Organocatalyzed atom transfer radical polymerization driven by visible light

See allHide authors and affiliations

Science  27 May 2016:
Vol. 352, Issue 6289, pp. 1082-1086
DOI: 10.1126/science.aaf3935

Precise control from a metal-free catalyst

Polymerization can be a rather dangerous free for all, with molecules joining randomly in chains at a chaotic pace. One of modern chemistry's great accomplishments has been the development of methods to assemble polymers in steady, orderly steps. However, order comes at a price, and often it's the need for metal catalysts that are hard to remove from the plastic product. Theriot et al. used theory to guide the design of a metal-free light-activated catalyst that offers precise control in atom transfer radical polymerization, alleviating concerns about residual metal contamination (see the Perspective by Shanmugam and Boyer).

Science, this issue p. 1082; see also p. 1053


Atom transfer radical polymerization (ATRP) has become one of the most implemented methods for polymer synthesis, owing to impressive control over polymer composition and associated properties. However, contamination of the polymer by the metal catalyst remains a major limitation. Organic ATRP photoredox catalysts have been sought to address this difficult challenge but have not achieved the precision performance of metal catalysts. Here, we introduce diaryl dihydrophenazines, identified through computationally directed discovery, as a class of strongly reducing photoredox catalysts. These catalysts achieve high initiator efficiencies through activation by visible light to synthesize polymers with tunable molecular weights and low dispersities.

Over the past two decades, atom transfer radical polymerization (ATRP) (14) has matured into one of the most powerful methodologies for precision polymer synthesis (5). Strict control over the equilibrium between a dormant alkyl halide and an active propagating radical dictates a low concentration of radicals and minimizes bimolecular termination to achieve controlled polymer chain growth (6). ATRP has historically relied on transition-metal catalysts to mediate this equilibrium and polymerize monomers with diverse functionality into macromolecules with controlled molecular weight (MW), low MW dispersity (Ð), defined chemical composition, and complex architecture (7).

The caveat of traditional ATRP has been that the transition-metal catalysts present purification challenges for the polymer products and impede their use in biomedical and electronic applications (8). Despite substantial strides in lowering catalyst loading (9, 10) and facilitating purification (11), organocatalyzed methods remain highly desirable for circumventing the need for metal removal, reducing toxicity concerns, and avoiding interference with electronic systems. Organocatalyzed variants of ATRP by use of alkyl iodide initiators have been established, although they are not a broadly applicable replacement for metal-catalyzed ATRP (1214).

Our interest in this field originated in 2013 with the discovery that perylene could serve as an organic visible-light photoredox catalyst (PC) to mediate an ATRP mechanism with alkyl bromide initiators, albeit with less control over the polymerization than has become the benchmark for traditional metal-catalyzed ATRP (1517). Our ongoing work has striven to establish organocatalyzed ATRP (O-ATRP) for the synthesis of polymers with the precision of traditional ATRP, using visible-light PCs to realize energy-efficient polymerization methods that eliminate a major limitation of ATRP. Although photoredox catalysis has been established for decades, visible-light photoredox catalysis has drawn increasing attention by presenting the opportunity to harness solar energy to mediate chemical transformations under mild conditions (18, 19). Phenyl phenothiazine derivatives have since also proven effective as PCs for the ATRP of methacrylates (20) and acrylonitrile (21) but require irradiation by ultraviolet light and leave much room for improvement for generating polymers with higher molecular weights and lower dispersities coupled with increased initiator efficiency.

Our proposed mechanism of photoredox O-ATRP posits reversible electron transfer (ET) from the photoexcited PC in order to reversibly activate an alkyl bromide initiator (Fig. 1C). In addition to the requirement that the excited triplet state 3PC* possess sufficiently strong reducing power to activate the initiator, a delicate interplay must be balanced between the stability of the radical cation 2PC•+ and its capacity to oxidize the propagating radical so as to efficiently deactivate the propagating polymer and yield a controlled radical polymerization.

Fig. 1 PC development for O-ATRP.

(A) Polymerization of MMA to well-defined polymers by using photoredox O-ATRP driven by sunlight. (B) Structures of the diphenyl dihydrophenazine PCs 1 to 4 used in this study. (C) A proposed mechanism for ATRP mediated by a PC via photoexcitation to 1PC*, intersystem crossing (ISC) to the triplet state 3PC*, ET to form the radical cation doublet 2PC•+, and back ET to regenerate PC and reversibly terminate polymerization.

Computationally directed discovery (22, 23) inspired us to focus on 5,10-diphenyl-5,10-dihydrophenazines as a potential class of PCs for O-ATRP (Fig. 1B). The phenazine core is shared by several biologically relevant molecules that serve as redox-active antibiotics (24, 25), whereas synthetic derivatives have drawn interest in organic photovoltaics (2628) and organic ferromagnets (29, 30). We hypothesized that an appropriate union between the excited-state reduction potential (E0*) and the stability of the radical cation PC•+ resulting from ET to the initiator would be required for the production of polymers with controlled MW and low Ð. As such, we investigated electron-donating (OMe, 1), neutral (H, 2), and electron-withdrawing (CF3, 3, and CN, 4) moieties on the N-phenyl substituents.

Density functional theory (DFT) was used to calculate the reduction potentials of the triplet excited-state PCs, initiator, and propagating radicals (Fig. 1B) (31). We found that 2 possesses a triplet excited-state reduction potential of E0(PC•+/3PC*) = –2.34 V versus saturated calomel electrode (SCE). Functionalization of the phenyl substituents with an electron-donating group OMe (1) strengthened the E0* to –2.36 V, whereas introduction of CF3 or CN electron-withdrawing groups (EWGs) weakened the E0* to –2.24 and –2.06 V for 3 and 4, respectively, all of which is corroborated by the measured values within experimental error (table S1). The triplet excited states of these PCs are all strongly reducing with respect to 1e transfer to the ethyl α-bromophenylacetate (EBP) initiator; we calculated that E0(EBP/EBP•–) = –0.74 V versus SCE for an adiabatic ET, which is consistent with our cyclic voltammetry results, which show that the onset of EBP reduction occurs at ~–0.8 V versus SCE (fig. S28). These phenazine derivatives are significantly more reducing than are commonly used metal PCs (18), including polypyridyl iridium complexes (E0* as negative as –1.73 V versus SCE) that have been used in photomediated ATRP (32, 33). Iridium PCs are expensive, challenging to remove from the product, and have only been demonstrated to produce polymers with Ð as low as 1.19.

The remarkable reducing power of these dihydrophenazine-based PCs arises from a distinct combination of their high triplet-state energies (~2.2 to 2.4 eV) and the formation of relatively stable radical cations [E0(PC•+/PC) = ~–0.1 to 0.2 V] upon their oxidation. These radical cations are also sufficiently oxidizing to deactivate the propagating chains. We computed E0s for propagating radicals with n monomer repeat unit (or units) bound to ethyl phenylacetate (EPA) of E0[(EPA – MMAn)/(EPA – MMAn)•–] = –0.74, –0.86, and –0.71 V for n = 0, 1, and 2, respectively. These E0s are sufficiently negative with respect to oxidization by the radical cations to drive rapid radical deactivation and regeneration of the PC to complete the photocatalytic cycle.

An initial series of target PCs (1 to 4) were synthesized in two steps from commercial reagents in good yields. Under otherwise identical conditions, all of the PCs were tested in the polymerization of methyl methacrylate (MMA) by using EBP as the initiator and white light-emitting diodes (LEDs) for irradiation in dimethylacetamide (Table 1, run 1, and table S2, runs S1 to S3). All four PCs proved effective in polymerization after 8 hours of irradiation, with the PCs bearing EWGs exhibiting the best catalytic performance. PC 3 proved superior in producing polymers with a combination of not only the lowest dispersity (Ð = 1.17) but also the highest initiator efficiency (I* = 65.9%) (I* is the ratio between the theoretical and experimentally measured number average molecular weight) (Table 1, run 1). Using methyl α-bromoisobutyrate as the initiator was also efficient but did not achieve the same level of control of the polymerization achieved with EBP (table S2, run S5). Additionally, polymerization could be driven by sunlight to produce poly(methyl methacrylate) (PMMA) with a low dispersity of Ð = 1.10 (run 2).

Table 1 Results for the organocatalyzed ATRP of MMA catalyzed by 3 using white LEDs or sunlight.

Asterisk indicates use of sunlight. Experimental details are provided in the supplementary materials.

View this table:

Time-point aliquots were taken during polymerization to monitor the MW and Ð progression as a function of monomer conversion (Fig. 2, A and B). The control provided by 3 was evidenced by the linear increase in polymer MW and low Ð throughout the course of polymerization. However, the y intercept of the number-average molecular weight (Mn) versus conversion plot was 3.46 kDa, suggesting an uncontrolled chain-growth period with addition of ~32 MMA equivalents during the onset of polymerization before precise control was attained, whereas an ideal polymerization would have a y intercept equal to the mass of the initiator (MW of EBP = 243 Da).

Fig. 2 Polymerization results using PC 3.

(A) Plot of molecular weight as a function of monomer conversion and (B) plot of dispersity as a function of monomer conversion for the polymerization of MMA mediated by 3. (C) Chain-extension from a PMMA macro-initiator (black) to produce block copolymers with MMA (green), BMA (blue), and BA (red). (D) GPC traces of each polymer depicted in (C) (color coded).

We next examined the effect of adjusting the initiator ratio relative to monomer and PC (runs 3 to 6). The weight-average molecular weight (Mw) of the resulting PMMA could be modulated from 7.12 to 85.5 kDa. High EBP ratios resulted in controlled polymerizations and low dispersities (Ð = 1.26 to 1.17), and despite the moderate loss of precise control over the polymerization at low EBP ratios (Ð = 1.54), high-MW polymer was produced with high initiator efficiency (Mw = 85.5 kDa, I* = 86.3%). Alternatively, adjusting the monomer ratio relative to EBP and PC regulated polymer MW while also maintaining low Ð (runs 7 to 10).

One of the greatest strengths of traditional ATRP is its capacity to synthesize advanced polymeric architectures, including block copolymers. The reversible-deactivation mechanism enforced in ATRP repeatedly reinstalls the Br chain-end group onto the polymer, and thus, isolated polymers can be used to reinitiate polymerization. A combination of nuclear magnetic resonance spectroscopy and matrix-assisted laser desorption ionization mass spectroscopy were used to confirm the expected EBP-derived polymer chain-end groups for a polymer produced through the proposed photoredox O-ATRP mechanism (figs. S12 and S13). Additionally, to further support the posited O-ATRP mechanism, a series of block polymerizations were performed to probe the Br chain-end group fidelity.

First, after initial polymerization of MMA proceeded for 12 hours, additional MMA was added to the reaction mixture. Gel permeation chromatography (GPC) analysis revealed that the MW of the resulting polymer quantitatively increased (fig. S22). Second, after polymerization of MMA was allowed to proceed for 8 hours, the reaction mixture was placed in the dark for 8 hours, and subsequently, additional MMA, benzyl methacrylate (BMA), or butyl acrylate (BA) was added. No polymerization took place during the dark period, whereas the subsequent addition of monomer and further illumination resulted in continued and controlled polymer chain growth (figs. S23, S24, and S26). Third, an isolated polymer was reintroduced to a solution of monomer and catalyst and exposed to light in order to ascertain whether it would serve as a macro-initiator for the synthesis of block polymers. This chain-extension proved successful with MMA, BMA, and BA (Fig. 2, C and D). The chain-extension polymerization from an isolated polymer produced from this polymerization method firmly supports the conclusion that this methodology proceeds through the O-ATRP mechanism, whereas all of these experiments revealed baseline resolved peaks in the GPC traces, demonstrating high chain-end group fidelity.

DFT calculations were performed in order to gain insight into the differences in the performances of the PCs, all of which possess similar E0(PC•+/3PC*)s and E0(PC•+/PC)s that are sufficiently reducing and oxidizing, respectively, to drive the photocatalytic cycle of Fig. 1C. As such, we reasoned that the superior performances of 4 and, in particular, 3 must be qualitatively different from those of 1 and 2 and result from a more complex effect.

Inspection of the triplet state (3PC*) frontier orbitals reveals qualitative differences in these PCs (Fig. 3). The low-lying singly occupied molecular orbitals (SOMOs) of all the PCs are similar, with the electron localized over the phenazine π system. For PCs 1 (OMe) and 2 (H), the high-lying SOMO is also localized on the phenazine rings; in contrast, for 3 (CF3) and 4 (CN), the high-lying SOMO, occupied by the reducing e, resides on the phenyl ring (or rings). We contend that the CF3 and CN EWGs of 3 and 4 stabilize their π* orbitals localized on the phenyl rings relative to the phenazine-localized π* orbital that is the high-lying SOMO of 1 and 2. This reorders the energies of the π* orbitals so that a π* orbital localized on the phenyls becomes the high-lying SOMO of 3 and 4, although the low-lying SOMO localized on the phenazine moiety remains singly occupied. Thus, 3 and 4 differ qualitatively from 1 and 2 in that their two triplet electrons reside on either the phenazine or the phenyl substituent and are thus spatially separated.

Fig. 3 Calculated triplet state (3PC*) frontier orbitals and excited-state reduction potentials E0* of diphenyl dihydrophenazine PCs 1 to 4.

(Top) The higher-lying SOMO. (Bottom) The low-lying SOMO. Phenyl functionalization with electron-withdrawing groups (CF3 and CN) localizes the high-lying SOMO on the phenyl.

Furthermore, a comparison of 3 to 4 elucidates another important distinction. For 3, the high-lying SOMO is localized on one of the phenyl rings, whereas in 4, the reducing e is delocalized over both phenyl rings. Surprisingly, calculations revealed one of the C–F bonds of the CF3 functionalized phenyl that bears the high-lying SOMO of 3 is lengthened from 1.35 to 1.40 Å, indicating partial localization of electron density on the C–F antibond. This symmetry-breaking effect in the triplet state of 3 creates a more localized, higher electron density of the reducing electron of 3 relative to 4 while also maintaining the spatial separation between the two SOMO electrons that preserves the reducing potential of the triplet.

With the above observations in mind, we attempted to discover even more efficient PCs to mediate O-ATRP using computational chemistry to design diaryl dihydrophenazines that possess sufficiently strong E0*s and spatially separated excited-state SOMOs, with the higher-energy SOMO localized over only one of the aromatic substituents off the dihydrophenazine core. Using these principles, we designed and synthesized 2-napthyl (5) and 1-napthyl (6) derivatives—with strong E0*s of –2.20 and –2.12 V, respectively—and SOMOs with the targeted desirable geometric features (Fig. 4). Using EBP as the initiator, both PCs proved successful in the polymerization of MMA (table S2, runs S8 and S9). Although 5 produced PMMA with an impressively low Đ of 1.03 (Mw = 9.35 kDa, I* = 46.1%)—rivaling metal ATRP catalysts—6 produced PMMA with a slightly higher I* (47.5%), faster polymerization rates, and a similarly low Đ of 1.08 (Mw = 12.3 kDa). The plot of Mn versus monomer conversion exhibits a y intercept of 850 Da, demonstrating the attainment of control over polymerization after the addition of only ~6 MMA units, which is correspondingly much more efficient control in the O-ATRP mediated by 6 than is achieved with 3 (Fig. 5A). Thus, we investigated 6 in more detail as the PC in the polymerization of MMA.

Fig. 4 Computationally directed discovery of PCs 5 and 6.

(A) Structures of 5 and 6 and the calculated E0*. (B) Triplet-state frontier orbitals of 5 and 6 showing the (top) higher-lying SOMO and (bottom) low-lying SOMO.

Fig. 5 Results for the polymerization of MMA using PC 6.

(A) Plot of Mn and Đ versus monomer conversion for the polymerization of MMA under continuous irradiation. (B) Plot of monomer conversion versus time and (C) plot of Mn and Đ (solid symbols indicate after irradiation, and open symbols indicate after dark period) versus monomer conversion, using 6 as the PC during pulsed light irradiation with white LEDs. Experimental details are provided in the supplementary materials.

A survey of initiators commonly used in traditional metal-catalyzed ATRP in conjunction with 6 (Table 2, run 11, and table S2, runs 9 to 12) revealed that methyl 2-bromopropionate (MBP) provided the best overall results for the polymerization of MMA (Mw = 10.6 kDa; Đ = 1.28; I* = 88.1%). Furthermore, temporal control was realized by using a pulsed-irradiation sequence (Fig. 5, B and C). Polymerization was only observed during irradiation and paused during dark periods, and the MW steadily increased with continued irradiation while producing a polymer with a low Đ of 1.17. Efficient control over the polymerization by 6 is highlighted by the consistently high I* achieved over broad reaction conditions to produce polymers with tunable MWs through varying initiator (runs 11 to 14) or monomer (runs 15 to 17) ratios.

Table 2 Results for the organocatalyzed ATRP of MMA catalyzed by 6 using white LEDs.

Experimental details are provided in the supplementary materials.

View this table:

We envision that this O-ATRP catalyst platform will expand the application scope for polymers beyond those synthesized by metal-catalyzed ATRP, and their impressively strong reducing power presents great promise for their application toward other challenging chemical transformations. We also anticipate that the governing principles that afford these organic photocatalysts with their desirable properties will be exploited through computational design to discover additional photochemical platforms with capacities for a variety of applications.

Supplementary Materials

Materials and Methods

Figs. S1 to S28

Tables S1 to S3

Coordinates of Calculated Molecular Structures

References (3539)

References and Notes

  1. G. M. Miyake, U.S. Patent application no. 61,212 (2013).
  2. G. M. Miyake, Organocatalyzed photoredox mediated polymerization using visible light. U.S. Patent no. US 9,156,921 B2 (2015).
  3. Recently, it was suggested that the singlet excited state of phenyl phenothiazines may be responsible for dehalogenations. Because of the oxygen intolerance of this polymerization system and the well-known phenomenon that the triplet state possesses a substantially longer lifetime sufficient for photocatalysis, we have based our computational interpretations on this polymer mechanism proceeding via the triplet state of the diphenyl dihydrophenazines (34).
Acknowledgments: G.M.M. is grateful for financial support from the University of Colorado Boulder and Advanced Research Projects Agency–Energy. This work was supported in part by NSF grant CHE-1214131 (C.B.M. and C.-H.L.). J.C.T. is thankful for a NSF Graduate Research Fellowship Program fellowship. M.D.R. acknowledges support from a Graduate Assistance in Areas of National Need fellowship. We also gratefully acknowledge use of Extreme Science and Engineering Development Environment supercomputing resources (NSF ACI-1053575) and the Janus supercomputer, which is supported by NSF (CNS-0821794) and the University of Colorado Boulder. We thank L. Hansman, A. Lockwood, S. Fatur, and N. Damrauer for technical assistance and enlightening discussions. We have filed a provisional patent application on the work described here.

Stay Connected to Science

Navigate This Article