acid. 75 This observation is attributed to the presence of crystalline domains of different size and different perfection.

Conclusion
A sustainable route for star-shaped poly(ε-caprolactone) synthesis was implemented using renewable feedstock in clean solvents (scCO 2 and bulk). The effect of the catalytic system (Sn(Oct) 2  Moreover, the eROP route leaves residual unreacted hydroxy groups on the core, something that cannot be achieved by a metal catalyst.