“Like Recycles Like”: Selective Ring‐Closing Depolymerization of Poly(L‐Lactic Acid) to L‐Lactide

Abstract Chemical recycling of poly(L‐lactic acid) to the cyclic monomer L‐lactide is hampered by low selectivity and by epimerization and elimination reactions, impeding its use on a large scale. The high number of side reactions originates from the high ceiling temperature (T c) of L‐lactide, which necessitates high temperatures or multistep reactions to achieve recycling to L‐lactide. To circumvent this issue, we utilized the impact of solvent interactions on the monomer–polymer equilibrium to decrease the T c of L‐lactide. Analyzing the observed T c in different solvents in relation to their Hildebrand solubility parameter revealed a “like recycles like” relationship. The decreased T c, obtained by selecting solvents that interact strongly with the monomer (dimethyl formamide or the green solvent γ‐valerolactone), allowed chemical recycling of high‐molecular‐weight poly(L‐lactic acid) directly to L‐lactide, within 1–4 h at 140 °C, with >95 % conversion and 98–99 % selectivity. Recycled L‐lactide was isolated and repolymerized with high control over molecular weight and dispersity, closing the polymer loop.


Polymerization of LLA
Synthesis of PLLA PLLA was synthesized in bulk by varying the ratio of initiator (BnOH) to monomer (LLA) to target different molecular weights (Supporting Table 1.1-.1.4). As an example of the general procedure, LLA (20.0 g, 0.139 mol) was added to a round-bottom flask together with BnOH (0.150 g, 0.00139 mol) and Sn(Oct)2 (0.563 g, 0.00139 mol) as catalysts. ROP was carried out under a N2 atmosphere at 100 °C for 1 h, after which the reaction was quenched by cooling. The polymer was dissolved in CHCl3 and precipitated in a mixture of cold heptane and methanol (10 mol% methanol). The precipitation was repeated four times to remove Sn(Oct)2 and unreacted monomers. The purified polymer was dried in a fume hood overnight and thereafter under vacuum (4 days at room temperature and 2 days at 60 °C). The polymer was stored in a glove box under a N2 atmosphere.
End capping of PLLA PLLA (Mn,SEC = 14.1 kDa, Ð = 1.13; Supporting Table 1.1) (1.0 g, 0.42 mmol -OH end groups) was added to a 25 mL round bottom flask equipped with a magnetic stirrer together with CH2Cl2 (10 mL) and triethylamine (290 μL, 2.1 mmol). Acetyl chloride (75 μL, 1.1 mmol) was added dropwise under stirring. The flask was kept under a N2 atmosphere, and the reaction was let to proceed for 1 h at RT (Supporting Figure 6). The solution was washed with HClaq (10 mL, 0.5 M) followed by water (10 mL) and brine (3x10 mL). CH2Cl2 was removed on the rotary evaporator followed by drying under vacuum for 2 days.

Calculation of polymerization thermodynamic parameters
To calculate the thermodynamic parameters (ΔHp and ΔSp) of LLA polymerization in different solvents (DMF, GVL, DMSO and PhCl), polymerization was carried out in solution ([M]0 = 0.5 M LLA) at four different temperatures per solvent. As a general procedure, LLA (0.144 g, 1.00 mmol) was added to a 10 mL flask equipped with a magnetic stirrer together with BnOH (0.1 mL of 1 mM DMF stock solution, 0.01 mmol) and 1.7 mL of DMF (Supporting Table 3). The preparation of the reaction vessel was performed inside a glove box under a N2 atmosphere. The vial was sealed with an aluminum/Teflon crimp cap with a rubber septum before being transferred out of the glove box into the open laboratory. The vial was immersed in a preheated oil bath, and the reaction was started by adding DBU (0.2 mL of 0.5 M DMF stock solution, 0.1 mmol) through the septum with a syringe. Aliquots of 0.1 mL were removed through the septum at regular time intervals, quenched by cooling in 0.6 mL of CDCl3, and analyzed by 1 H NMR spectroscopy. The reaction was allowed to proceed until an equilibrium monomer concentration was established. This process was repeated at four different temperatures, and the equilibrium monomer concentration

Depolymerization of PLA
Six different polymers with varying molecular weights were utilized for depolymerization experiments. Four polymers were synthesized as described above, while the fifth was commercial grade PLA (PLA Ingeo TM Biopolymer 4043D, NatureWorks) (Supporting Fig. 1). The fifth polymer was end capped as described above. All polymers were used as crushed powders (Supporting Fig. 2).
The solvent, polymer concentration, catalyst, catalyst concentration and temperature were varied to explore the effect on the depolymerization and side reactions. In general, PLLA (0.072 g or 0.144 g; 0.5 or 1.0 mmol) was added to a 10 mL flask equipped with a magnetic stirrer together with catalyst (1-10 mol %; Sn(Oct)2, DBU or TBD) and 2.0 mL of solvent (DMF, GVL, DMSO, PhCl, DX, acetone, CHCl3 or PhMe), resulting in a concentration of 0.5 M (based on the LLA repeating unit). The preparation was performed inside a glove box under a N2 atmosphere. The vial was sealed with an aluminum/Teflon crimp cap with a rubber septum before being transferred out of the glove box. The vial was quickly heated with a heating gun to obtain a homogenous solution directly before it was immersed in a thermostatic oil bath (65 °C, 140 °C 160 °C or 180 °C). Aliquots of 0.1 mL were removed through the septum at regular time intervals, quenched by cooling in 0.6 mL of CDCl3, and analysed by 1 H NMR spectroscopy without further purification.
For depolymerization in a mixed plastic waste stream, pieces of postconsumer plastic composed of PE, PP, PET and PC (Supporting Fig. 10) and pellets of PA6.6 were used. The experiments were performed with both PLLA (Mn = 14,000 g mol -1 , Ð = 1.13) and with pieces cut from a PLA cup (Supporting Figure 10). The mixed plastics were added to DMF (0.5 PLLA/PLA based on the LL/LA repeating unit). After depolymerization the insoluble plastic pieces were separated from the solution by filtration. The pieces were photographed before and after the reaction (Figure 3c, Supporting Figure 10).

Repolymerization of LA
Direct repolymerization PLLA (Mn,SEC = 14.1 kDa, Ð = 1.13; Supporting Table 1.1) was depolymerized in DMF (0.5 M based on the LLA repeating unit) at 140 °C (details described above). 10 mol% TBD was added directly to the reaction mixture. The polymerization was performed at RT, and aliquots were withdrawn at regular time intervals, quenched by acetic acid, and analyzed by 1 H NMR spectroscopy without further purification.
Isolation and repolymerization of LA PLA (3.60 g, 25.0 mmol: Ingeo TM Biopolymer 4043D, NatureWorks, Mn = 110 kDa, Ð = 2.02; Supporting Figure 1) was depolymerized in DMF (50 mL) with Sn(Oct)2 (1.01g, 2.50 mmol) at 140 °C. The sample was concentrated from 0.5 M to 3 M LLA by distillation of DMF at 25 °C under dynamic vacuum (0.2-0.3 mbar) for 10 h. Thereafter, 50 mL of n-heptane was added and distilled off at 25 °C. This was repeated four times until a thick viscous slurry was obtained. LA was recrystallized from the slurry in toluene, and the crystals were dried under vacuum overnight. ROP of recycled LA was performed as described above (details in Supporting Table  1.5). The crude polymer was analysed by 1 H NMR and SEC without further purification (Supporting Figure 1).

Characterization
Nuclear magnetic resonance (NMR) spectroscopy 1 H NMR spectra were recorded on a Bruker Avance III HD (400 MHz) spectrometer. All experiments were performed at RT with CDCl3 as the solvent and with CHCl3 as the internal reference.

Size exclusion chromatography (SEC)
Molecular weight analysis via SEC was performed on a Malvern GPCMAX instrument equipped with an autosampler, a PLgel 5 μm guard column (7.5 x 50 mm) and two PLgel 5 μm MIXED-D (300 x 7.5 mm) columns. The polymer sample (4-5 mg mL -1 ) was dissolved in chloroform containing 2% v/v toluene, which also was used as eluent. The flow rate was 0.5 mL min -1 , and the temperature was kept at 35 °C. Narrow disperse polystyrene standards with molecular weights in the range of 1,200-400,000 g mol -1 were used for calibration.

Molecular dynamics (MD) simulations
All MD simulations were performed using GROMACS [1] version 2020.2 using a stochastic integration algorithm [2] with a basic time step of 2 fs. Nonbonded interactions were cut off at 1.2 nm and shifted to ensure zero potential at the boundary. Electrostatic interactions were treated with PME [3,4] using a real-space cutoff of 1.2 nm. Pressure was maintained at 1 atm using a Parrinello-Rahman barostat [5] and a compressibility of 5·10 -5 bar -1 , while temperature was controlled by the integration algorithm. All covalent bonds were constrained to their equilibrium value using P-LINCS. [6] All simulations employed a replica exchange protocol [7] using eight temperatures ranging from 323 K to 363 K in steps of 8 K. Exchange between neighboring replicas was attempted every 1000 steps.
The simulations were run on cyclic L-lactide and PLLA oligomers of DP 5 or 6 in different solvents using a fully periodic computational box in the shape of a truncated octahedron with an approximate nearest-image distance of 3.5 nm. The liquid molecules included in the simulations were DMSO, DMF, DX, and PhCl. All compounds were modeled using the general CHARMM force field (CGenFF) version 2.4.0. [8][9][10] The liquid densities showed good agreement with the experimental values (Supporting Fig. 11).
Solvation free energies were calculated using Computational Alchemy [11] in which solute-solvent interactions were linearly decoupled using a single coupling parameter. The decoupling was performed in 20 discrete steps, where the first ten were used to decouple electrostatic interactions, followed by van der Waals interactions in the remaining steps. Thus, the fully decoupled state corresponds to the gas phase of the solute and the pure liquid state of the solvent. For each value of the coupling parameter, a 2 ns equilibrium simulation was performed during which the derivative of the total potential with respect to the coupling parameter was sampled. This quantity can finally be connected to the free energy difference between the start and end states, GS, using Bennett's acceptance ratio. [12] S5

Supporting Note 1. Calculation of relative amounts of LLA, DLA meso-LLA and PLLA.
The fraction of meso-LA (b) in the crude lactide mixture was calculated from the integrals of the CH3 peaks for meso-LA and LLA/DLA: Based on the probability law, the LLA (L) and DLA (D) content in the crude lactide can be calculated from b accordingly: [14] = (1+√1−2 )

(SE3)
The total conversion to monomer was calculated from the CH peaks for LLA/DLA and meso-LA/PLLA:

(SE15)
For long linear chains, the term ( − 1)⁄ will be approaching 1 and can, therefore, be neglected (Flory's assumption). However, for shorter oligomeric chains (approx. DPn ≤ 20) the effect of DPn on the thermodynamic equilibrium between polymer and monomer is noteworthy, and the [M]eq will decrease with decreasing DPn. This explains the observed decline in conversion of polymer to monomer with decreased feed molecular weight (Supporting Tab. 1.1-1.3