Computational and experimental investigations of one-step conversion of poly(carbonate)s into value-added poly(aryl ether sulfone)s

Polymer Formation.

We report a procedurally simple cascade reaction sequence involving depolymerization followed by polycondensation for the in situ synthesis of value-added, engineering thermoplastics. PCs were depolymerized and converted into PSUs in a single step. PSUs are high-T_g, high-performance thermoplastics traditionally formed by the conversion of bisphenols into more reactive metal phenoxides (or other surrogates) that undergo S_NAr condensation with aryl fluorides proceeding through Meisenhemer intermediates (14, 15). Thus, poly(aryl ether sulfone)s have been produced from the reactions of aryl halides with bisphenols in the presence of stoichiometric amounts of metallic bases, or with silyl-protected bisphenols in the presence of metallic fluoride catalysts or organocatalysts (Fig. 1A) (15⇓–17). Our approach involves the in situ generation of phenoxides from depolymerized poly(carbonate) waste and conversion of these intermediates into poly(aryl ether sulfone)s (Fig. 1B).

The depolymerization of poly[bisphenol A (BPA) carbonate] to liberate phenolic end groups and subsequent electrophilic trapping with 4,4′-difluorodiphenylsulfone allowed access to PSU in a one-step procedure, with loss of carbon dioxide as the byproduct. The depolymerization of PC to PSU proceeds cleanly (¹H NMR spectra in Fig. 2, Inset) in this timeframe and produces a polymer with a single T_g and molecular mass comparable with starting material [gel permeation chromatography (GPC) traces are shown in Fig. 2 and the SI Appendix]. Notably, the T_g of PSU is ∼30 °C higher than that of PC [T_g(PC) ∼ 151 °C; T_g(PSU) ∼ 180 °C]. This approach uses a cascade reaction involving the depolymerization of PC by K₂CO₃ and formation of reactive phenoxides, which are then polycondensed in situ with 4,4′-difluorodiphenylsulfone. The reaction is complete within 18 h at 190 °C in N-cyclohexyl-2-pyrrolidone. Crucially, the stringent requirements for step growth polymerizations are met under these conditions, which is a surprising and unexpected result; from a mechanistic standpoint, this transformation represents the first example, to our knowledge, of a macromolecule that is used as a monomer source for polycondensation.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

GPC traces for PC (red) before in situ depolymerization by K₂CO₃ in the presence of 4,4'-sulfonylbis(fluorobenzene) and PSU (blue) after reaction completion. (Inset) ¹H NMR of PC before depolymerization (upper trace) and PSU after repolymerization (lower trace). CHP, N-cyclohexyl-2-pyrrolidone.

Under the same conditions but when the monomer BPA is used instead of the polymer PC, the M_n (number average molar mass) of the isolated PSU is ∼42.5 kDa, which suggests that the kinetics of the reaction with PC is slower, which is to be expected given that PC must be depolymerized before polycondensation. The relative ratio of free OH end groups to F was ∼1:2.6, consistent with observations for slow polycondensations that are subject to backbiting and formation of cyclic PSU side products during the polymerization (17). When the scale of the polymerization was increased, the average molecular mass increased by ∼3 kDa (Table 1), likely because of improvement in massing accuracy during experimental setup. To rule out effects from differences in retention volumes on molecular mass when measured against polystyrene standards, the PC starting material and PSU product were analyzed by GPC coupled with a light scattering detector. The PSU produced by this method had M_n = 27 kDa, whereas the commercial PC starting material had an M_n = 20 kDa (18).

The crude reaction mixture was analyzed after first depolymerizing PC with 1 eq K₂CO₃ at 190 °C for 18 h followed by addition of 4,4′-difluorodiphenylsulfone in N-methyl pyrrolidone (NMP) and continued heating for an additional 18 h. PSU oligomers were formed in low molecular masses (∼1 kDa and below); however, 4,4′-difluorodiphenylsulfone was completely consumed (details in SI Appendix). We attribute the low molecular masses observed in this regime to the decomposition of solvent, which hampers the kinetics of polycondensation. In addition, when PC was heated in the absence of the aryl fluoride and the carbonate salt, no observable changes in PC signals were observed by ¹H NMR analysis on the sample in d₆-DMSO within 3.5 h of heating at 185 °C (details in SI Appendix), which suggests that the carbonate salt is required for PC decomposition. However, on the addition of Li₂CO₃ (which exhibits some solubility in DMSO), ¹H NMR signals corresponding to BPA (∼6.7 and 7.0 ppm) were observed within 40 min of heating (∼5% conversion), and nearly complete depolymerization to BPA was observed after 19.5 h (∼88% conversion) (Fig. 3). In addition to depolymerization of PC by carbonate salts, fluoride decomposition was also possible. When PC was treated with LiF, PC was depolymerized to a mixture of oligomers and BPA within the experimental timeframe (details in SI Appendix), suggesting that fluoride salts, the elimination byproduct of the S_NAr step, are capable of depolymerizing PC. We hypothesize that PC decomposition by fluoride is caused by nucleophilic attack of fluoride on carbonate, leading to formation of carbonofluoric acid and carbonofluoric ester intermediates that can, in turn, react with water or hydroxide to form gaseous CO₂ (details in SI Appendix).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

¹H NMR study of depolymerization of PC by stoichiometric Li₂CO₃ (d₆-DMSO; 185 °C).

These results prompted us to use substoichiometric quantities of K₂CO₃ as an initiator given the ability of the LiF byproduct to also depolymerize PC. Thus, the depolymerization/polycondensation reaction was attempted with catalytic K₂CO₃ (12 mol %). Under catalytic conditions, ∼30% of 4,4′-difluorosulfone reacted to form ether linkages within 24 h, which is unsurprising given that two phenoxide end groups are formed on cleavage of the carbonate linker (vide infra) (details in SI Appendix). Fewer than 2% of the signals corresponded to phenoxides derived from unreacted BPA or PC oligomers, indicating that ether formation is dependent on the initial depolymerization of PC and that phenoxide-terminated products, which may be oligomeric, are instantly consumed through S_NAr with the aryl fluoride. Although PSU signals were not detected under these conditions, multiple signals characteristic of gem-dimethyl groups were observed by ¹H NMR. Differential scanning calorimetry analysis on the product, which showed a single T_g at ∼100 °C, suggested that PC decomposition was incomplete and formed miscible oligomers with varied molecular masses that react with the aryl fluoride. Furthermore, the molecular mass of the product was diminished from the initial molecular mass of PC (M_n decreased from ∼12 to ∼4 kDa by GPC) (details in SI Appendix). The isolation of this intermediate suggests that the reaction of PC oligomers with 4,4′-difluorodiphenylsulfone was kinetically favored over depolymerization and that PSU formed after repeated depolymerizations and condensations over 18 h.

It is noteworthy that, in this reaction, K₂CO₃ not only depolymerizes PC but is also itself degraded by the depolymerization process and serves as a catalyst for polyether formation in the subsequent step. To our knowledge, the methodology presented herein represents the first example of complete conversion of a step growth polymer into a dissimilar step growth polymer through in situ polymerization.

Finally, to assess the ability to transform materials comprised of PC into PSU with this method, a compact disk was exposed to the depolymerization/repolymerization reaction conditions in the presence of the aryl fluoride and K₂CO₃ (Fig. 4). Conversion to PSU in this fashion was equally as efficient and produced polymers with comparable molecular masses as when purified PC pellets were used as starting materials [M_n (PSU) ∼ 11 kDa]. We envision that this method could also be applied to baby bottles formulated from poly(BPA carbonate) or other materials in a one-pot fashion.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Repurposing a compact disk as a monomer source for the depolymerization of PC and repolymerization to PSU. Compact disk (left) and PSU powder (right) are shown before and after the depolymerization/repolymerization process.

Computational Results.

Computational investigations were performed di-p-tolyl carbonate (PTC) and 1-fluoro-4-(methylsulfonyl)benzene (FSB) as computational models for PC and the bis(aryl fluoride) used in experiments, respectively (Fig. 5). The model reaction requires the use of 1 eq each PTC and sodium carbonate and 2 eq FSB and results in the formation of 2 eq each ether, gaseous CO₂, and sodium fluoride. These investigations involved the use of B3LYP-D3/aug-cc-pVTZ//6–311+G(2d,p) (19⇓⇓⇓⇓⇓⇓–26) in implicit DMA solvent with the solvation model based on density (SMD) (27) to determine the mechanism and energetics for the depolymerization of PC by Na₂CO₃ and subsequent polyether formation (Figs. 6 and 7 and details in SI Appendix).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Computational models for ether formation from the reaction of PTC with FSB in the presence of Na₂CO₃.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 6.

Stationary points and free energies for ether formation from the reaction of PTC with FSB in the presence of Na₂CO₃.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 7.

Free energy surface for ether formation from the reaction of PTC with FSB in the presence of Na₂CO₃.

The mechanism begins with the formation of intermediate 1 (INT1), a complex that is 23 kcal/mol more stable than the reactants on the free energy surface and comprises Na₂CO₃ coordinated to PTC through a very favorable interaction between the sodium atoms of Na₂CO₃ and the oxygen atoms of the organic carbonate. The free energy barrier for nucleophilic attack by Na₂CO₃ on PTC in transition state 1 (TS1) is 17 kcal/mol with respect to INT1. TS1 leads to the formation of an oxodimethoxy carbonate intermediate, INT2, with a tetrahedral carbon attached to four oxygen atoms.

INT2 is energetically unstable, is in a shallow minimum, and quickly decomposes to INT3; TS2, the transition state for the decomposition of INT2, possesses a free energy barrier of only ∼5 kcal/mol with respect to INT2 (∼20 kcal/mol with respect to INT1). Two concomitant processes occur in TS2: breakage of the C2–O4 bond, which leads to CO₂ formation, and breakage of the C5–O6 bond, which converts the tetrahedral C(sp³)–O4 into a trigonal C(sp²)–O3 (Fig. 6). The intermediate resulting from CO₂ formation, INT3, is ∼47 kcal/mol more stable than the reactants; it comprises a covalently bonded disodium complexed to the CO₂ molecule, cresolate, and PTC.

Dissociation of CO₂ from INT3 results in the formation of INT4, which is ∼11 kcal/mol more stable than INT3. The exergonicity for CO₂ dissociation from INT3 is likely caused by the entropy gained by dissociation of the gaseous CO₂ molecule from the solvated complex. The second molecule of CO₂ is formed from INT4 in TS3 and results in the formation of INT5, in which the sodium cresolate dimer is bound to the CO₂ molecule. TS3 possesses a free energy barrier of ∼26 kcal/mol with respect to INT4, and INT5 is ∼21 kcal/mol less stable than INT4. Formation of the second molecule of CO₂ from the second molecule is notably less favorable than formation of the first molecule, presumably because the first molecule of CO₂ is formed from an energetically unstable oxodimethoxy carbonate comprising a tetrahedral carbon atom bonded to four oxygen atoms in INT2. By contrast, the second molecule of CO₂ is formed from a comparatively more stable ionic carbonate, which is less likely to release carbon dioxide. Dissociation of the second molecule of CO₂ from INT5 results in the exergonic formation of the sodium cresolate dimer, INT6, which is ∼36 kcal/mol more stable than INT5 (∼24 kcal/mol below INT4). As before, this observed exergonicity is likely caused by the entropy gain from gaseous CO₂ dissociation from the solvated complex.

We next considered whether formation of the initial ether molecule proceeds directly from the sodium cresolate dimer, INT6 (i.e., whether the dimer dissociates and forms sodium cresolate before reacting with an aryl fluoride present in the reaction or whether INT6 is directly involved in ether formation). The former scenario could be ruled out because of the fact that formation of two molecules of sodium cresolate from INT6 is endergonic (41 kcal/mol above INT6), presumably because of the presence of the strong disodium bond. In contrast, formation of INT7 by the coordination of the sodium cresolate dimer with FSB is exergonic by about 17 kcal/mol with respect to INT6. The transition state for ether formation proceeding directly from the covalently bound sodium cresolate dimer, TS4, possesses a free energy barrier of ∼30 kcal/mol with respect to INT7 and forms INT8, in which the ether is bound to a datively coordinated sodium fluoride–sodium cresolate complex. Dissociation of ether from INT8 is endergonic by about 10 kcal/mol owing to the favorable binding of sodium with oxygen; dissociation leads to the formation of INT9, a datively coordinated sodium fluoride–sodium cresolate complex. Coordination of the second molecule of FSB to INT9 produces INT10, which is ∼21 kcal/mol more stable than INT9. Notably, there is a net free energy gain after an incoming FSB molecule coordinates to the sodium fluoride–sodium cresolate complex formed after ether dissociation; this free energy gain mitigates the fact that ether dissociation from the sodium fluoride–sodium cresolate complex is an endergonic process.

The second ether molecule is formed from INT10, which proceeds through TS5 to produce INT11, a complex comprising the ether molecule bound to two molecules of sodium fluoride. The free energy barrier for formation of the second ether molecule in TS5 is ∼28 kcal/mol with respect to INT10. The release of the second ether molecule from INT11 is expected to be slightly endergonic (similar to the release of the first ether molecule from INT8 to form INT9) because of the favorable binding of sodium to the oxygen belonging to the ether, but in general, the free energy profile is exergonic starting from the reactants.

Overall, these calculations signify that the decomposition of organic carbonates by carbonate salts into phenolate dimers is a facile process. In addition, these phenolate dimers are competent substrates for ether formation from reactions involving aryl fluorides.

In conclusion, we have shown formation of high-value PSUs from the in situ depolymerization of PCs in a single operation in the presence of a carbonate salt and bis(aryl fluoride)s. This strategy has important implications for the formation of value-added polymeric materials from plastic waste (for example, recycling). Given the stringent criteria for step growth polymerization, including the absence of side reactions, perfect stoichiometry, and quantitative conversion of monomer to polymer, the PSU formation shown herein attests to the robustness of the depolymerization and repolymerization process. Computational studies used to support these findings show that carbonate salts decompose organic carbonates through initial nucleophilic attack followed by the loss of CO₂ and formation of phenoxide dimers and subsequently, ether products. We envision that careful control over depolymerization reactions to form materials by coupling computation with experiments will ultimately enable repurposing of spent polymeric materials.