Research ArticleChemistry

Coordination-induced polymerization of P═C bonds leads to regular (P─C)n polycarbophosphanes

See allHide authors and affiliations

Science Advances  10 Mar 2021:
Vol. 7, no. 11, eabf4272
DOI: 10.1126/sciadv.abf4272


The replacement of carbon in (C─C)n chains of polyolefins by phosphorus leads to polycarbophosphanes (P─C)n, which may possess unique chemical and physical properties. However, macromolecules with a regular (P─C)n chain have never been unambiguously identified. Here, we demonstrate that addition polymerization, a general concept to polymerize olefins, can be extended to P═C double bonds. The polymerization of monomeric 2-phosphanaphthalenes is mediated by copper(I) halides and leads to polycarbophosphanes with an Mn of 14 to 34 kDa. Each phosphorus is coordinated to Cu(I), which can be easily removed. Unlike long-term durable polyolefins, the metal-free polymers depolymerize rapidly back to monomers under sunlight or ultraviolet irradiation at λ = 365 nm. The monomers can be recycled for repolymerization, demonstrating a cradle-to-cradle life cycle for polycarbophosphanes.


The polymerization of olefins was discovered in 1839 (1). Today, hundred million tons of polyolefins are produced every year using sophisticated processes that allow the controlled polymerizations of the C═C double bonds in alkenes (2). The incorporation of noncarbon elements into the polymer chains gives access to previously unidentified materials with unique physical and chemical properties (3, 4). Many monomers with a C═E bond, for example, E═N (5), O (6, 7), S (8), and Se (9), can be polymerized in a head-to-tail pattern (addition polymerization), usually with an initiator. This process, schematically shown in Fig. 1A, necessitates monomers (AB), which are sufficiently kinetically inert to allow isolation and purification. However, many doubly bonded species of noncarbon elements as potential monomers are highly reactive and short-lived. They require steric protection or electronic stabilization, for example, incorporation into a conjugated π system to make them stable and manageable, which, in turn, makes them reluctant to polymerize. Thus, many inorganic polymers such as polysiloxanes [─R2Si─O─]n, polyphosphazenes [─R2P═N─]n, or polyphosphates [─(RO)P(O)─O─]n are prepared by different methods such as polycondensation or ring opening polymerization, which do not require monomers of multiple bonds (3, 4, 10). The head-to-tail polymerization of phosphaalkenes, R2C═PR′, would allow the preparation of a polymer with a regular (P─C)n backbone but early attempts only led to ill-defined materials of “polymeric nature” (11). Later, in particular, Gates et al. (3, 12, 13) intensively investigated possibilities to polymerize phosphaalkenes. Several sterically stabilized phosphaalkenes (Fig. 1B) with mesityl (Mes) substituent at P and variable substituents R on carbon (R = phenyl, 9-phenanthryl, 4-methoxyphenyl, 1-naphthyl, ferrocenyl, pyrenyl, and oxazoline-derived groups) were polymerized or copolymerized. The initial assumption that these polymers contain a regular (P─C)n backbone had to be corrected (14, 15). Instead, a complex polymerization process that involves isomerization takes place, leading to polymers with an irregular microstructure as shown in Fig. 1C.

Fig. 1 Polymerization of double bonds.

(A) Schematic presentation of an addition polymerization of doubly bonded species AB using an initiator I*. (B) Previously used phosphaalkenes for polymerization. (C) Observed microstructure resulting from the polymerization of phosphaalkenes shown in (B).

The stability of hydrocarbon polymers and their massive and still increasing use causes severe environmental problems (2). A simple solution would be to mechanically recycle plastics, which, however, is not without problems (2, 16). Another strategy consists in chemically breaking down a polymer into its monomer units for reuse (17). Some poly(phthalaldehyde)s, polyesters, polycarbonates, polyurethanes, polythioesters, poly(enol ether)s, and polymers based on covalent diketoenamine bonds follow this “cradle-to-cradle” design (1826). Their degradation/depolymerization relies on thermolysis, catalytic ring-closing depolymerization, or bond exchange reactions that require the use of quantitative amounts of additional reagents. Polymers that unzip exclusively back to the monomers at low temperature or triggered by physical stimuli remain of great interest (27). Specifically, depolymerization under irradiation with light has great potential to tackle the problem of plastic pollution (28), which is an even more sustainable approach compared to the photodegradation of polymers (29).

Here, we report the successful polymerization of P═C bonds in 2-phosphanaphthalenes to regular (P─C)n chains promoted by copper(I) halides. Removal of Cu(I) allows the isolation of the free polymers that are stable as a solid. In solution, the polymers depolymerize exclusively to the monomers. This process is very slow at room temperature, but becomes faster at elevated temperatures, and is very fast under irradiation with sunlight or at λ = 365 nm. The obtained monomers can be fully reverted to polymers simply by adding copper(I) halides again without any further purification.



Sodium phosphaethynolate, Na(OCP) 2, became a versatile synthon (30) and can be used to synthesize sodium 2-phosphanaphthalene-3-olate 3 from commercially available phthalazine 1 (31). Upon reaction with various electrophiles, the neutral 2-phosphanaphthalenes M1 to M3 are produced in good yields (Fig. 2A).

Fig. 2 Synthesis and structures of monomers and their metal complexes.

(A) Synthesis of 2-phosphanaphthalenes M1 to M3. (B and D) Crystal structures of M3 and 4. (C and E) Synthesis of copper complexes 4 to 6.

The 31P chemical shift of the λ3-Pring nucleus in the ring is at relatively high frequency [M1: δ(31P) = 172.8 parts per million (ppm); M2: δ(31P) = 171.3 ppm; M3: δ(31P) = 158.9 ppm], which is typical for 2-phosphanaphathalenes (32). Phosphanaphthalene M3 was further characterized by single-crystal x-ray diffraction, and its crystal structure is shown in Fig. 2B. The P1─C9 bond length [1.70(2) Å] in M3 is in the typical range of P═C double bonds [∑rcov(C═P) = 1.69 Å] (33) and much shorter than the P1─C1 bond [1.78(1) Å]. This and all other structural features are comparable to the previously described ones for M1 (31). Calculations indicate that both the PC5 and C6 ring in M1 are aromatic (31).

Phosphanaphthalene M1 was reacted with CuCl in methylene chloride as a solvent to give a single product 4 (Fig. 2C), which is sparingly soluble and precipitated as a bright yellow powder from the reaction mixture (92% isolated yield). The structure of 4 was determined by single-crystal x-ray diffraction (Fig. 2D). In the solid state, complex 4, (M1CuCl)2, contains a Cu22-P)2 core, where each phosphanaphthalene ligand bridges two copper centers. The structural parameters of the coordinated phosphanaphthalene are almost identical to the ones in free M1 (table S1). Each P═O unit of the phosphoryl substituents coordinates via the oxygen center O2 to a copper center, which, as a result, resides in a distorted tetrahedral coordination sphere. Overall, the core of complex 4 consists of a tricyclic structure in a zigzag confirmation. The reaction of M2 or M3 with CuCl initially gives Cu phosphanaphthalene complexes as indicated by the 31P nuclear magnetic resonance (NMR) spectra, but they rapidly react further (vide infra). However, the Cu triphenyl phosphane precursors [CuX(PPh3)] (X = Cl, Br) lead to stable dinuclear copper complexes 5 and 6 with M2 and M3 as terminal η1-binding ligands, respectively (Fig. 2E). The halogen centers (X = Cl, Br) take μ2-bridging positions (figs. S122 to S123). Again, the structure metrics of the coordinated phosphanaphthalene rings are almost identical to the unbound ones (table S1).

The coordination shift [Δcoord = δ(31Pring)complex − δ(31Pring)unbound] allows differentiating between the terminal η1-binding and bridging μ2-binding mode of a phosphinine ligand. It also helps to decide whether the solid-state structure of a complex is retained when dissolved in solution (34). A large negative Δcoord indicates a bridging phosphinine, and a small negative Δcoord manifests a terminal one. The solid-state 31P MAS NMR of 4 shows a resonance at δ = 104 ppm for the λ3-Pring nucleus (Δcoord = −68 ppm), which is very different from the one in solution [δ(31Pring) = 141.6 ppm; Δcoord = −32 ppm]. This suggests that in solution, 4 converts to a species with a η1-phosphanaphthalene as seen in 5 or 6. The solid-state 31P MAS NMR spectrum of 5 shows a signal at 143 ppm (Δcoord = −28 ppm), while in solution, the 31Pring nuclei is detected at 162.1 ppm (Δcoord = −9 ppm). The small Δcoord indicates η1-bound phosphanaphthalene in both the solid and solution, but the different 31P chemical shifts infer that the solid-state structure of 5 is likely not retained either when dissolved. The hydrodynamic volume of 5 in solution was determined to be 268 Å3 in CD2Cl2 and 333 Å3 in CDCl3 by diffusion ordered NMR spectroscopy. This corresponds roughly to the half of the volume of 5 in the solid state (1013 Å3), which indicates that 5 dissociates into a mononuclear complex in solution. Typically, the resonances of the 31Pring nuclei are broadened by quadrupolar coupling to the directly bound 63Cu/65Cu nuclei (half-widths ν½ of 123 to 637 Hz; table S2).


Mixing phosphanaphthalene M2 or M3 with copper halides CuX (X = Cl, Br) in dichloromethane (DCM) or N,N′-dimethylformamide at room temperature rapidly gives yellow solutions, which show broadened 31Pring NMR signals [M2CuCl δ(31Pring) = 166.5 ppm, Δcoord = −5 ppm; M2CuBr δ(31Pring) = 169.2 ppm, Δcoord = −2 ppm; M3CuCl δ(31Pring) = 142.6 ppm, Δcoord = −16 ppm], indicating the formation of η1-phosphanaphthalene Cu complexes. The intensity of these signals decreases with time while simultaneously broad signals at significantly lower frequencies increase in intensity indicating the conversion of the P═C unit in M2 or M3 to P─C single bonds and the formation of polymers P2CuCl [δ(31Pring) = −30.9 ppm, ν½ = 714 Hz], P2CuBr [δ(31Pring) = −31.6 ppm, ν½ = 514 Hz], and P3CuCl [δ(31Pring) = −26.0 ppm, ν½ = 823 Hz]. Concomitantly, the 1H and 13C NMR signals of the relevant P,C units (table S3) are shifted to significantly lower frequencies likewise indicating the conversion of the double bonds to P─C single bonds. Full conversion is reached after some hours at room temperature and evaporation of the solvent. As an example, Fig. 3A displays the 31P NMR spectra of the phosphoryl substituted phosphanaphthalene M2 (top), the reaction mixture of M2 with a slight excess of CuCl (middle), and the pure product P2CuCl (bottom). The products P2CuCl, P2CuBr, and P3CuCl were obtained as brittle yellow solids in good yields (60 to 85%). Slow evaporation of a DCM solution of P2CuCl yielded a brittle yellow film.

Fig. 3 NMR spectra and mass analysis by matrix-assisted laser desorption/ionization (MALDI).

(A) 31P NMR spectra of M2 (top), the reaction mixture of M2 with CuCl in DCM (middle), and the pure P2CuCl (bottom). [The 31P NMR resonance for the PO(OEt)2 group is not significantly affected by complexation or polymerization.] (B) High-mass MALDI-MS spectrum of P2CuCl. m/z, mass/charge ratio.

That M2 and M3 undergo a coordination-induced polymerization is verified by gel permeation chromatography (GPC, figs. S102 to S105) and high-mass matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) (figs. S106 to S110) of the resulting polymers. Figure 3B shows, as an example, the high-mass MALDI-MS spectrum of the polymer P2CuCl. The number average molecular weight (Mn) of P2CuCl (Mn = 26.8 kDa), P2CuBr (Mn = 14.4 kDa), and P3CuCl (Mn = 34.2 kDa) indicates medium-size polymers with a size distribution in the range of Đ = 1.88 (P2CuBr) to 2.41 (P3CuCl) (Đ = polydispersity index). Two-dimensional (2D) NMR spectra allow us to unequivocally determine the alternating sequence of carbon and phosphorus nuclei (see the Supplementary Materials for details). Elemental analysis data indicate that the phosphorus of each repeating unit binds to one CuX unit.

The initial difficulty in achieving reproducible kinetic data led to the discovery that traces of water influence the rate of the polymerization. The deliberate addition of small amounts of water to a DCM solution of the copper complex M2CuCl significantly accelerates the polymerization and the conversion reaches 100% in 5 min at ambient temperature. GPC analysis of this polymer P2CuCl gave Mn = 29.0 kDa with Đ of 1.84, which is comparable to previously obtained data. The polymerization of a 0.2 M solution of M2CuCl in DCM was followed by 31P NMR, showing an apparent rate constant of kap = 5.2 × 10−4 s−1 (see the Supplementary Materials for details).

Ionic substances like HBF4(OEt2) or [Ph4P]+Cl were found to retard the polymerization. The use of Ag(O3SCF3) leads to a stable colorless coordination polymer [Ag(O3SCF3)(M2)] (see the Supplementary Materials for details). [AuCl(SMe2)] promotes the polymerization of M2 but a less defined material is obtained. Metal salts like MnBr2(THF)2, Fe2Cl4(THF)3, CoCl2, NiCl2(DME), or ZnCl2 do not boost the polymerization.

The coordinated copper can be easily removed from P2CuCl, P2CuBr, and P3CuCl by adding an excess of phosphanes under the conditions indicated in Fig. 4. These conditions were optimized such that the polymers can be easily obtained in pure form as insoluble white solids while the by-products remain in solution. Alternatively, aqueous NaCN can be used to displace the Cu+ ions. The loss of copper is indicated by a substantial narrowing of the 31P NMR signals due to the absence of coupling to the quadrupolar Cu nuclei. Furthermore, the 31Pring resonance is shifted to significantly lower frequencies in P2 [δ(31Pring) = −47.4 ppm; ν½ = 40 Hz] and P3 (−46.1 ppm; ν½ = 136 Hz). The regular (P─C)n sequence in P2 and P3 was again proven by 1D (1H, 13C, and 31P) and 2D NMR (heteronuclear single-quantum coherence, heteronuclear multiple-bond correlation, and correlation spectroscopy) (figs. S71 to S79 for P2 and figs. S84 to S93 for P3). We were not able to obtain satisfactory GPC data of P2 or P3. However, the reaction of P2 with CuCl quantitatively gives P2CuCl and the oxidation of P2 using aqueous hydrogen peroxide gives polymer P2O, which could be characterized by GPC and high-mass MALDI-MS experiments.

Fig. 4 Synthesis of polymers P2CuCl, P2CuBr, and P3CuCl by coordination-induced polymerization (CIP), their decoordination to give the metalfree polymers P2 and P3, the depolymerization and oxidation.

The NMR spectra for P2 show only one relatively narrow peak for the 13C and 31P nuclei in the (P─C)n backbone. This strongly suggests that P2 shows remarkable stereoregularity (vide infra). On the contrary, polymer P3 consists of atactic (P─C)n backbones, which show several 13C and 31P resonances.

In the solid state, polymers P2 and P3 are stable for several months at room temperature in the absence of ultraviolet light, monitored by Fourier transform infrared spectroscopy. A very slow depolymerization of P2 and P3 to the monomers M2 and M3 was observed at room temperature in solution. The full conversion of P3 takes 2 months but quantitatively leads to analytically pure M3. The even slower depolymerization of P2 was followed by 31P and 1H NMR spectroscopy in 1,1,2,2-tetrachloroethane-d2 at four different temperatures (353, 373, 393, and 410 K) (see the Supplementary Materials for details). The depolymerization reaction follows first-order kinetics, and Arrhenius plots allow us to determine the activation energy Ea to be 32.9 ± 5.3 kJ mol−1G#298K = 104.9 ± 9.3 kJ mol−1) and 39.8 ± 7.9 kJ mol−1G#298K = 106.1 ± 14.3 kJ mol−1) based on 1H and 31P NMR, respectively (ΔG# = Gibbs energy of activation). Note that the Ea for the depolymerization of polyolefins is much larger (200 to 300 kJ mol−1) (35). Under the irradiation with light of λ = 365 nm or simply sunlight, quantitative conversion of P2 and P3 to M2 and M3 is achieved at room temperature within 10 min. When CuCl is added to these reaction mixtures, the quantitative reconstitution of the polymers P2CuCl and P3CuCl is observed.

Mechanism and DFT calculations

It is well known that coordination to a metal center renders the P═C bonds of phosphinines more susceptible to addition of nucleophiles such as water (36, 37). When (M1CuCl)2 4, which does not polymerize, is reacted with water, complex 7 is obtained in high yield (Fig. 5A). When the polymerization of M2CuCl with the addition of water was retarded by [Ph4P]+Cl, the formation of an analog of 7 was observed (see the Supplementary Materials for details).

Fig. 5 Proposed simplified mechanism of the polymerization.

(A) Addition of water to the C═P bond in 4. (B) Possible mechanism for the polymerization of the phosphanaphthalene copper complexes MCuX, represented by M3CuCl. (C) Calculated isotactic pentamer of P3CuCl (left) and P3 (right).

According to these experimental observations, further supported by end group analysis by MALDI-TOF (time-of-flight) study of P2 (see the Supplementary Materials for details), we therefore assume that hydroxide ions initiate the polymerization. A possible mechanism, which is bolstered by density functional theory (DFT) calculations [M06-2X/6-31G(d,p)], is depicted in Fig. 5B. The addition of OH to the phosphorus center of M3CuCl is exergonic (ΔG = −60.7 kcal mol−1) and leads first to intermediate I1, which rearranges to its more stable tautomer I2G = −11.4 kcal mol−1). This can either be protonated to give 7′ or react with monomer M3CuCl, which is in large excess at the beginning of the reaction, to give intermediate I3 in an exergonic reaction (ΔG = −29.0 kcal mol−1). Note that a related water-promoted dimerization of a coordinated phosphinine has been recently reported (38). Chain propagation starts from I3 at the nucleophilic carbon center. Protonation terminates the active chain ends. According to the calculations, each chain propagation step is exergonic by about −23 kcal mol−1 (figs. S131 to S132). DFT calculations on a pentamer of P3CuCl (Fig. 5C) indicate that the isotactic polymer is slightly more stable than the syndiotactic or atactic one. This may explain why polymer P2 carrying the bulkier PO(OEt)2 substituent shows remarkable stereoregularity as indicated by a single and only slightly broadened 31P NMR resonance. DFT calculations also predict that the metal-free polymer P3 is thermodynamically unstable and that each depolymerization step is exergonic by about −4 kcal mol−1 (figs. S133 to S134).

In conclusion, the P═C bonds in 2-phosphanaphthalenes can be polymerized in the presence of Cu(I) halides to give, as a first example, a family of polymers with a regular (P─C)n backbone of three-valent phosphorus. Under irradiation with light, the quantitative depolymerization of the metal-free polymers is triggered. In solution, both polymerization and depolymerization proceed under very mild conditions at room temperature and can be repeatedly performed. However, in the solid state, these stimulus-responsive polymers remain stable, which may allow for potential applications in the course of their lifetime. For example, because they contain coordination sites for metals, various metallopolymers with tunable properties are potentially accessible, which can be used for different applications such as immobilized catalysts or optoelectronic materials (39). At the end of their lifetime, these polymers can be easily unzipped in solution and the monomers can be fully recycled. Although the general nature of these observations remains open to further investigation, the results described here do show that monomers can be found to prepare fully recyclable polymers under mild conditions exemplifying the concept of cradle-to-cradle processes in polymer science.


Because of the lengthy description of the experimental methods, synthesis, and characterization of all the compounds, these details can be found in the Supplementary Materials.


Supplementary material for this article is available at

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We thank C. Zhu and J. Nicolas for the help with GPC tests. Y.M. acknowledges China Scholarship Council for fellowship. Funding: This work was supported by the ETH Zürich through grant 0-20406-18. Author contributions: Y.M. conceived the idea, performed the experiments, and interpreted the data. H.G. directed the project. Y.M., Y.B., and H.G. discussed the results and prepared the manuscript. J.J.G.-C. performed computational studies. Y.B., G.L.C., and N.W. measured GPC and DSC, solid-state NMR, and high-mass MALDI-MS, respectively. Competing interests: The authors declare that they have no competing interests. Data and materials availability: Experimental details and characterization data are provided in the Supplementary Materials. Crystallographic data for this paper (CCDC number in SM) can be obtained free of charge from The Cambridge Crystallographic Data Centre.

Stay Connected to Science Advances

Navigate This Article