Stereoselective depolymerization of chiral polyesters

Abstract

Enzymes precisely structured chiral active pockets can achieve highly selective recognition of chirality in small molecules as well as macromolecule, enabling asymmetric transformation. Inspired by enzymic catalysis, the design of synthetic catalysts has enabled diverse asymmetric reactions of small molecules. However, stereoselective transformations of macromolecules catalyzed by synthetic catalysts remain unexplored. This work establishes a significant advancement in the field by demonstrating chiral recognition and conversion of polymers using synthetic catalysts. Employing the representative chiral polymer polylactic acid (PLA) as a model substrate, a specially designed catalyst BisSalen-Al with confined chiral cavity facilitated the stereoselective depolymerization of PLA into chiral lactide (LA) monomers. In depth exploration of the relationship between catalyst structure and stereoselectivity revealed the chiral recognition mechanism of polymer stereochemistry and catalyst chiral cavity matching. This work provides a strategy for the stereoselective conversion of chiral polymers and significantly advances the expansion of asymmetric catalysis into macromolecular systems.

Lactide is enantiomeric, and depolymerization of poly(lactic acid) (PLA) with mixed chirality typically produces a racemic mixture. Here, the authors use a chiral Aluminium catalyst for the stereoselective depolymerization of PLLA and PDLA towards chiral lactide with high optical purity.

Introduction

Chirality is a fundamental characteristic of nature. At the molecular level, this chiral feature is manifested in the precise stereoselective transformation of small biomolecules and biomacromolecules (e.g., DNA, RNA, proteins), ensuring efficient metabolic pathways, accurate genetic information transfer, and proper protein function¹. The key executors of such high selectivity are enzymes^2–4. Enzymes utilize well-defined chiral active pockets to distinguish the stereochemistry of chiral/prochiral substrates and specific sequences on polymer chains, thereby enabling asymmetric transformation for constructing and degrading chiral biomolecules^5,6. Such a chiral recognition and conversion process is thus one of the most essential functions in maintaining life.

Remarkably, asymmetric catalysis based on chiral discrimination extends beyond natural biomolecules. Enzymes also exhibit exceptional recognition toward synthetic substrates^7,8 (Fig. 1A). For example, proteases (e.g., Proteinase K) can selectively recognize poly(L-lactic acid) (PLLA) and complete its catalytic hydrolysis but show minimal activity toward poly(D-lactic acid) (PDLA)⁹. The inevitability of this phenomenon stems from that proteases must evolve specific recognition for proteins (L-amino acid polymers), which share structural similarities with PLLA¹⁰. The strong chiral recognition exhibited by enzymes towards both biological and synthetic macromolecules highlights the structural sophistication and reveals intrinsic connections between living systems and synthetic chemistry.

Fig. 1. Asymmetric catalysis for small molecules and macromolecules using enzyme and synthetic catalysts.

A Enzyme catalysts. B Synthetic catalysts.

Achieving asymmetric catalysis using synthetic catalysts (e.g., metal complexes, organocatalysts) represents a major pursuit in chemistry^11–14 (Fig. 1B). For example, in the pioneering work of Sharpless, asymmetric dihydroxylation of olefins by introducing chiral dihydroquinine ligand into osmium tetroxide was achieved^15,16. Sharpless was awarded the Nobel Prize in Chemistry in 2001 for his contributions on asymmetric catalytic oxidation. In 2021, List and MacMillan were awarded the Nobel Prize in Chemistry for their contributions on asymmetric organocatalysis as well. A well-known work of List was the design of catalysts such as imidodiphosphorimidate (IDPi) with chiral biaxial skeletons to construct confined spaces similar to biological enzymes, achieving a series of advances in asymmetric functionalization^17,18. Previous studies have fully demonstrated that catalyst design is the key to achieving high stereoselectivity. However, despite these advances in asymmetric catalysis, research remains focused on small-molecule transformations^19,20. Compared to small molecules, polymers possess complex spatial arrangements and multi-chiral sites. This feature places higher demands on the stereo-discrimination of synthetic catalysts, while also making traditional small molecule catalytic mechanisms difficult to apply. Consequently, catalytic recognition of chiral macromolecules by synthetic catalysts remains virtually unexplored.

This study focuses on the key challenge of whether a synthetic catalyst can simulate enzymatic function for precise macromolecular recognition. The most representative synthetic chiral polymers, PLLA and PDLA, were used as a model chiral polymer. Inspired by the structural design concept of enzyme active pockets, we engineered a metal Al complex with dual chiral Salen ligands (BisSalen-Al) to create a confined chiral cavity^21–24. The (R,R)-BisSalen-Al enables stereoselective depolymerization of PDLA into chiral monomer D-lactide (D-LA). Mechanistic studies reveal BisSalen-Al mediated recognition and activation pathways analogous to enzymatic catalysis. Our work fills a critical gap in synthetic catalysis, expands the frontiers of chiral macromolecule conversion, and deepens insights into enzymatic recognition mechanisms.

Results

Structure design and optimization of confined chiral catalysts

Before investigating the stereoselective catalysis toward macromolecules, the selection of model substrates and conversion pathways were first considered. Polylactic acid (PLA) is one of the most representative synthetic chiral polymers, and its stereochemical diversity (PLLA, PDLA, PLLA-PDLA stereocomplex etc.) provides an ideal model for studying the stereoselective recognition²⁵. Under catalytic conditions, PLA can be depolymerized into the initial monomer lactide (LA). This conversion process enables closed-loop transformation of PLA and represents its most crucial depolymerization pathway^26–32. Current researches focus on how to achieve efficient depolymerization or improve the optical purity of LA by controlling the racemization side reaction, while largely overlooking the aspect of chiral recognition of the polymer^28,32. Concurrently, analyzing the generation and optical purity of the obtained LA can directly reflect stereoselectivity³³. Therefore, depolymerization into monomer is established as the reaction model.

Depolymerization was conducted in a solvent medium to better investigate the catalyst-substrate interaction mechanism. Nevertheless, solvent selection should first be optimized to ensure successful depolymerization of PLA into LA, thereby circumventing limitations imposed by polymerization-depolymerization equilibrium^34,35. Under solvent-free conditions, the ceiling temperature (T_c, the temperature at which rates of polymerization and depolymerization are equal) of PLA is as high as 641 °C³⁶. The addition of solvent affects monomer concentration, which consequently influences the T_c. Thus, the solvent should play an efficient dilution role, so that the T_c is significantly decreased^37,38. Besides, the solvent will affect the interaction between the solvent, monomer and polymer^39,40. In the solvent conducive to depolymerization, the chain propagation rate (k_p) associated with the solvent-monomer interaction should be slowed down to highlight the effect of depropagation (k_d). Based on the above solvent selection principles, solvent screening was conducted. It was found that acetonitrile (MeCN), a solvent with high polarity, possessed favorable potential. First, the thermodynamics was verified through the polymerization of L-LA. The thermodynamic parameters ΔH_p⁰ and ΔS_p⁰ were obtained as ΔH_p⁰ = −17.4 kJ·mol⁻¹ and ΔS_p⁰ = −36.9 J·mol⁻¹·K⁻¹, respectively (Supplementary Fig. S23). It could be further deduced that when the initial monomer concentration [M]₀ was 0.5 mol·L⁻¹, T_c was reduced to be 135 °C. Second, the polymerization rate under the same catalytic conditions in different solvents was also compared. The results revealed a lower chain propagation rate in MeCN before reaching equilibrium compared to low-polarity solvents such as DCM and toluene (Supplementary Fig. S23). Based on the above analysis, depolymerization was first attempted with trimethylaluminum (AlMe₃) as the catalyst at 140 °C in 0.5 M MeCN. As expected, both PLLA and PDLA achieved 86% LA conversion within 6 h, confirming the viability of MeCN as the solvent (Supplementary Table S9, entries 1 and 2).

To be noted, in the absence of a catalyst, it was found that depolymerization was difficult to occur (Supplementary Table S9, entries 3 and 4), indicating that the PLA polymer is kinetically trapped. That is to say, the catalyst is crucial for activating the depolymerization process. Certain commercially available catalysts, such as tin(II) octanoate (Sn(Oct)₂) and magnesium chloride (MgCl₂)⁴⁰, have been experimentally proven to catalyze the depolymerization of PLLA to L-LA. However, these catalysts equally promote the depolymerization of PDLA (Supplementary Table S10). Consequently, achieving stereoselective recognition of PLA necessitates rational catalyst design.

As we known, the chiral unit arrangement and spatial conformation provide the enzyme-specific chiral cavity for asymmetric transformation of macromolecules. Inspired by enzymic catalysis, if tighter chiral ligands are introduced to endow the metal catalyst with constrained chiral space surrounding the active site, it is possible to achieve the stereoselective recognition. Salen ligand featuring cyclohexanediamine linker was initially considered (Fig. 2). The derived (R)-Al-1 catalyst, as demonstrated in prior studies, could achieve chiral recognition of lactide monomers, in which (R)-Al-1 polymerizes L-LA significantly faster than its D-enantiomer, with a polymerization rate difference of k_L-LA/k_D-LA = 14⁴¹. However, when it was applied to the selective depolymerization of PLA, this catalyst exhibited only slight selectivity between the two polymeric configurations (PLLA and PDLA). The analysis of the depolymerization process conformed to the approximatively first-order kinetics, and the slope comparison showed that the depolymerization rate difference k_PLLA/k_PDLA was merely 1.4 (Table 1, entries 1 and 2, Fig. 3a, Supplementary Tables S11, 12). This result indicates that although there is a certain stereoselective correlation between polymerization and depolymerization, chiral recognition of macromolecular PLA presents a greater challenge than that of monomeric LA^42–45.

Fig. 2.

Catalyst structures used in the selective depolymerization of PLA.

Table 1.

Study on the influence of catalyst structure on selective depolymerization

Entrya	Polymer	Cat.b	t (h)	Conv.c (%)	meso-LAd (%)	L-LA: D-LAe	kf	kfast/kslow
1	PLLA	(R)-Al-1	96	80	5	99.8: 0.2	0.0226	1.4
2	PDLA		96	79	6	0.7: 99.3	0.0157
3	PLLA	(R)-Al-2	1	10	/	/	0.0514	17.4
4	PDLA		1	93	2	0.1: 99.9	0.8922
5	PLLA	(S,S)-Al-3	2	94	1	98.4: 1.6	0.3459	32.0
6	PDLA		2	5	/	/	0.0108
7	PLLA	(R,R)-Al-3	2	5	1	/	0.0108	30.3
8	PDLA		2	90	/	1.0: 99.0	0.3270
9	PLLA	(R)-Al-4	8	6	/	/	0.0039	19.7
10	PDLA		8	89	1	0.4: 99.6	0.0770
11	PLLA	(S,S)-Al-5	8	84	2	99.8: 0.2	0.0860	15.9
12	PDLA		8	9	/	/	0.0054
13	PLLA	(S,S)-Al-6	8	83	2	99.7: 0.3	0.1263	22.6
14	PDLA		8	9	/	/	0.0056
15	PLLA	(R,R)-Al-7	1.5	4	/	/	0.0146	37.5
16	PDLA		1.5	91	< 1	0.6: 99.4	0.5479

^aThe reaction temperature was 140 °C, using laboratory-prepared PLLA-100 and PDLA-100 as the raw material, catalyst (2 mol%), and the PLA repeat unit concentration was 0.5 mol·L⁻¹ in MeCN. The reaction was conducted in a sealed pressure-resistant glass tube. The mass of all components is conserved within the sealed system, and the process involves a dynamic equilibrium of solvent vaporization and condensation. Therefore, the reaction concentration is based on the initially charged amounts. ^b Details of the catalysts used are illustrated in Fig. 2. ^c Depolymerization conversion was determined by ¹H NMR spectrum of the crude product in CDCl₃ by comparing the integration of methine signal peaks. ^d The content of meso-LA was determined and calculated by ¹H NMR spectrum by comparing the integration of methyl signal peaks. ^e L-LA: D-LA of the recovered monomer was measured by chiral HPLC. ^f The depolymerization catalyzed by Al-1 follows first-order kinetics, thus, its rate constant k is the slope of the ln[1/(1-conv.)] versus time relationship graph, carries units of h⁻¹. In contrast, Al-2 to Al-7 adhere to zeroth-order kinetics, its rate constant equals the product of the slope of the conversion-time graph and the initial concentration (0.5 mol·L⁻¹). resulting in a rate constant expressed in mol·L⁻¹·h⁻¹.

Fig. 3. Stereoselective conversion to chiral monomers for PLA.

a The depolymerization conversion versus time for PLLA and PDLA catalyzed by (R)-Al-1 (k_PLLA = 0.0226 h^-1, k_PDLA = 0.0157 h^-1). b The depolymerization conversion versus time for PLLA and PDLA catalyzed by (R)-Al-2 (k_PLLA = 0.0514 mol·L⁻¹·h⁻¹, k_PDLA = 0.8922 mol·L⁻¹·h⁻¹). c GPC traces for PLLA before and after depolymerization, using THF as the eluent (flow rate: 1 mL/min, at 40 °C), and the sample concentration was 1 mg/mL. d The depolymerization conversion versus time for PLLA and PDLA catalyzed by (S,S)-Al-3 (k_PLLA = 0.3459 mol·L⁻¹·h⁻¹, k_PDLA = 0.0108 mol·L⁻¹·h⁻¹). e HPLC plots of recovered LA for PLLA catalyzed by (S,S)-Al-3 and PDLA catalyzed by (R,R)-Al-3. f The depolymerization conversion versus time for PLLA and PDLA catalyzed by (R,R)-Al-7 (k_PLLA = 0.0146 mol·L⁻¹·h⁻¹, k_PDLA = 0.5479 mol·L⁻¹·h⁻¹).

Further investigation introduced the more rigid binaphthyl-based catalyst (R)-Al-2 into the depolymerization system (Table 1, entries 3 and 4). Monitoring at 1 h revealed 93% of PDLA was converted to D-LA, while only 10% of PLLA underwent conversion to L-LA. The difference in the depolymerization activity of PDLA and PLLA indicated that the chiral matching plays an important role in promoting the stereoselective depolymerization. Upon investigating the reaction kinetics, the depolymerization conversion exhibited a nearly linear relationship with reaction time, conforming to an approximate zeroth-order kinetic behavior⁴⁶. The rate difference between the two polymeric configurations was calculated as 17.4 (Fig. 3b and Supplementary Tables S13, 14). This enhanced selectivity indicated that the rigid binaphthyl-based ligand provided better chiral spatial matching. This result inspired us to further enhance stereoselectivity through ligand regulation, thus, a dual-ligand strategy in catalyst design to construct BisSalen-Al was implemented, postulating that simultaneous encapsulation of the metal center by two ligands would create a more rigid recognition pocket. To our delight, when employing the (S,S)-Al-3 for the depolymerization of PLLA, the depolymerization to L-LA achieved a 94% conversion within 2 h, showing that the catalyst still maintains good catalytic activity (Table 1, entry 5). Conversely, when the depolymerization object was replaced by PDLA, the depolymerization process was significantly hindered, with only 5% D-LA produced after 2 h (Table 1, entry 6). Analysis of molecular weight and dispersity for PDLA revealed almost no detectable change after depolymerization, confirming minimal depolymerization of PDLA (Fig. 3c). Comparing the depolymerization kinetics of PLLA and PDLA, the reaction rate difference k_PLLA/k_PDLA reached up to 32.0 (Fig. 3d and Supplementary Tables S15, 16). The superior stereoselectivity of the dual-ligand Al-3 catalyst compared to its single-ligand counterpart Al-2 (reaction rate difference = 32.0 vs. 17.5), demonstrated the efficacy of the dual-ligand strategy in creating a more confined chiral space. Furthermore, the (R,R)-Al-3 was used as well (Table 1, entries 7 and 8, Supplementary Tables S17, 18). It catalyzed the CRcM (chemical recycling to chiral monomers) for PDLA smoothly with the conversion reached 90% after 2 h, while the PLLA was unable to depolymerization in this case (2 h, conv. = 5%). These distinct preferences further revealed the high stereoselectivity of the BisSalen-Al catalyst. Notably, the chiral high performance liquid chromatography (HPLC) analysis showed that the enantiomeric excess (ee) of the recovered lactide monomers is high for Al-2 and Al-3 (Table 1, entries 3–8, Fig. 3e), consistent with the initial PLA configuration, indicating that there was almost no epimerization side reaction during this highly controllable depolymerization process.

Furthermore, the study shifted focus to the influence of the second ligand structure on stereoselective recognition. Initially, the achiral second ligand N-salicylideneaniline was considered ((R)-Al-4). Under this condition, a decrease in reactivity for chirality-matched PLA was observed. After a prolonged time of 8 h, the depolymerization conversion of PDLA reached 89% (Table 1, entry 9), while that of PLLA was only 6% (Table 1, entry 10). The reaction rate difference was calculated as descend to 19.7 (Supplementary Tables S19, 20 and Supplementary Fig. S29). It was speculated that the chirality of the second ligand may play a crucial role in enhancing selectivity. Therefore, second ligands featuring a binaphthyl chiral skeleton ((S,S)-Al-5 and (S,S)-Al-6) were introduced for polymerization. Their catalytic depolymerization activity resembled that of Al-4, as the chirality-matched PLLA was basically completed the depolymerization after 8 h. The stereoselectivity analysis revealed that with no additional substituents on the binaphthyl skeleton ((S,S)-Al-5), reaction rate difference was merely 15.9 (Table 1, entries 11 and 12, Supplementary Fig. S30). Upon introducing a methoxy substituent ((S,S)-Al-6), reaction rate difference increased to 22.6 yet remained below the stereoselectivity level of Al-3 (Table 1, entries 13 and 14, Supplementary Fig. S31). These results suggested that the two salicylaldimine units in the second ligand appear to be key to enhancing stereoselectivity. To further verify this conclusion, catalyst (R,R)-Al-7, whose second ligand also contains two salicylaldimine units but features a partially hydrogenated binaphthyl skeleton, was synthesized and applied. In this case, depolymerization activity for chirality-matched PLA returned to a high level. After 1.5 h, depolymerization conversion for PDLA reached 91% with a rate constant of 0.5479 mol·L⁻¹·h⁻¹ (Table 1, entry 15), while the chirality-mismatched PLLA exhibited only 4% conversion (Table 1, entry 16). Kinetic studies revealed the rate difference was further elevated to 37.5 (Fig. 3f and Supplementary Tables S25, 26). These findings further illustrate the influence of the structure of the second ligand on depolymerization selectivity and highlight the critical role of the dual ligand strategy in constructing confined chiral pockets.

Investigation of stereoselective depolymerization mechanism

When considering the depolymerization mechanism, two paths are mentioned: chain-end unzipping mode or random chain scission mode^47,48. The former refers to the depolymerization occurring at the chain-end through a sequential back-biting reaction, hence the concentration of active chain ends remains relatively constant during the reaction. Therefore, the reaction rate remains constant, exhibiting zeroth-order kinetic behavior. It is reflected in the depolymerization results as a linear correlation between reaction conversion and time (conv. vs. time). The latter refers to the depolymerization randomly occurring in the polymer chain, accompanied by the rapid decrease in molecular weight. The reaction rate is directly related to the concentration of ester bonds in the polymer. As the reaction proceeds, the concentration of ester groups in the polymer decreases, resulting in a decline in activity accordingly. It results in first-order kinetic behavior, which is reflected in the depolymerization data as a linear correlation between ln([C]₀/[C]_t) (where [C]_t = [C]₀*(1 - conv.), which can be further expressed as ln[1/(1 - conv.)]) and the reaction time. As mentioned above, the depolymerization kinetics catalyzed by Al-1 was approximated as a first-order kinetic model, while Al-2 to Al-7 followed kinetics closer to a zeroth-order model. This kinetic divergence indicated that Al-1 likely mediated random chain scission, whereas Al-2 to Al-7 activated depolymerization via an unzipping back-biting mechanism. The difference in reaction mechanisms further illustrates the influence of catalyst structure. Catalysts Al-2 to Al-7 possess a confined spatial structure, which enables specific interaction with polymer chain ends of the matched stereochemistry. This restricts the occurrence of side reactions like random transesterification. In contrast, the more open molecular structure of Al-1 lacks this end-group selectivity, making it more prone to engage in random attacks. The unzipping back-biting mechanism was further demonstrated by the following experimental facts. As shown in Fig. 4a, when higher molecular weight PDLA was depolymerized catalyzed by (R,R)-Al-3 (M_n = 47.5 kg·mol⁻¹, Đ = 1.15), the depolymerization rate decreased significantly. This phenomenon originated from that as longer chain lengths result in more chain-end back-biting steps, manifesting as a decline in overall rate. Meanwhile, a detailed trend analysis of the molecular weight revealed that the M_n of the polymer decreased linearly with depolymerization conversion, and the dispersity remained consistently narrow within the range of 1.29 (Figs. 4b, c and Supplementary Table S27). This phenomenon provided direct evidence for controllable chain-end back-biting mechanism^49,50. Additionally, the MALDI-TOF MS analysis of the depolymerized PLLA at 39% conversion catalyzed by (S,S)-Al-3 further confirmed the uniform decline in molecular weight, with the linear polymer chain structure and good chain-end fidelity during depolymerization (Fig. 4d). While, for non-preferred PDLA, the molecular weight remained almost unchanged as mentioned above (Fig. 3c). Furthermore, acetyl-capped PLLA (PLLA-Ac) was synthesized and subjected to depolymerization (Fig. 4e and Supplementary Fig.  S50). Experimental results demonstrated that when (S,S)-Al-3 was employed as the catalyst, the depolymerization conversion of capped PLLA-Ac reached only 12% after 12 h, with minimal change in molecular weight before and after depolymerization (Fig. 4f and Supplementary Fig. S51). In contrast, under identical conditions and duration, uncapped PLLA underwent nearly complete conversion (Table 1, entry 5). This outcome further confirms the activation effect on terminal hydroxyl groups. Based on the above information, the stereoselective depolymerization process could be depicted as follows. For the preferred PLA, the chiral catalyst matched with configuration of chiral chain-end to promote depolymerization. While for non-preferred PLA, the mismatching mode of chirality in the catalyst and chain-end hindered the depolymerization, thus preserving the PLA chain structure.

Fig. 4. Investigation of chain-end back-biting mechanism.

a The depolymerization conversion versus time for PDLA with different molecular weights catalyzed by (R,R)-Al-3 (100 DP: M_n = 16.4 kg·mol⁻¹, Đ = 1.19, k_100DP = 0.3270 mol·L⁻¹·h⁻¹; 300 DP: M_n = 47.5 kg·mol⁻¹, Đ = 1.15, k_300DP = 0.1497 mol·L⁻¹·h⁻¹). b The molecular weight (M_n) and dispersity (Đ) versus conversion for depolymerization of PDLA catalyzed by (R,R)-Al-3 (PDLA-300, M_n = 47.5 kg·mol⁻¹, Đ = 1.15). c, GPC traces for depolymerization of PDLA-300 at different conversion catalyzed by (R,R)-Al-3, using THF as the eluent (flow rate: 1 mL/min, at 40 °C) and the sample concentration was 1 mg/mL. d MALDI-TOF mass spectrum characterization of PLLA after depolymerization catalyzed by (S,S)-Al-3 (PLLA-100, 0.5 h of depolymerization, conv. = 39%). e Comparison of the ¹H NMR spectra of PLLA with hydroxyl end groups and acetyl-capped PLLA (PLLA-Ac) (400 MHz, Chloroform-d, 298 K). f Comparison of the GPC traces of PLLA-Ac before and after depolymerization for 2 h catalyzed by (S,S)-Al-3, using THF as the eluent (flow rate: 1 mL/min, at 40 °C), and the sample concentration was 1 mg/mL.

Research efforts further focus on elucidating how the catalyst activates the chain-end to achieve stereoselective catalytic transformation. It was hypothesized that the Al metal center would first react with the terminal hydroxy group to form an alkoxy-aluminum species, while the chiral cavity constructed by the ligand differentially modulates the backbiting barriers for chain ends of different chiralities. To verify this mechanism, experimental verification and theoretical calculation were conducted.

Firstly, the structure of the catalysts bearing dual ligands needed clarification to understand its catalytic process. We therefore cultivated single crystals of catalyst Al-4 and successfully resolved its crystal structure. As shown in Fig. 5, the (S)-Al-4 complex crystallizes in the enantiopure orthorhombic space group P2₁2₁2₁, with a Flack parameter of 0.071(18), confirming its chiral integrity. The central Al(III) ion adopts a distorted octahedral geometry coordinated by three oxygen atoms (O1–O3, Al-O: 1.809–1.843 Å) and three nitrogen atoms (N1–N3, Al-N: 2.047–2.073 Å) from two distinct ligands. The bond lengths and bond angles both conform to the characteristics of typical organometallic complexes^24,51. Critically, the binaphthyl scaffold (bonded via N1 and N2) adopts a helical twist with a dihedral angle of 68.5(2)°, while the two strategically positioned salicylaldimine units (linked through N1/O1 and N3/O3) project their phenolate oxygen atoms toward each other with an O···O separation of 6.73 Å, forming a rigid C₂-symmetric chiral pocket. Crystallographic analysis revealed that the chirality of the ligand and the six-coordinate distorted octahedral geometry around the metal center are pivotal in constructing the chiral cavity. Notably, a single metal Al center could coordinate only three oxygen atoms simultaneously. Thus, for catalysts Al-3 and Al-7, one additional equivalent of the salicylaldimine unit may act as a bridging structure, promoting the aggregation of multiple Al centers. This structural arrangement further regulates the chiral confinement space around the metal centers, enhancing stereoselective control.

Fig. 5. X-ray structure of complex (S)-Al-4.

Hydrogen atoms have been omitted for clarity. Selected bond lengths (in Å) and bond angles (in °): Al1-O1 1.809(4), Al1-O2 1.843(4), Al1-O3 1.833(4), Al1-N1 2.047(5), Al1-N2 2.063(5), Al1-N3 2.073(5); O1-Al1-O2 92.96(19), O1-Al1-O3 93.07(19), O1-Al1-N1 90.01(19), O1-Al1-N2 174.78(19), O1-Al1-N3 91.9(2), O2-Al1-N1 89.66(17), O2-Al1-N2 86.97(19), O2-Al1-N3 88.60(18), O3-Al1-O1 173.8(2), O3-Al1-N1 91.58(18), O3-Al1-N2 87.14(19), O3-Al1-N3 89.96(19), N1-Al1-N2 84.78(18), N1-Al1-N3 177.5(2), N2-Al1-N3 93.29(19).

To further elucidate the nature of the active species during catalysis, the influence of catalyst concentration on the reaction activity was investigated. The depolymerization catalyzed by (R,R)-Al-3 were first investigated with increasing catalyst concentration. As shown in Supplementary Fig. S52, the apparent rate constants (k_obs) at different catalyst concentrations were obtained from reaction kinetics. As expected, the reaction rate accelerated with increasing catalyst concentration. The logarithmic values of the catalyst concentration [Cat.]_tot and k_obs were further calculated, as summarized in Supplementary Table S36. By plotting ln (k_obs) versus ln([Cat.]_tot), an order in total catalyst of 0.66 was discovered (Supplementary Fig. S53). Using the same research approach, the influence of catalyst concentration on the reaction rate was also explored for (R)-Al-4 (Supplementary Fig. S54 and Supplementary Table S37), and an order in total catalyst of 1.11 was calculated (Supplementary Fig. S55). As reported in the literature, if the catalyst operates via a well-behaved monomolecular catalytic process, the kinetic order to the catalyst should be 1. In contrast, if the catalyst exists as a bimetallic dimer in the ground state, but the actual catalysis proceeds through a mono-metallic transition state, implying a dissociation process of metal aggregates, the kinetic order relative to the catalyst would be 0.5^52,53. Therefore, in our experimental study, the order close to 1 for (R)-Al-4 suggests a well-defined monometallic catalytic process. On the other hand, the order for (R,R)-Al-3 is closer to 0.5, indicating that the catalyst exists in an aggregated state at the metal centers, and a dissociation process occurs during the actual catalysis, further supporting our hypothesis regarding the catalyst structure²².

Further investigation via in situ NMR spectroscopy employed a lactic acid dimer (dimer-OH) as a simplified model substrate to probe the catalyst-polymer terminal hydroxy group interaction mechanism. Reactions were conducted in deuterated MeCN to accurately simulate actual catalytic conditions. As shown in Fig. 6, when (R)-Al-2 was reacted with both enantiomers of dimer-OH in a 1:1 molar ratio, it was observed that the methyl signal of Al-2 disappeared, and the hydroxy peak of dimer-OH vanished with concurrent chemical shift variation. This result provides definitive evidence for methyl/hydroxy exchange, forming alkoxy-Al species. Crucially, distinct NMR profiles emerged for (R)-Al-2 complexes with two dimer-OH enantiomers ((R,R)-dimer-OH and (S,S)-dimer-OH), indicating differential coordination modes and activation levels based on chirality matching. The significant LA formation confirmed backbiting cyclization, while enantiomeric divergence in LA yield demonstrated stereoselective backbiting. Further investigations were conducted on the interaction between (S,S)-Al-3 and dimer-OH. NMR analysis revealed partial conversion to alkoxy-aluminum activated species, while significant amounts of free dimer-OH remained in the system. This observation may be attributed to the enhanced stability of the Al-O bond to Al-C, making it more difficult to be replaced by hydroxyl groups. Notably, the chirality-matched S-configured dimer-OH with (S,S)-Al-3 exhibited substantial LA formation, whereas the mismatched R-configured counterpart showed nearly undetectable lactide production. This striking contrast unambiguously demonstrates the highly stereoselective backbiting cyclization enabled by Al-3 catalysis.

Fig. 6. Analysis of the interaction between dimer-OH and catalyst.

A Analysis of the interaction between dimer-OH and (R)-Al-2. ¹H NMR for (a) (S)-dimer-OH + (R)-Al-2, (b) (R)-dimer-OH + (R)-Al-2, (c) (R)-Al-2, (d) dimer-OH (400 MHz, Chloroform-d, 298 K). B Analysis of the interaction between dimer-OH and (S,S)-Al-3. ¹H NMR for (a) (S)-dimer-OH + (S,S)-Al-3, (b) (R)-dimer-OH + (S,S)-Al-3, (c) dimer-OH (400 MHz, Chloroform-d, 298 K).

To further probe the stereoselective depolymerization mechanism and energetic landscape, detailed Gibbs free-energy profiles were computed by density functional theory (DFT) calculations (Fig. 7). The lactate trimers with RRR and SSS configurations were selected as model substrates to simulate PLA chains for simplified computation (Fig. 7a, -P is simplified as -Me). The (R,R)-Al-3 was constructed referred to the lowest energy structure based on the crystal structure of the ligand from experimental work^22,23. In the first step, the hydroxyl end of the lactate trimer is bound to the metal aluminum center of the (R,R)-Al-3 to form (R,R)-Al-3-RRR and (R,R)-Al-3-SSS. This step is extremely exothermic. As the ensemble (R,R)-Al-3-RRR is used as a zero reference, and the ΔG for the RRR lactate trimers binding process is calculated to be − 37.8 kcal·mol⁻¹. For (R,R)-Al-3-SSS formed from SSS lactate trimers substrate, its relative Gibbs free energy is located at 7.2 kcal·mol⁻¹. The lower Gibbs free energy of (R,R)-Al-3-RRR indicates a preference for the binding of (R,R)-Al-3 with the R-configuration chain-end.

Fig. 7. DFT calculations for the stereoselective depolymerization of PLA catalyzed by (R,R)-Al-3.

a DFT-calculated free-energy paths, intermediates (INT) and transition states (TS) in the stereoselective depolymerization of PLA (-P is simplified as -Me), SSS represents PLLA, RRR represents PDLA. b Calculated free energy of each INT and TS.

Next, the carbonyl group coordinates with the Lewis acidic Al center to formed TS1. An energy barrier difference is observed, giving the RRR substrate activation energy of 11.9 kcal·mol⁻¹ (TS1_RRR) and the SSS substrate activation energy of 13.6 kcal·mol⁻¹ (TS1_SSS) compared with the preceding state. Subsequently, the back-biting step is followed from the activated hydroxyl end to the carbonyl group. After experiencing the intermediates INT1 and INT2, it evolves to the transition state TS2. At this stage, the two substrates display a huge energy barrier difference. For TS2_RRR, the Gibbs free energy barrier is only 2.1 kcal·mol⁻¹, while for TS2_SSS, an activation energy of 14.0 kcal·mol⁻¹ is calculated. The apparent energy barrier difference (11.9 kcal·mol⁻¹) in the intramolecular cyclization stage is the origin of high stereoselectivity. After the ring-closing reaction, it releases one molecule of lactide (LA), thereby forging a new intermediate INT3, which enters the next catalytic cycle.

After performing DFT calculations, it is evident that the BisSalen-Al catalyst exhibits stereoselective recognition towards the chiral chain-end of PLA. The substantial energy barrier difference observed in the intramolecular back-biting reaction catalyzed by the BisSalen-Al catalyst is the key in achieving stereoselective ring-closing depolymerization for PLA. The above DFT calculations deepen the understanding of the stereoselective depolymerization mechanism.

Stereoselective depolymerization of PLLA and PDLA blends

After demonstrating that this dual-ligand strategy enables the construction of confined chiral cavities for distinguishing PLLA and PDLA, its capability to recognize specific chiral PLA within PLLA and PDLA blends and catalyze kinetic resolution depolymerization (KRDP) was verified. The study initially focused on an equimolar mixture of PLLA and PDLA, which usually forms stereocomplexed PLA (SC-PLA) with enhanced heat resistance and crystallinity compared to homocrystalline structure⁵⁴. To confirm the feasibility of this KRDP of SC-PLA, the (R,R)-Al-3 was firstly employed to catalyze the depolymerization process (Fig. 8a). By monitoring the depolymerization conversion, an apparent inflection point at around 40% conversion was revealed, where the depolymerization rate slowed down from fast to slow and became almost stagnant at around 50% conversion (Fig. 8b and Supplementary Table S38). This phenomenon indicated that the (R,R)-Al-3 may display excellent stereoselectivity for different configurations of PLA in SC-PLA. Meanwhile, the HPLC analysis was used to track the enantiomeric excess (ee) of the obtained LA. It was found that the recovered LA was predominantly D-LA with an ee value exceeding 90% (Fig. 8c). For example, after 2 h, the conversion reached 45%, in which the ee value of LA was 91%. Therefore, it could be concluded that the (R,R)-Al-3 highly stereoselective recognized and depolymerized PDLA in SC-PLA, and thus D-LA was formed. While PLLA was almost not selected and depolymerized during this process, suggesting the excellent stereoselectivity of BisSalen-Al in SC-PLA.

Fig. 8. Stereoselective depolymerization of stereocomplexed PLA.

a Schematic diagram of stereoselective depolymerization of stereocomplexed PLA. b The depolymerization conversion versus time for SC-PLA catalyzed by (R,R)-Al-3. c ee value and meso-LA ratio of the recovered LA at different conversions. d P_m value for the PLA at different conversions. e Differential scanning calorimetry (DSC) of SC-PLA, PLLA, and SC-PLA after depolymerization under first heating curves, T_m is the melting point. f WAXD of SC-PLA, PLLA, and SC-PLA after depolymerization, SC represents β-form stereocomplex crystallites, HC represents α-form homo-crystallites. g HPLC plots of recovered LA for SC-PLA catalyzed by (R,R)-Al-3 (Conv. = 46%), (S,S)-Al-3 (Conv. = 48%), and Rac-Al-3 (Conv. = 94%).

In addition, analyzing the remaining PLA during the SC-PLA depolymerization process could aid in better understanding this KRDP process. Through the homonuclear decoupled ¹H NMR spectra of the PLA continuously monitored during the depolymerization, it was observed that the P_m (the probability of meso linkages between monomer units in polymer) remained above 0.99 throughout the whole depolymerization process (Fig. 8d and Supplementary Fig. S71). Moreover, the reaction was terminated when the depolymerization conversion plateaued at approximately 50%. The recovered lactide monomers and unreacted PLA were separated via precipitation by cooling the mixture in MeCN solvent. The precipitate (unreacted PLLA) was washed with cold MeCN, then dried to obtain pure PLLA with an isolate yield of 94%. The liquid supernatant was concentrated and then sublimation to obtain pure L-LA with an isolate yield of 82% (Supplementary Fig. S82, 83). The M_n of the unreacted PLLA was tested as 11.9 kg·mol⁻¹, with a narrow dispersity of 1.15 (Supplementary Fig. S84). The subtle changes in molecular weight and dispersity showed that the non-preferred PLLA was well preserved, highlighting the high stereoselectivity and controllability of the depolymerization process. Besides, the thermal properties tests of the PLA samples were conducted by differential scanning calorimetry (DSC) (Fig. 8e). For the SC-PLA, the melting point (T_m) reached 219.9 °C due to the stereocomplex crystallites. Whereas the T_m of the PLA obtained after depolymerization was 166.4 °C, which is consistent with the T_m of commercial PLLA (160-170 °C)^55,56. The difference in T_m further indicated that the specific depolymerization of PDLA in SC-PLA was achieved. Furthermore, the crystallinity of the PLA was explored through the wide-angle X-ray diffraction (WAXD) spectra (Fig. 8f). For SC-PLA, the diffraction peaks at 11.9°, 20.7° and 23.9°, which were assigned to the diffractions from (110), (300)/(030) and (220) planes of β-form stereocomplex crystallites (SC), respectively⁵⁷. While, after depolymerization, these diffraction peaks were not observed, but peaks at 14.8°, 16.8°, 19.1° and 22.3° were instead appeared, which could be correlated to (010), (110)/(200), (203) and (015) crystal planes of α-form homo-crystallites (HC), respectively. All the above characterization results of residual PLA conclusively demonstrate that this catalyst enables specific recognition of chiral polymers within SC-PLA and facilitates highly stereoselective depolymerization.

Subsequently, to further investigate the generality of the stereoselective depolymerization of BisSalen-Al, (S,S)-Al-3 was also investigated (Supplementary Table S39, entry 1). As shown in Fig. 8g, a conversion of 48% was obtained after 2.5 h, with an 86.6% ee of L-LA was detected through HPLC analysis, showing the high PLLA selectivity of (S,S)-Al-3 in the stereoselective depolymerization of SC-PLA. Notably, the recovered highly optically pure L-LA was introduced into repolymerization using (S,S)-Al-3 as a catalyst (Supplementary Table S40). Under optimized polymerization conditions (80 °C, [M]₀ = 1 M in toluene) with diphenylmethanol as the initiator ([M]₀: [Cat.]₀: [OH]₀ = 100: 1: 1), 93% monomer conversion was achieved after 2.5 h. The resulting polymer exhibited a molecular weight consistent with the theoretical value of 11.1 kg·mol⁻¹ with a narrow dispersity (Đ = 1.11, Supplementary Fig. S91). Further analysis demonstrated that the polymer was highly isotactic (P_m = 0.99, Supplementary Fig. S92), while DSC analysis revealed a melting point of 165 °C (Fig. S93). These results confirm that recycled LA monomers can be reused to synthesize highly stereoregular PLA.

Besides, when a racemic catalyst consisting of a 1:1 mixture of (R,R)-Al-3 and (S,S)-Al-3 was used, it was found that the conversion reached 94% after 4 h (Supplementary Table S39, entry 2), which means (S,S)-Al-3 and (R,R)-Al-3 depolymerized PLLA and PDLA, respectively. In addition, SC-PLA with higher M_n (M_n = 47.9 kg·mol⁻¹, Đ = 1.15) was introduced into depolymerization as well (Supplementary Table S39, entry 3). After 8 h, the (R,R)-Al-3 catalyzed depolymerization conversion reached 50% with an ee value for recovered D-LA of 86.2%. The consistent selectivity indicated the molecular weight independence for KRDP of SC-PLA.

Apart from the SC-PLA, the stereoselective depolymerization of diblock PLA also attracted our attention. The PLLA-b-PDLA with R-configured chiral chain-end and PDLA-b-PLLA with S-configured chiral chain-end were prepared and introduced into depolymerization. At first, (R,R)-Al-3 was employed to catalyze the depolymerization of PLLA-b-PDLA (Fig. 9a). Referring to the conversion, the depolymerization proceeded smoothly before 50% conversion, while showed significant stagnation afterwards, indicating that the (R,R)-Al-3 displayed highly efficient identification of the R-configured chain-ends and achieved catalytic depolymerization of the PDLA block (Fig. 9b and Supplementary Table S41). Similarly, the ee of the recovered LA was analyzed through HPLC (Fig. 9c). It was observed that the obtained D-LA consistently exhibited a high ee value. For example, at a reaction time of 1 h and a conversion of 36%, the ee value of D-LA was 97.8%. After 3 h, the conversion reached 45% and an ee value of 91.8% was detected. Notably, the highly stereoselective kinetic resolution depolymerization for the diblock PLA catalyzed by the BisSalen-Al catalyst was intuitively presented. Moreover, from the perspective of molecular weight change, the M_n exhibited a linear decrease trend before 50% conversion, consistent with the controlled depolymerization process as well (Fig. 9d).

Fig. 9. Stereoselective depolymerization of diblock PLA.

a Schematic diagram of the stereoselective depolymerization of diblock PLA. b The depolymerization conversion versus time for PLLA-b-PDLA catalyzed by (R,R)-Al-3. c ee value and meso-LA ratio of the recovered LA at different conversions. d The molecular weight (M_n) and dispersity (Đ) versus conversion for depolymerization of PLLA-b-PDLA.

Furthermore, the depolymerization of PDLA-b-PLLA catalyzed by (R,R)-Al-3 was also considered. In this case, the depolymerization became significantly stagnant, with a depolymerization conversion of only 4% after 1 h and 6% after 2 h (Supplementary Table S42). This phenomenon further revealed that the depolymerization initiation originated from the chiral matching of the catalyst and the polymer chain-end. When a non-preferred chiral configuration occupies the polymer chain-end, the mismatched mode made the depolymerization difficult to proceed. These results provide convincing evidence for the stereoselectivity of BisSalen-Al complex.

To verify the applicability of the catalytic system, the depolymerization of commercially available PLA cups was also investigated. The commercial-grade PLA cups used in this study had a lactic acid unit composition of 96.4% L-configured units, and the polymer had a molecular weight of 68.2 kg/mol with a Đ of 1.95⁵⁸. Experimental results demonstrated a significant difference in catalytic activity depending on the catalyst stereochemistry (Supplementary Table S43). When the depolymerization was catalyzed by (S,S)-Al-3 (matched with the prevailing L-configuration of the polymer), an 85% conversion was achieved within 8 h. In contrast, using (R,R)-Al-3 resulted in a markedly lower conversion of only 9% in 8 h. This disparity in reactivity clearly indicates that our catalytic system maintains high stereoselectivity for commercial-grade PLA product. This finding suggests that the catalytic system exhibits a degree of tolerance towards additives present in real-world polymer materials, thereby expanding its substrate adaptability.

Discussion

In summary, to the best of our knowledge, this work presents the case of chiral recognition and catalytic transformation of macromolecules using a synthetic catalyst. The BisSalen-Al catalysts, featuring a confined chiral cavity constructed via a dual-ligand strategy, selectively recognize stereochemically matched PLA and catalyze its conversion to chiral lactide monomers, while remaining almost unresponsive to mismatched PLA. The stereoselective depolymerization mechanism was revealed as uan nzipping back-biting mechanism, in which the chiral BisSalen-Al catalyst stereoselectively activates the chiral chain-end of PLA to trigger the depolymerization. The DFT calculation indicates the huge difference in the energy barriers of the intramolecular back-biting process. Benefiting from the high stereoselectivity, this strategy has been further extended to the kinetic resolution depolymerization of stereocomplexed and diblock PLA to obtain high optical purity lactide and remaining PLA homo-crystallites. This study extends asymmetric catalysis to the field of polymer enantioselective transformation, promoting the development of polymer and chiral science. Achieving the practical application of the reaction process represents an important goal for future research.

Methods

General procedure for the depolymerization of PLA:

A dry 5 mL pressure-resistant tube was first introduced into an Argon-filled glovebox. (S,S)-Al-3 (10 mg, 10 μmol, 1 equiv.) was added and dissolved in MeCN (1 mL). Then, PLLA-100 (36 mg, 0.5 mmol relative to repeating unit, 50 equiv.) was added. Seal the reaction tube, then depolymerization was conducted at 140 °C in an oil bath. At a specific time, remove the reaction system from the oil bath and quickly cool down, then expose it to air to quench the reaction. The depolymerization conversion was determined and calculated by ¹H NMR spectrum of the crude product in CDCl₃ by comparing the integral areas of the methine groups of LA and PLA. Subsequently, after separating the monomer from the unreacted polymer and catalyst using a silica gel plate, chiral HPLC analysis was performed to determine the relative contents of L- and D-configured lactide.

Author contributions

R.Y., G.X., and Q.W. conceived the work and designed the experiments. R.Y., W.W., and Y.J. performed the laboratory experiments. X.G. checked the experiment data. R.Y., G.X., X.G., W.W., Y.J., and Q.W. analyzed the data and co-wrote the manuscript.

Peer review

Peer review information

Nature Communications thanks Nagaraju Barsu, and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

Details on the procedures, optimization and characterization are available in Supplementary Information. Crystallographic data for compound (S)-Al-4 are available in the Supplementary Data 1 and from the Cambridge Crystallographic Data Center under deposition number CCDC 2481918. Copies of the data can be obtained free of charge via www.ccdc.cam.ac.uk/data_request/cif. Cartesian coordinates (Å) for the computed structures are provided as the xyz data in Supplementary Data 2. Source Data are provided with this paper. The crystallographic data, cartesian coordinates and source Data generated in this study have also been deposited in the Figshare database via 10.6084/m9.figshare.31249927. All data are available from the corresponding authors upon request. Source data are provided in this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-70164-1.