Nitroxide-mediated radical polymerization of carbon dioxide-expanded methyl methacrylate


  • Da Wei Pu,

    1. Centre for Advanced Macromolecular Design (CAMD), School of Chemical Engineering, The University of New South Wales, UNSW Sydney, NSW 2052, Australia
    Search for more papers by this author
  • Frank P. Lucien,

    Corresponding author
    1. Centre for Advanced Macromolecular Design (CAMD), School of Chemical Engineering, The University of New South Wales, UNSW Sydney, NSW 2052, Australia
    • Centre for Advanced Macromolecular Design (CAMD), School of Chemical Engineering, The University of New South Wales, Sydney, NSW 2052, Australia
    Search for more papers by this author
  • Per B. Zetterlund

    Corresponding author
    1. Centre for Advanced Macromolecular Design (CAMD), School of Chemical Engineering, The University of New South Wales, UNSW Sydney, NSW 2052, Australia
    • Centre for Advanced Macromolecular Design (CAMD), School of Chemical Engineering, The University of New South Wales, Sydney, NSW 2052, Australia
    Search for more papers by this author


original image

Nitroxide-mediated radical polymerization based on TEMPO has been performed for methyl methacrylate under conditions of CO2-expanded monomer at 7 MPa and 90°C. The fraction of propagating radicals lost due to disproportionation with nitroxide, a well-known problem in NMP of methacrylates, decreased in the presence of CO2. This is consistent with deactivation between propagating radical and nitroxide proceeding more rapidly in the lower viscosity induced by CO2. However, the polymerization ceased at relatively low conversions of approximately 19% both with and without CO2, indicating that disproportionation between propagating radical and nitroxide is a major problem also in CO2-expanded MMA.


Nitroxide-mediated radical polymerization (NMP),1, 2 one of the techniques for controlled/living radical polymerization,3, 4 is based on the reversible deactivation of propagating radicals by a stable nitroxide radical. A significant drawback of NMP is that it is in general incompatible with alkyl methacrylates5–9 primarily due to the tendency of nitroxide radicals and the corresponding propagating radicals to undergo disproportionation generating a hydroxylamine and a dead chain with an unsaturated end group.10, 11 This reaction may also proceed unimolecularly, although it has been shown for the present system that the bimolecular pathway is predominant.6, 12, 13 It was reported in 2007 that the nitroxide known as 2,2-diphenyl-3-phenylimino-2,3-dihydroindol-1-yloxyl nitroxide (DPAIO)14 can be successfully used for NMP of methyl methacrylate (MMA) because of the absence of such disproportionation reactions.13 However, NMP of alkyl methacrylates using common nitroxides such as 2,2,6,6-tetramethylpiperidine-N-oxyl (TEMPO) and N-tert-butyl-N-[1-diethylphosphono-(2,2-dimethylpropyl)] (SG1) does proceed satisfactorily in the presence of a relatively small amount (approx. 8 mol %) of styrene, that is, as a copolymerization.15–17

The value of the rate coefficient of deactivation in NMP (kdeact), that is, the radical coupling reaction of a propagating radical and nitroxide, is generally in the range 106–108 M−1 s−1 at low monomer conversion.18 For a variety of low-molecular weight carbon-centered radicals and TEMPO, the radical coupling reaction has been reported to be close to diffusion-controlled, yet influenced by solvent viscosity.19 The value of kdeact for TEMPO and a poly(MMA) radical has not been determined, but kdeact for TEMPO and a low-molecular weight model radical of poly(MMA), the 2-carbomethoxy-2-propyl radical, has been reported as 5.9 × 108 M−1 s−1 at 120 °C,20 and kdeact for TEMPO and R-MMA, where R = 2-tert-butoxy-carbonyl-2-propyl, has been estimated as kdeact = 2 × 107 M−1 s−1 at 100 °C.6 In a polymerization, the viscosity of the system gradually increases as monomer is converted to polymer, and it is conceivable that the deactivation reaction becomes diffusion-controlled at some stage.

In catalytic chain transfer (CCT) polymerization, reactions between propagating radicals and cobaloxime complexes are considered to be diffusion-controlled in the case of methacrylates. The chain transfer step (rate coefficient ktr) involves abstraction of a hydrogen atom from the propagating radical, leading to the formation of a dead chain with an unsaturated end group and a cobalt hydride species.21 The hydride subsequently reacts with monomer to initiate a new monomeric radical and regenerate the cobaloxime catalyst. It has been demonstrated that an inverse relationship exists between the chain transfer constant (ktr/kp) and monomer viscosity, thereby providing evidence of a diffusion-controlled chain transfer reaction.22 This is further supported by work on CCT polymerization of MMA in supercritical CO2.23 The chain transfer constant in supercritical CO2 is an order of magnitude higher than that in bulk MMA and this is attributed to the gas-like viscosity of the supercritical medium.

In more recent studies, measurements of the chain transfer constant in CCT have been performed in CO2-expanded methacrylates and styrene.24, 25 This technique uses greatly reduced operating pressures in comparison with the reaction performed in supercritical CO2 under homogeneous conditions. Typical values of ktr for CO2-expanded methacrylates vary in the range 107–108 M−1 s−1 and are significantly higher than the values in bulk monomer. A useful feature of CO2-expanded monomer systems is that values of ktr can be determined over a wide range of viscosity by controlling the degree of expansion of the monomer. This permits a more detailed investigation of the inverse relationship between rate coefficients and viscosity.

The propagation reaction in conventional (nonliving) radical polymerization of MMA becomes diffusion-controlled at approximately 60% conversion at 33 °C.26 The deactivation reaction in TEMPO-mediated polymerization of MMA (also involving a polymer chain and a small molecule) would become diffusion-controlled at a significantly lower conversion, because the value of kdeact (under chemical control) would be orders of magnitude higher than the rate coefficient for propagation (under chemical control; 1620 M−1 s−1 at 90 °C).27

The propagating radical (P) can thus react with the nitroxide (T) either by (i) combination (deactivation) or by (ii) disproportionation (rate coefficient kβHtr). Depending on conditions (temperature, conversion), the deactivation reaction may be diffusion-controlled, whereas the disproportionation reaction is under chemical control (kβHtr = 1.69 × 103 M−1 s−1 for MMA/SG1 at 70 °C,11 and kβHtr = 1.4 × 106 M−1 s−1 for n-butyl methacrylate/TEMPO at 130 °C).28 If the polymerization is carried out in CO2-expanded monomer, the reduced viscosity and increased diffusion rates may increase the rate of deactivation, whereas the rate of disproportionation would remain unaffected (assuming no solvent effect). Therefore, under such conditions, the polymerization is predicted to proceed with better control/livingness (as confirmed by simulations,7) because the ratios kβHtr/kdeact and RβHtr/Rp would decrease (RβHtr = rate of H-abstraction by nitroxide; Rp = rate of monomer consumption; RβHtr/Rp = kβHtr[P][T]/kp[P][M] = kβHtr[T]/kp[M]; An increase in kdeact leads to a decrease in [T], thus RβHtr/Rp decreases).

In this work, NMP of MMA using TEMPO has been carried out in bulk and under expansion by CO2, with the objective of investigating whether the increased diffusivity in the CO2-expanded medium significantly influences the course of the polymerization.



2,2,6,6-tetramethylpiperidinyl-1-oxy (TEMPO, Sigma–Aldrich, 98%) and liquid CO2 (Linde Gases, 99.5%) were used as received. MMA (Sigma–Aldrich, 99%) was passed through a column of activated basic alumina to remove the inhibitor 4-methoxyphenol. 2,2-Azobis(isobutyronitrile) (AIBN, DuPont) was purified by recrystallization twice from methanol.

Polymerization Procedure

A schematic diagram of the experimental apparatus is shown in Figure 1. Polymerizations were conducted in a sight gauge reactor with an internal volume of 40 mL. Mixing was achieved with a magnetic stirrer designed to operate at a maximum speed of 2000 rpm. The pressure in the reactor was measured with a pressure transducer (Druck, PDCR 911), and the temperature was monitored with a Pt100 sensor. The temperature in the reactor was controlled by circulating oil from a heater/cooler unit (Julabo, FP35) through internal channels encased in the reactor block. The heater/cooler unit was operated under an external control mode in which the Pt100 sensor was used as the input signal. In this way, the temperature of the oil was regulated independently of the operator. High-pressure CO2 was added to the reactor via an inlet tube using a syringe pump (Isco, 260D).

Figure 1.

Schematic diagram of the polymerization reactor. CO2 = carbon dioxide cylinder; H/C = heater/cooler unit; MS = magnetic stirrer; P = pressure transducer; R = reactor vessel; SP = syringe pump; T = Temperature probe (Pt100 sensor); V = valve.

The reactor was cooled to 15 °C before the addition of any materials. A typical polymerization was conducted as follows: A solution of MMA (9.44 g, 9.429 × 10−2 mol), TEMPO (7 mg, 4.714 × 10−4 mol), and AIBN (6 mg, 3.929 × 10−4 mol) was added to the reactor. The reactor was sealed and stirring was initiated. The reactor contents were purged with CO2 (0.5 MPa) to remove oxygen. For atmospheric polymerizations, the solution was purged with N2 (0.5 MPa) to remove oxygen. Continuous stirring (1400 rpm) was maintained to ensure equilibrium between the liquid and vapor phases. The system was heated to 90 °C with stirring followed by pressurization with CO2 to the desired operating pressure of 7 MPa (it took approximately 15 min to reach 90 °C). In preliminary work, it was established that negligible polymerization occurred during this heating stage. The polymerizations were stopped at prescribed times by cooling of the reactor to ambient temperature and subsequent depressurization. Monomer conversion was measured by gravimetry after evaporation of unreacted monomer (confirmed by 1H NMR).

Polymer Characterization

Molecular weights (MWs) and molecular weight distributions (MWDs) were determined by gel permeation chromatography (GPC) with a Shimadzu modular system with tetrahydrofuran as eluent at 40 °C at a flow rate of 1.0 mL/min with injection volume of 100 μL. The GPC was equipped with a DGU-12A solvent degasser, a LC-10AT pump, a CTO-10A column oven and an ECR 7515-A refractive index detector, and a Polymer Laboratories 5.0 μm bead-size guard column (50 × 7.8 mm) followed by four 300 × 7.8 mm linear Phenogel columns. The system was calibrated against polystyrene standards ranging from 500 to 106 g/mol. Absolute MWs were obtained using the universal calibration method with the following Mark Houwink Sakurada constants; poly(styrene),29K = 14.1 × 10−5 dL g−1 and α = 0.70; PMMA,30K = 12.8 × 10−5 dL g−1 and α = 0.697. Theoretical number-average molecular weights (Mn,th) were calculated based on the number of chains being equal to the number of radicals generated by decomposition of AIBN, assuming an initiator efficiency (f) of 0.6, according to Mn,th = (αMMAMMMA,0/2f[AIBN], where MMMA,0 is the MW of MMA and αMMA is the MMA conversion.

1H NMR spectra were recorded on a Bruker DPX spectrometer (300 MHz) in CDCl3 with (CH3)4Si as the internal standard.


Polymerization Rate and MWD

TEMPO-based NMPs of MMA were performed in the absence of CO2 at 0.1 MPa (atmospheric pressure) and in the presence of CO2 at 7 MPa at 90 °C. In the presence of CO2 (and without), the solutions remained clear throughout the polymerizations, indicating that the homogeneous expansion limit was not reached. In the absence of CO2, it is well known that TEMPO-based NMP of MMA does not proceed beyond relatively low conversion because of the disproportionation reaction between propagating radicals and TEMPO.5–8 As outlined in the Introduction, it was speculated that the presence of CO2, and the thus reduced viscosity of the polymerization mixture, may result in a reduction in the rate of this disproportionation reaction relative to the rate of propagation, and thus higher final conversion and improved control/livingness. However, in both cases, the polymerizations reached limiting conversions of approximately 19% conversion (Fig. 2), and the conversion-time data were similar.

Figure 2.

Experimental and simulated conversion-time data for TEMPO-based NMP of MMA in the absence of CO2 at 0.1 MPa (atmospheric pressure; filled circles) and in the presence of CO2 at 7 MPa (open circles) at 90 °C. The full line represents simulated data (see Appendix for simulation details). [MMA]0 = 9.35 M (bulk); [AIBN]0 = 0.039 M; [TEMPO]0 = 0.047 M.

In addition to the effect of CO2 on viscosity, expansion of MMA with CO2 causes an increase in total volume, that is, there is a dilution effect. In the present system, introduction of CO2 was accompanied by an increase in volume by approximately 25%. PREDICI simulations (model described in Appendix) were carried out to elucidate how this level of dilution would impact the conversion-time data (assuming no effect of CO2 on the rate coefficients), revealing that the reduction in conversion caused by dilution at any given conversion was <1%. The value of kβHtr for MMA/TEMPO/90 °C has not been estimated, although it has been reported28 that kβHtr(TEMPO/butyl methacrylate/130 °C) = 1.4 × 106 M−1 s−1. In the present simulations, kβHtr was treated as an unknown, and reasonable agreement between simulated and experimental conversion-time data was obtained for kβHtr = 5 × 104 M−1 s−1. At the simulated limiting conversion of close to 22%, at which point the alkoxyamine concentration is close to zero (i.e., zero livingness), Mn = 4323 g/mol and Mw/Mn = 1.24 (dilution effect of CO2 accounted for). The simulated Mn value was close to experiment (3600 g/mol), although the experimental MWD (Mw/Mn = 1.93) in the presence of CO2 was broader than the simulated one. However, the simulated MWDs do not account for experimental GPC broadening,31 but the difference is too great for this to be the only reason of the discrepancy. PREDICI simulations were also carried out to investigate whether an increase in the termination rate coefficient (kt) in the CO2-expanded MMA may affect the results (kt is diffusion-controlled, and higher diffusion rates in the CO2-expanded medium may thus increase kt). It has previously been estimated that kt increases by a factor of approximately three in the case of CCT polymerization of MMA in CO2-expanded monomer.32 However, the extent of CO2-expansion was considerably greater in that work than in this work (expansion of ≫50 vol % rel. to 25 vol %). Simulations using kt values (ktc and ktd—see Appendix) increased by a factor of two resulted in a reduction in polymerization rate of only 1%, that is, the effect is negligible.

Figure 3 shows MWDs obtained at different polymerization times in the absence of CO2 at 0.1 MPa (atmospheric pressure) and in the presence of CO2 at 7 MPa at 90 °C. In both cases, the MWDs shifted to higher MWs with increasing conversion. The Mn values increased with conversion and were in reasonable agreement with Mn,th, and no major systematic difference could be discerned between the two data sets (Fig. 4). The Mw/Mn values increased with increasing conversion, and were somewhat lower in the presence of CO2 (Fig. 4). The main chain end-forming events would be bimolecular termination by disproportionation and disproportionation with TEMPO, and thus the number of chains would be expected to remain close to constant, and one would therefore anticipate that MnMn,th, as confirmed by experiment.

Figure 3.

Molecular weight distributions obtained from TEMPO-based NMP of MMA in the absence of CO2 at 0.1 MPa (atmospheric pressure; 1 bar) and in the presence of CO2 at 7 MPa (70 bar) at 90 °C. [MMA]0 = 9.35 M (bulk); [AIBN]0 = 0.039 M; [TEMPO]0 = 0.047 M.

Figure 4.

Mw/Mn and Mn values versus conversion for TEMPO-based NMPs of MMA in the absence of CO2 at 0.1 MPa (atmospheric pressure (1 bar); filled circles) and in the presence of CO2 at 7 MPa (70 bar; open circles) at 90 °C. [MMA]0 = 9.35 M (bulk); [AIBN]0 = 0.039 M; [TEMPO]0 = 0.047 M. The dotted line is the theoretical Mn (Mn,th).

End-Group Analysis

1H NMR analysis revealed the presence of unsaturated functionality as evidenced by resonances in the vinyl region of the spectrum (Fig. 5), assigned to the unsaturated ω end group of PMMA.10Mn(NMR) was estimated by comparison of the relative intensities of the 1H NMR resonances of the unsaturated end group with the resonance of the methoxy (OCH3) group, that is, based on the assumption that each polymer chain bears one unsaturated end group. The number of unsaturated end groups per polymer chain (f) was subsequently estimated via eq 110:

equation image(1)

The ratio of the rate coefficient for termination by disproportionation to that of combination (ktd/ktc) for MMA has been reported as 4.37 at 90 °C,33 which translates to approximately 81% of the propagating radicals terminating by disproportionation. Each termination event by disproportionation results in formation of one unsaturated chain and one saturated chain. If termination by disproportionation is the only pathway to unsaturated chain ends, the maximum value of f is 0.5, and ktd/ktc = 4.37 corresponds to f = 0.45.10

Figure 5.

1H NMR resonances corresponding to the unsaturated end group observed in the TEMPO-based NMP of MMA in the presence of CO2 at 7 MPa at 90 °C.

The value of f was estimated for polymerizations with and without CO2 (the data point at the highest conversion in each data series), resulting in f = 0.84 (without CO2) and 0.78 (with CO2). Thus, the average number of unsaturated end groups per chain was slightly lower in the presence of CO2. The fact that f ≫ 0.5 indicates the existence of a pathway other than termination by disproportionation to PMMA bearing an unsaturated end group, consistent with disproportionation between nitroxide and propagating radicals. Propagating radicals would undergo addition reaction with the unsaturated end group. However, the generated adduct radical undergoes fragmentation (not monomer addition) to regenerate the unsaturated end group, and the estimation of f is thus not affected.34

Assuming that T/P disproportionation does occur, and that the number of chains capped by nitroxide is negligible (reasonable considering the very minor changes in Mn, Mw/Mn and conversion after 20 h polymerization time) there are four different types polymer chains with regards to formation mechanism10: (a) A saturated chain by combination of two P; (b) A saturated chain by two P terminating by disproportionation; (c) An ω-end unsaturated chain by the reaction in b; (d) An ω-end unsaturated chain by nitroxide/P disproportionation. The following relationships between the number fractions (N) of various chains hold (Ntotal = 1):

equation image(2)
equation image(3)
equation image(4)

Considering that approximately 80% of P terminate by disproportionation in a conventional MMA polymerization,33 it follows that:

equation image(5)

Solving eqs 25 yielded the number fraction of chains formed by TEMPO/PMMA disproportionation (Nd): Nd = 0.72 (without CO2) and 0.59 (with CO2). In other words, the fraction of propagating radicals that lost their activity permanently by disproportionation with nitroxide was slightly reduced in the presence of CO2. This is consistent with the presence of CO2 causing a reduction in the rate coefficient ratio kβHtr/kdeact via an increase in kdeact because of lower diffusion resistance as outlined in the Introduction. However, the observed effects of CO2 on f and Nd are relatively small and may be within experimental error. This is supported by the fact that the limiting conversion was not significantly affected by CO2 (Fig. 2). It is worth noting that the values of Nd were considerably lower than for the system MMA(bulk)/SG1 in the presence of a vast excess of SG1,10 presumably mainly caused by the excess nitroxide in that study, although intrinsic differences between TEMPO and SG1 cannot be excluded. Finally, it is worth noting that based on theoretical arguments by Casey et al. for the analogous situation of propagation versus chain transfer to monomer (i.e., two different reactions undergone by the same two reactants),35 it can be predicted that the ratio kβHtr/kdeact would in fact be unaffected by viscosity (i.e., the two reactions would come under diffusion-control at the same conversion).


NMP based on the nitroxide TEMPO has been performed for MMA under conditions of CO2-expanded monomer at 7 MPa and 90 °C. Based on 1H NMR and GPC analysis, the fraction of propagating radicals that irreversibly lost their activity because of the disproportionation reaction with nitroxide decreased somewhat in the presence of CO2. This finding is consistent with the deactivation reaction between propagating radical and nitroxide proceeding more rapidly in the lower viscosity induced by the presence of CO2. However, similar to what has been previously observed in bulk NMP of MMA with TEMPO and other nitroxides, the polymerization ceased at relatively low conversions of approximately 19% both with and without CO2, indicating that disproportionation between propagating radical and nitroxide is a major problem also in CO2-expanded MMA. Modeling and simulations allowed estimation of the rate coefficient for the disproportionation reaction to be of the order of 5 × 104 M−1 s−1 (with or without CO2).


PBZ is grateful to the Australian Research Council for a Discovery Grant (DP1093343).


PREDICI Simulations

TEMPO-based NMP of MMA was simulated using the software PREDICI36 based on the following reaction steps:

Initiator Decomposition

equation image

Primary Radical Addition to Monomer

equation image


equation image
equation image


equation image
equation image


equation image


equation image
equation image
equation image
equation image

Disproportionation with Nitroxide

equation image
equation image

where R denotes the cyanoisopropyl radical from AIBN decomposition, P(denotes propagating radicals, RT and PT denote low MW and polymeric alkoxyamines, respectively, M is monomer and I is initiator. Parameter values (90 °C): kd(AIBN) = 4.47 × 10−4 s−137; AIBN initiator efficiency = 0.6; ki = 5.93 × 103 M−1 s−138 (recalculated for 90 °C based on Ei = 26.1 kJ/mol for styrene/2-cyano-2-propyl radical38); kp = 1620 M−1 s−127; ktc = 6.5 × 106 M−1 s−134; ktd = 2.8 × 107 M−1 s−134; kact = 1.87 × 10−4 s−16 (recalculated for 90 °C assuming Eact the same as for polystyrene-TEMPO39); kdeact = 2 × 107 M−1 s−16; kβHtr = 5 × 104 M−1 s−1 (estimated in these simulations by comparison with experimental conversion-time data). [MMA]0 = 9.35 M (bulk); [AIBN]0 = 0.039 M; [TEMPO]0 = 0.047 M. The values of f,40kd,41 and kp41 have been reported to change somewhat in supercritical CO2 relative to conditions of bulk monomer. However, the CO2 pressure in this work (not supercritical) is markedly lower, and thus such effects are deemed negligible. The value of kt (ktd + ktc) would be higher in CO2-expanded monomer (addressed in Results and Discussion). All rate coefficients were assumed to be independent of chain-length and conversion. In addition to the reaction steps above, the hydroxylamine species generated from the disproportionation reaction between nitroxide and propagating radical may transfer its hydrogen to a propagating radical.42 This would cause additional retardation and loss of control/livingness. However, in the absence of accurate estimates of the relevant rate coefficient, this reaction was not included in the simulations. It is believed the estimated value of kβHtr is accurate at least within an order of magnitude.