Mechanism and Kinetics of Dithiobenzoate-Mediated RAFT Polymerization. I. The Current Situation

Transcription

1 HIGHLIGHT Mechanism and Kinetics of Dithiobenzoate-Mediated RAFT Polymerization. I. The Current Situation CHRISTOPHER BARNER-KOWOLLIK, 1 MICHAEL BUBACK, 2 BERNADETTE CHARLEUX, 3 MICHELLE L. COOTE, 4 MARCO DRACHE, 5 TAKESHI FUKUDA, 6 ATSUSHI GOTO, 6 BERT KLUMPERMAN, 7 ANDREW B. LOWE, 8 JAMES B. MCLEARY, 9 GRAEME MOAD, 10 MICHAEL J. MONTEIRO, 11 RONALD D. SANDERSON, 9 MATTHEW P. TONGE, 9,12 PHILIPP VANA 2 1 Centre for Advanced Macromolecular Design, School of Chemical Engineering and Industrial Chemistry, The University of New South Wales, Sydney, New South Wales 2052, Australia 2 Institut für Physikalische Chemie, Georg-August-Universität Göttingen, Tammannstraße 6, Göttingen, Germany 3 Laboratoire de Chimie des Polymères, UMR CNRS-UPMC 7610, Université Pierre et Marie Curie, T44, E1, 4, Place Jussieu, Paris Cedex 05, France 4 Research School of Chemistry, Australian National University, Canberra ACT 0200, Australia 5 Institut für Technische Chemie, Technische Universität Clausthal, Erzstraße 18, Clausthal-Zellerfeld, Germany 6 Institute for Chemical Research, Kyoto University, Uji, Kyoto , Japan 7 Laboratory of Polymer Chemistry, Eindhoven University of Technology, P. O. Box 513, 5600 MB Eindhoven, The Netherlands 8 Department of Chemistry and Biochemistry, University of Southern Mississippi, Hattiesburg, Mississippi UNESCO Centre for Macromolecules and Materials, Department of Chemistry and Polymer Science, University of Stellenbosch, Private Bag X1, Matieland 7602, South Africa Correspondence to: P. Vana ( pvana@unigoettingen.de), Vol. 44, (2006) VC 2006 Wiley Periodicals, Inc. 5809

2 5810 J. POLYM. SCI. PART A: POLYM. CHEM.: VOL. 44 (2006) 10 CSIRO Molecular Science, Bag 10, Clayton South, Victoria 3169, Australia 11 School of Molecular and Microbial Sciences, Australian Institute of Bioengineering and Nanotechnology, University of Queensland, Brisbane, Queensland 4072, Australia 12 Key Centre for Polymer Colloids, Chemistry School F11, University of Sydney, New South Wales 2006, Australia Received 24 April 2006; accepted 1 June 2006 DOI: /pola Published online in Wiley InterScience ( ABSTRACT: Investigations into the kinetics and mechanism of dithiobenzoate-mediated Reversible Addition Fragmentation Chain Transfer (RAFT) polymerizations, which exhibit nonideal kinetic behavior, such as induction periods and rate retardation, are comprehensively reviewed. The appreciable uncertainty in the rate coefficients associated with the RAFT equilibrium is discussed and methods for obtaining RAFT-specific rate coefficients are detailed. In addition, mechanistic studies are presented, which target the elucidation of the fundamental cause of rate retarding effects. The experimental and theoretical data existing in the literature are critically evaluated and apparent discrepancies between the results of different studies into the kinetics of RAFT polymerizations are discussed. Finally, recommendations for further work are given. VC 2006 Wiley Periodicals, Inc. J Polym Sci Part A: Polym Chem 44: , 2006 Keywords: controlled living radical polymerization; kinetics (polymer); polymerization mechanism; rate retardation; reversible addition fragmentation chain transfer (RAFT) IUPAC TASK GROUP Task group members (from left to right): T. Fukuda, M. Coote, P. Vana (chairman), B. Charleux, B. Klumperman, M. Monteiro, G. Moad, C. Barner-Kowollik, J. McLeary; (not on the photo: M. Buback, M.Drache,A.Goto,A.Lowe,R.Sanderson, M. Tonge). [Color figure can be viewed in the online issue, which is available at The IUPAC task group \Towards a Holistic Mechanistic Model for RAFT Polymerizations: Dithiobenzoates as Mediating Agents" was established in 2005 and emerged from activities of the \IUPAC Subcommittee on Modeling of Polymerization Kinetics and Processes." The work of the task group is to provide a detailed understanding of the mechanism of RAFT polymerization and at the determination of the associated kinetic coefficients. The main intention of the project is to improve the currently obscure situation via jointly discussing the evidence gathered by different scientific groups and via collating and critically evaluating experimental results for various RAFT systems. In addition, recommendations are developed how to rationally perform and to present future RAFT experiments to guarantee comparability. The project finally aims at providing a more encompassing mechanistic and kinetic picture of the RAFT process.

3 HIGHLIGHT 5811 INTRODUCTION Reversible Addition Fragmentation chain Transfer (RAFT) polymerization has developed into one of the leading controlled/living radical polymerization techniques since its invention by the CSIRO group in It is arguably the most versatile controlled polymerization process with respect to the types of monomers and the reaction conditions that enables the formation of polymer with controlled molecular weight, low polydispersity, and complex polymeric microstructure with relative ease. The success of this powerful technique is demonstrated by a constantly growing body of work that deals with various RAFT processes leading to advanced polymeric materials, as has comprehensively been reviewed very recently. 2,3 Parallel to utilizing the RAFT process for polymer synthesis, a profound mechanistic and kinetic understanding of the individual RAFT-specific reactions was sought from the very beginning of RAFT to establish structure-rate correlations for specific RAFT agents. Such information is mandatory for the design of novel mediating compounds as well as for modeling the polymerization process, whereby costly and time-consuming experiments may be minimized. Shortly after the introduction of RAFT, the CSIRO team proposed a reaction scheme for the process (see Scheme 1) and proved the occurrence of intermediate adduct radicals (3 and 6) by electron spin resonance (ESR) spectroscopy. 4 The overall RAFT polymerization is generally divided into two sets of reactions, that is, the so-called pre-equilibrium, which involves the initial RAFT agent and includes the initialization of the living process, and the main equilibrium between growing and dormant polymer chains (see Scheme 1). In the pre-equilibrium, which takes place in the early stages of the RAFT polymerization, propagating macroradicals 1 which are mainly of oligomeric nature in this period add to the sulfur carbon double bond of the initial RAFT agent 2, resulting in a carbon-centered intermediate RAFT radical 3. The addition of primary initiator-derived radicals to the initial RAFT agent has been generally neglected, but may become prominent in case of high initiating radical concentrations. Such reactions induce the transformation of the leaving group of the initial RAFT agent and may have an impact on the kinetics in the early phase of the RAFT polymerization. The adduct radical 3 formed in the pre-equilibrium in turn undergoes b-scission, either yielding back the reactants or releasing an initiating leaving group radical 5 under concomitant formation of a polymeric dithioester compound 4, which constitutes the dormant species. A similar set of reactions is operating in the main equilibrium, in which a propagating macroradical 1 reacts with the polymeric RAFT agent 4. Recurring RAFT events establish the equilibrium between dormant and living chains, by which living/controlled characteristics are induced in the polymerization. The individual reactions of these equilibriums are described kinetically via so-called addition rate coefficients, k ad, and fragmentation rate coefficients, k b (see Scheme 1). The k ad and k b values of the asymmetric pre-equilibrium have to be considered individually and as being different from those of the main equilibrium, Scheme 1. Basic reaction steps of the Reversible Addition Fragmentation Chain Transfer (RAFT) processes occurring in dithioester-mediated radical polymerization.

4 5812 J. POLYM. SCI. PART A: POLYM. CHEM.: VOL. 44 (2006) Scheme 2. A selection of dithioesters used as RAFT agents in kinetic and mechanistic studies. because chemically different radical species are involved as attacking and leaving group radicals. The main equilibrium, on the other hand, is symmetrical apart from small differences in the chain length of the participating macroradicals, which, however, were generally neglected up to now. The basic reaction scheme depicted in Scheme 1 is superimposed on that of a conventional radical polymerization including the elementary steps of initiation, propagation, and termination, by which the experimental ease of simply adding a RAFT agent to a conventional polymerization mixture is reflected. The basic RAFT reaction steps given in Scheme 1 are sufficient to describe the living/controlled behavior observed in RAFT polymerizations, which leads to low polydispersities of the generated polymer and increasing molecular weights with monomer conversion. As the ideal RAFT process depicted in Scheme 1 does not alter the propagating radical concentration, the rate of polymerization in the steady state remains unchanged in comparison with a conventional polymerization system. The mediating compounds employed in RAFT polymerizations are thiocarbonyl thio compounds, Z C( ¼S)S R, which have been developed in great structural variety with respect to their leaving R-groups and to their stabilizing Z-moieties. 5,6 Many of these

5 HIGHLIGHT 5813 Figure 1. Pseudo first-order rate plots for bulk polymerizations of methyl acrylate at 80 8C mediated by CDB in various concentrations and initiated by AIBN ([AIBN] ¼ mol L 1 ). (Reproduced from ref. 10, Copyright 2005, Elsevier.) RAFT agents induce controlled polymerizations, exhibiting kinetics that can be satisfactorily described by Scheme 1. Theoretically, the degenerative nature of the main equilibrium yields educts and products that are chemically identical and the propagating radical concentration remains unaffected, which complicates the determination of the individual rate parameters. Kinetic anomalies, however, such as unexpected reductions of the propagating radical concentration, which occur in some RAFT systems, provide additional information for accessing the kinetics of the elementary steps. Dithiobenzoates, in which Z is phenyl (see Scheme 2), have generally been found to be very effective RAFT agents for controlling the molecular weight in polymerizations of, for example, styrenics, acrylates, methacrylates, and acrylamides, 7 and are thus used very frequently. In addition, they have been found to exhibit nonideal kinetic behavior, and thus are metaphorically speaking the \fruit fly" for the study of kinetics and mechanism of the RAFT process. It should, however, be stressed that findings for dithiobenzoate-mediated polymerizations presented in this article cannot be generalized a priori to all RAFT processes, as the stabilization of the intermediate radical by the phenyl moiety provides specific characteristics that may not occur with RAFT agents carrying other Z-groups. However, such findings may be used as a guide to possible side reactions that may occur with lower frequencies for other RAFT agents. Soon after the advent of RAFT, it was recognized by the CSIRO group that the rate of polymerization, R p,in RAFT polymerizations using dithiobenzoates as the mediating agents was retarded, that is, R p decreased with increasing initial RAFT agent concentrations. 8 Two individual rate retarding effects were identified: (i) an induction period 9 in the initial phase of the polymerization with virtually no polymerization activity (in literature about RAFT kinetics sometimes also denoted as \inhibition"), (ii) followed by a polymerization being slower in rate in comparison with the corresponding conventional radical polymerization system in the absence of RAFT agent, an effect termed rate retardation. 9 These two kinetic effects are demonstrated in Figure 1, in which pseudo first-order rate plots for CDB-mediated methyl acrylate polymerizations are depicted. Both the induction period and the rate retardation effect clearly depend in their extent on the concentration of the initial RAFT agent. When performing RAFT polymerizations using a polymeric RAFT agent, that is, effectively skipping the pre-equilibrium, no induction period can be observed (see Fig. 2), whereas rate retardation still occurs. The induction period can hence be attributed to the preequilibrium. Pre-equilibrium characteristics, however, seem not to be the cause of rate retardation in the main equilibrium. In its early work, 8 the CSIRO team attributed the induction period to a low reinitiation rate of the leaving group radical 5 and explained the rate retardation occurring in the main equilibrium by slow fragmentation of the polymeric intermediate radical 6. Soon after, Monteiro and de Brouwer 12 introduced the concept of irreversible termination between intermediate radicals 3 and 6 with propagating radicals, termed cross-termination (see Scheme 3), and postulated that such termination Figure 2. Pseudo first-order rate plot for the bulk polymerization of styrene at 60 8C mediated by pstdb (M n ¼ 1100 g mol 1, PDI ¼ 1.08); [AIBN] 0 ¼ 10 2 mol L 1 ; pst-br was added to keep the total polymer concentration constant at mol L 1. (Reproduced from ref. 11, Copyright 2004, American Chemical Society.)

6 5814 J. POLYM. SCI. PART A: POLYM. CHEM.: VOL. 44 (2006) Scheme 3. Termination (cross-termination) between propagating and intermediate radicals of the pre-equilibrium (upper part) and of the main equilibrium (lower part). A possible reversible pathway [see section Reversible Termination of the Intermediate Radical (Nonstationary State Model)] is indicated in grey. reactions are inducing the rate retardation effect in RAFT polymerization. They performed simulations of rate of polymerization data on the basis of a kinetic model that includes cross-termination with the associated rate coefficient assumed being equal to that of conventional termination, that is, k t,cross ¼ k t, and thereby obtained estimates for the fragmentation rate coefficient rate k b ¼ s 1, by inserting k ad ¼ L mol 1 s 1 that was estimated independently by Goto et al. 13 using a model-free experimental method, in which the addition reaction was kinetically isolated from the other reactions. 14 This set of coefficients combined to an equilibrium constant of K ¼ k ad /k b ¼ 80 L mol 1. In parallel work, the CDB-mediated styrene polymerization was also modeled by the CAMD team, 15 using the original kinetic model given in Scheme 1, without taking any side reactions of the intermediate radicals into account, that is, assuming k t,cross ¼ 0. In that study, rate and molecular weight data were modeled simultaneously, which allowed the estimation of both RAFTspecific rate coefficients, and values of k ad ¼ L mol 1 s 1 and of k b ¼ s 1 were reported, resulting in K ¼ L mol 1 for the main equilibrium. Comparison of these two sets of values revealed a discrepancy of more than six orders of magnitude in the fragmentation rate, whereas the addition rate coefficient differed to a much lower extent. The enormous difference in the reported k b values, stemming from different assumptions regarding the size of k t,cross, stimulated a number of studies into this issue: Fukuda and coworkers, 11,16 for instance, used rate of polymerization data in conjunction with intermediate radical concentrations obtained via ESR spectroscopy for the experimental determination of the main equilibrium constant and obtained K ¼ 55 L mol 1, which is very close to the value reported by Monteiro et al. (see above) and supported the concept of cross-termination. On the other hand, the finding of low k b values obtained by modeling RAFT polymerizations without assuming termination of the intermediates was supported by quantum-chemical ab initio calculations by Coote et al. 17,18 for small model species, as occurring in the pre-equilibrium, as well as by a nonconstant rate of polymerization in the early polymerization phase, 10,19 alongside mass spectroscopic evidence. 20,21 The proposed slow fragmentation of the RAFT intermediate radicals, however, could not be harmonized with ESR-derived intermediate radical concentrations 4,10,16,22 24 and the quasistationary state in radical concentrations during the main equilibrium. 16 These apparently contradictory findings initiated a lively scientific discussion about the underpinning cause of rate retardation in RAFT. 25,26 It should, however, be stressed that in conjunction with the associated kinetic model both of the aforementioned sets of rate parameters are capable of describing RAFT polymerizations equally well within experimental uncertainties: The rate of polymerization, the molecular weight distribution of polymeric RAFT agent, and the evolution of its molecular weight with time can be predicted satisfactorily via both of the mechanisms assumed for rate retardation. 27,28 Differences between the models become noticeable only when considering additional experimental quantities such as concentrations of individual radical species and type and concentration of potential side-products. When aiming at getting further insight into the RAFT mechanism, it is hence mandatory to trace these species, which are generally very low in concentration. This situation already demonstrates the challenge for studies into the RAFT mechanism. The fact that the intense research efforts undertaken by various groups could not resolve the issue during the last 5 years demonstrates the complexity of the problem and suggests that some key information is not yet available for the complete understanding of all aspects of the RAFT mechanism. It thus appeared beneficial to establish a task group, which was founded within the framework of IUPAC, in which the key points of the controversy are the subject of common consideration. In the present article, the work performed on the rate retardation effect in dithiobenzoate-mediated RAFT polymerization, including mechanistic studies, will be reviewed up to the beginning of 2006 and points of agreement will be clarified. The critical evaluation of the experimental and theoretical data presented in this communication will help to identify apparent discrepancies

7 HIGHLIGHT 5815 and to precisely define the outstanding inconsistencies regarding the mechanism of dithiobenzoate-mediated RAFT polymerization. We are well aware that the controversy may not be settled immediately, but it is our intention to provide a well-defined starting point for further work into this issue and to stimulate novel approaches for getting better insight into this puzzling situation. Recommendations will be given on how to rationally perform, analyze, and present future kinetic data. THE FUNDAMENTAL CAUSE OF RATE RETARDATION Rate retardation, caused by a lower propagating radical concentration than observed in an analogous reaction without RAFT agent, is seen in many RAFT polymerizations. It seems important to explicitly state the common agreement on what fundamentally causes the rate retardation effect in RAFT: In all theories presented up to now, the retardation effect increased with increasing stability of the intermediate radicals 3 and 6. On the one hand, greater stability of these adduct radicals induces an increase in their concentrations, resulting in an enhanced probability of terminating side reactions with propagating radicals or themselves, whereby the polymerization rate is reduced. On the other hand, an extremely high stability of the intermediates induces rate retardation by itself, as the establishment of the stationary propagating radical concentration is delayed for an extensive time period during which the rate of polymerization is lower than in the stationary state. When the stationary state conditions are reached, however, no further retardation is induced by this mechanism. In either model, rate retardation is enhanced by an increase in intermediate radical stability. The pronounced rate retardation effects occurring in dithiobenzoate-mediated polymerizations may thus be assigned to the stabilization of the intermediate radical by the aromatic phenyl-moiety, which allows for the delocalization of the radical site into the aromatic system, 29 as depicted in Scheme 4. This finding was verified by employing CPDA as RAFT agent (see Scheme 2), in which the phenyl Z-group is replaced by a benzyl group. 30 This mediating compound exhibits a much lower capability of stabilizing the associated intermediate radical, because the delocalization of the radical center into the aromatic system is effectively prevented. The rate retardation effect using CPDA is significantly lower than in the dithiobenzoate system. A high stability of the intermediate radical is, however, a strong thermodynamic driving force for the addition reaction of propagating radicals toward the RAFT agent. It may hence be surmised that most of the RAFT agents that exhibit a high addition rate toward propagating radicals and which are consequently very effective controlling compounds, such as the dithiobenzoates, are inherently prone to rate retardation. In addition to these energetic aspects, the pronounced rate retardation occurring with dithiobenzoates may additionally be caused by the structural effect of radical delocalization, 29 whereby the radical center is shifted to more exposed sites of the intermediate radical, such as the para-position of the phenylgroup (see Scheme 4), which may enhance otherwise sterically demanding side reactions (see section Composite Slow Fragmentation/Irreversible Termination Model: Varying K for the Pre-equilibrium and Main Equilibrium). MODELS FOR RATE RETARDATION In what follows, models for explaining the rate retardation in dithiobenzoate-mediated RAFT polymerization will be described in detail and key differences as well as implications for expected experimental behavior will be illustrated. In sections Slow Fragmentation (Nonstationary State Model) and Irreversible Termination of Intermediate Radicals (Stationary State Model), the two representative models suggested earlier, that is, slow fragmentation and irreversible termination of the intermediate radicals, are introduced and the time evolution of the propagating radical concentration expected for each of these models are discussed. In sections Reversible Termination of the Intermediate Radical (Nonstationary State Model) and Composite Slow Fragmentation/ Scheme 4. Resonance structures of the intermediate radical occurring in dithiobenzoatemediated RAFT polymerization.

8 5816 J. POLYM. SCI. PART A: POLYM. CHEM.: VOL. 44 (2006) Irreversible Termination Model: Varying K for the Preequilibrium and Main Equilibrium, two models that have been proposed to harmonize the apparently conflicting evidence are described. All simulation results presented in this section are for systems employing polymeric RAFT agent, that is, they refer to a main equilibrium situation. Slow Fragmentation (Nonstationary State Model) This model 8,15,19,20,26,27,30 35 includes five types of reactions, that is, the RAFT equilibrium reactions (see Scheme 1), conventional initiation, propagation, and conventional termination between propagating radicals. In a system with monomer, conventional initiator, and polymeric RAFT agent 4, the concentration of propagating and intermediate radicals is zero at t ¼ 0. When the polymerization is started by allowing conventional initiation to occur, the produced propagating radical 1 will undergo RAFT addition reactions to yield the intermediate radical 6, which will accumulate until the addition fragmentation equilibrium is established. During this period, most of the propagating radicals will be stored as intermediate radicals, which may not propagate and therefore result in rate retardation. When the equilibrium constant K is large (i.e., fragmentation is slow), the system will spend a long time-span in the nonstationary state until a large amount of 6 is accumulated and a significant part of the polymerization will consequently occur in a nonstationary state. Figure 3(a) shows an example of the time evolution of the propagating radical concentration, simulated for slow fragmentation with k b ¼ s 1, along with the parameters k ad ¼ L mol 1 s 1, k t ¼ L mol 1 s 1, the rate of initiation being R i ¼ mol L 1 s 1, and a RAFT agent concentration of 0.01 mol L 1, which is typical for a dithiobenzoate-mediated RAFT system. Figure 3(a) illustrates that the propagating radical concentration increases with time, constituting a pronounced nonstationary state region, 27 until it approaches the value found in the conventional system with no further rate retardation. The period of the nonstationary state depends on K and is longer for a larger K value (slower fragmentation). The nonstationarity of this model currently impedes any analytical description of the process, because of the complex and coupled reaction scheme. Figure 3. Simulated time evolution of the propagating radical concentration, [1], in a retarded RAFT polymerization (full lines) for (a) slow fragmentation of the intermediate (k b ¼ s 1, k t,cross ¼ 0), and (b) for fast fragmentation and irreversible cross-termination of the intermediate radicals (k b ¼ s 1, k t,cross ¼ L mol 1 s 1 ). The dotted line indicates the propagating radical concentration of the corresponding conventional/nonretarded polymerization. The other simulation parameters are: [4] ¼ 0.01 mol L 1, R i ¼ mol L 1 s 1, k ad ¼ L mol 1 s 1, and k t ¼ L mol 1 s 1. Irreversible Termination of Intermediate Radicals (Stationary State Model) This model 10 12,16,22 25,28,36 44 includes all the reactions mentioned in section Slow Fragmentation (Nonstationary State Model), as well as irreversible cross-termination between propagating and intermediate radicals (see Scheme 3), and irreversible self-termination between two intermediate radicals (see Scheme 5). This model focuses on a stationary state in which all reversible reactions are in equilibrium (or in quasiequilibrium) and the concentrations of all radical species are (approximately) invariant with time. Once the steady state is attained, the mathematics of the reaction kinetics is dramatically simplified. In case the addition fragmentation quasiequilibrium approximately holds and the radical concentrations are assumed to be in a stationary state, the po-

9 HIGHLIGHT 5817 Scheme 5. Self-termination reactions between intermediate radicals of the main equilibrium proceeding via different resonance structures as depicted in Scheme 4. lymerization rate, R p, in a RAFT polymerization is given by eq 1, 11 R p R p;0 ¼ f1 þ 2ðk t;cross =k t ÞK½RAFTŠþðk t;self =k t ÞK 2 ½RAFTŠ 2 g 1=2 ð1þ where R p, 0 is the polymerization rate of the conventional/ nonretarded system, k t,cross and k t,self are the rate constants of the cross-termination (Scheme 3) and the self-termination (Scheme 5), respectively, and [RAFT] denotes the concentration of dithioester groups. According to this model, a RAFT system suffers a retardation in R p by a factor of (1 þ 2(k t,cross /k t )K[RAFT] þ (k t,self /k t )K 2 [RAFT] 2 ) 1/2. To further facilitate the kinetic analysis, borderline situations may be considered: In the case that cross-termination is the main cause for rate retardation, k t,self in eq 1 may be set to zero, yielding eq 2, 11,16,22 R p ¼ R p;0 f1 þ 2ðk t;cross =k t ÞK½RAFTŠg 1=2 ð2þ whereas in the case of self-termination being prominent, eq 1 with k t,cross ¼ 0 becomes 11 R p ¼ R p;0 f1 þðk t;self =k t ÞK 2 ½RAFTŠ 2 g 1=2 ð3þ Figure 3(b) shows the propagating radical concentration versus time traces simulated for a system with fast fragmentation of the RAFT intermediate (k b ¼ s 1 ) and with irreversible intermediate termination occurring (k t,cross ¼ L mol 1 s 1 ). All other parameters are the same as in Figure 3(a). The concentration of 1 reaches a stationary state within a couple of seconds after the onset of the polymerization and R p is retarded by a constant factor throughout the entire polymerization. This clear difference in the retardation behavior between the slow fragmentation mechanism (in a nonstationary state) and the irreversible intermediate termination mechanism (in a stationary state) is an important probe for the underlying mechanism. For the main equilibrium, 16 both the propagating and intermediate radical concentrations in pstdb-mediated styrene polymerization were found to be in a steady state, which is established quickly and remains over a long time period [see Fig. (4)]. This finding resulted in the conclusion of a fast fragmenting intermediate and cross-termination being responsible for the rate retardation in the main equilibrium. 11,16 For the early polymerization period, however, a nonsteady state behavior of both the propagating radical concentration 19 [concluded from a nonconstant rate of polymerization as obvious from inspection of Fig. (8)] and a nonsteady state in the intermediate radical concentrations 22,23 was identified when using CDB as the mediating agent.

10 5818 J. POLYM. SCI. PART A: POLYM. CHEM.: VOL. 44 (2006) before the equilibrium (or quasiequilibrium) is reached, but after it has been established, the radical concentrations will be in a stationary state and the system would kinetically appear as if there was no reversible termination. In other words, the reversible termination has no effect on the stationary radical concentrations. This model therefore resembles in essence the slow fragmentation model with respect to the nonstationarity. Figure 5 illustrates the situation by simulated evolutions of the Figure 4. (a) The intermediate radical concentration, [P- (X )-P] (¼[6]), and (b) the propagating radical concentration, [P ](¼[1]), versus time for the bulk polymerization of styrene at 60 8C mediated by mol L 1 pstdb (M n ¼ 1100 g mol 1, PDI ¼ 1.08); [AIBN] 0 ¼ mol L 1. (Reproduced from ref. 16, Copyright 2002, American Chemical Society.) Reversible Termination of the Intermediate Radical (Nonstationary State Model) This model 19,20,26,27,30,31,33 35 consists of the reactions mentioned in section Irreversible Termination of Intermediate Radicals (Stationary State Model), but assumes that cross-termination and/or self-termination are reversible (see Schemes 3 and 5). The introduction of reversibility was driven by the idea of harmonizing low intermediate radical concentrations measured by ESR with the long life-times of intermediates predicted by the slow fragmentation model. 30 Under a scenario of reversible intermediate termination, it will take a certain time Figure 5. Simulated time evolution of the propagating radical concentration, [1], in RAFT polymerization at various levels of reversibility of cross-termination (k t,cross ¼ L mol 1 s 1 ), indicated by k cross values, (a) for the slow fragmentation (k b ¼ s 1 ), and (b) for the fast fragmentation (k b ¼ s 1 ) of the intermediate radical. The dotted line indicates the propagating radical concentration of the corresponding conventional/nonretarded polymerization. All other simulation parameters are: [4] ¼ 0.01 mol L 1, R i ¼ mol L 1 s 1, k ad ¼ L mol 1 s 1, k t ¼ L mol 1 s 1.

11 HIGHLIGHT 5819 propagating radical concentration on the example of reversible cross-termination for both the models of fast and slowly fragmenting intermediates. As the reversible termination reactions are hypothesized reactions with the associated rate coefficients being completely unknown, a variety of rate coefficients describing the reversibility of the termination reactions was chosen from the range, where the kinetic effects of interest occur. True values for the reversible termination reactions, if existing, may be different. In the case of slowly fragmenting intermediates [Fig. 5(a)], the introduction of reversible cross-termination further reduces the propagating radical concentration, while retaining the nonsteady state characteristics. When k cross is chosen as s 1, the intermediate radical concentration is calculated to be mol L 1 after 2500 min, which is about 25 times smaller than the value calculated for the slow fragmentation model {[6] ¼ mol L 1, calculated for the scenario depicted in Fig. 3(a)}, but still four orders of magnitude higher than the value calculated for the fast fragmentation model {[6] ¼ mol L 1, calculated for the situation depicted in Fig. 3(b)}. The propagating radical concentration, however, has in case of k cross ¼ s 1 already dropped to only 3% of the corresponding nonretarded value. It is hence clear that in the case of slow fragmentation, the intermediate radicals cannot be stored in reversible cross-termination products effectively without concomitantly lowering the propagating radical concentration. In case of fast fragmenting intermediates [Fig. 5(b)], the introduction of reversible cross-termination leads to hybrid characteristics between nonsteady state and steady state behavior: relatively low values of k cross in essence resemble the irreversible termination model, whereas relatively high k cross values, for example, of k cross ¼ 10 3 s 1 and above, do not induce any retardation. All in-between situations exhibit, as described above, nonstationary state characteristics. The kinetic picture is very similar in the case of reversible self-termination, and simulated propagating and intermediate radical concentration versus time curves for the range of s 1 < k self < s 1 for the slow fragmentation model (with k b ¼ s 1 )andforthe range of s 1 < k self < s 1 for the fast fragmentation model (with k b ¼ s 1 )resemblethe characteristics of the curves given in Figure 5. Composite Slow Fragmentation/Irreversible Termination Model: Varying K for the Pre-equilibrium and Main Equilibrium In actual polymerizations usually low molecular weight RAFT agents are employed, in which the leaving group is different both in its chemical nature and in its molecular weight to the macroradical leaving group that occurs in the main equilibrium. In the pre-equilibrium (see Scheme 1), which includes the initialization phase of the RAFT polymerization during which the entire initial RAFT agent is converted to its single monomer and higher adducts, three important facts have to be considered: 1. The kinetic coefficients for the addition and fragmentation reaction in the pre-equilibrium, described by k ad,1, k ad,2, k b,1, and k b,2 (see Scheme 1), may markedly be different in comparison with k ad and k b of the main equilibrium, as suggested by the CAMD group, 27 by Calitz et al. 23 and Drache et al., 10 which may result in largely different equilibrium constants in comparison with the main equilibrium. 2. The leaving group radical from the original RAFT agent occurring in the pre-equilibrium is different in its chemical nature from the main equilibrium macroradical leaving group. A difference in the addition rate toward the monomer (reinitiation vs. propagation rate) may well be envisaged. 8, The propagating radicals are controlled in molecular weight and have oligomeric chain lengths during the early polymerization phase. As the chain-length dependencies of kinetic coefficients are known to be especially pronounced in the very short chain-length regime, a large impact of chain length on the overall kinetics has to be considered under pre-equilibrium conditions. 23 Especially the termination rate coefficient, which is by more than one order of magnitude higher for small radicals in comparison with large macroradicals, 48,49 slows down the polymerization rate in the early polymerization phase. This effect, however, is not specific for a distinct RAFT agent, but common to all controlled RAFT polymerizations. A chain-length dependence of k b may also be envisaged, as it is known that fragmentation reactions in dithioesters are enhanced in rate when the fragmenting radicals are larger. 6,13,50 We have outlined above that two different models are capable of describing the observed rate retardation phenomena in the RAFT process. A situation where both models are fused into one more encompassing approach may also be envisaged. In fact, there is some experimental and theoretical evidence (see later) that does indeed suggest that such a composite model may be capable of explaining all available experimental data. Composite models of varying complexity can be hypothesized: (i) Initially, one may assume that K is differing greatly for the pre-equilibrium and main equilibrium, while at the same time being chain-length dependent. The equilibrium constant may initially be very large for relatively short polymer chains formed at low monomer

12 5820 J. POLYM. SCI. PART A: POLYM. CHEM.: VOL. 44 (2006) conversions thus leading to a nonsteady state situation and may then gradually decrease with increasing chain length of the polymeric RAFT agent. Also the opposite may occur, that is, increasing K values with increasing chain-length, as obtained from ESR spectroscopic studies into the CDB-mediated styrene polymerization. 23 The consequence of such a scenario is in any case that whether a RAFT polymerization is commenced with either initial or polymeric RAFT agent would significantly influence the polymerization kinetics, an observation that has been confirmed experimentally. 32,33 The chain-length dependence of the equilibrium constant can be caused by a chain-length dependent, potentially diffusion-controlled addition reaction as well as by a chain-length dependent fragmentation rate coefficient. (ii) In addition to a variable equilibrium constant, irreversible termination of the intermediate radicals may occur during both the pre-equilibrium and the main equilibrium, again with chain-length dependent rate coefficients. It is mandatory to comment on a mechanistic consequence arising from large equilibrium constants in conjunction with cross-termination: A large equilibrium constant leads to relatively high concentrations of intermediate radicals and thus to a high rate of cross-termination. Under such a scenario, where cross-termination proceeds via combination, significant amounts of star polymer material (either 3- or 4-armed, see section ESI-MS, MALDI-TOF) are expected. 10,21 (iii) The possibility of reversible termination of intermediate RAFT radicals playing a role in the RAFT process has been proposed for some time, 30 [see section Reversible Termination of the Intermediate Radical (Nonstationary State Model)], and there are experimental observations that are in agreement with such a possibility. 23,31,33 Reversible termination of the intermediate radical possibly proceeds via self-termination pathways, as depicted in the middle and lower part of Scheme 5. A termination reaction of intermediate RAFT radicals with their radical centers being delocalized into the aromatic ring may suffer less steric hindrance and the stability of the resonance-stabilized radical may allow for reversibility. 29 When using a dithiobenzoate with a substituent on the para-position of the Z-group, such as CPMDB (see Scheme 2), a significant reduction of the rate retardation effect has been observed. 35 As the para-methyl group is not expected to change the stability of the intermediate significantly, but makes the para-position less prone to radical attack due to sterical congestion, an involvement of delocalized radical sites in rate retarding reactions can be surmised. Direct evidence for such reactions, however, has not yet been found. In addition, a satisfactory embedment of such reaction pathways into kinetic schemes has not been achieved so far [see section Reversible Termination of the Intermediate Radical (Nonstationary State Model)]. APPROACHES FOR DETERMINING RATE COEFFICIENTS In this section, the currently available approaches for accessing RAFT specific rate coefficients, that is, k ad, k b, K ¼ k ad /k b, k t,cross and k t,self, along with their advantages and limitations are described, and a brief survey of the obtained values is given. A full collection of the rate coefficients will be presented in a subsequent paper of this IUPAC project. It should be noted that the rate coefficients presented herein were obtained for various systems and conditions. As stressed above, the rate coefficients may differ for polymer and low-mass mediating compounds and may depend on the chain length of the participating species. The results for different systems therefore have to be considered independently. Determination of k ad and k b High Level Ab Initio Calculation Quantitative information about the kinetics and thermodynamics of chemical reactions can be obtained via ab initio molecular orbital theory. 51 Besides the rates and equilibrium constants, the charge distributions, spin densities, geometries, and vibrational frequencies of all species including transition state structures may be estimated by such calculations. The ab initio approach consequently represents a useful partner to experiment. Its advantage for studying RAFT reactions is that the calculated predictions do not depend upon any kinetic model or empirical data, other than laws of quantum mechanics. Its predicted rate coefficients are therefore independent of the models described in section Models for Rate Retardation. The principle disadvantage of quantum chemistry is that the accuracy of the predictions is heavily dependent upon the simplifications and approximations introduced in solving the Schrödinger equation. Given an infinite basis of orbitals, a wave-function that includes contributions from all possible configurations of those orbitals, and the computer power necessary to process the data, one could in principle predict the chemical behavior of any chemical system with absolute accuracy. In practice, of course, this is not feasible and it is necessary to introduce simplifications, which are a potentially large source of error. It is extremely important to note that there is a range of possible methods to use and that these vary considerably in their accuracy and their computational expense. Errors can typically range from the sub-kj level (for methods such as W1 or W2) to more

13 HIGHLIGHT 5821 than 50 kj mol 1 (for DFT or semiempirical methods), and are dependent on the type of chemical reaction under investigation. Moreover, the most accurate methods are highly time-expensive, with their computational cost scaling exponentially with the size of the chemical system. Hence, applying computational chemistry to polymeric reactions necessarily involves a compromise in which reliable low-cost procedures are applied to small chemical models of the polymeric reaction under investigation. Both the method and the model reaction, however, must be chosen carefully on the basis of assessment studies. Such assessment studies have been recently performed by Coote and coworkers, both for the RAFT-relevant case of radical addition to S ¼C bonds, and for other reactions of significance in radical polymerization, such as propagation and chain transfer In general, it appears that geometries and frequencies can be calculated at relatively low levels of theory such as B3- LYP/6-31G(d), 53 provided transition structures are corrected via IRCmax. 57 However, very high-level composite procedures such as W1 are essential for accurate absolute values of the barriers and enthalpies. Low-cost approximations to this level of theory (to within kcalaccuracy) can be obtained via an ONIOM-based procedure in which the inner core is studied at W1, the core at G3(MP2)-RAD and the full system at ROMP2/6-311þG(3df,2p). 54 It should be stressed that popular DFT-methods (such as B3-LYP and BB1K) fail comprehensively to model the energetics of these polymerization-relevant reactions with errors in excess of 50 kj mol 1 in some cases. 58 As in the case of propagation, 59 the harmonic oscillator approximation is not adequate and it is therefore important to treat low frequency torsional modes as hindered internal rotations. 18,52 It is also advisable to use variational transition state theory rather than standard transition state theory, though the latter can be adopted for order of magnitude estimates. 52 Using these recommended procedures, chemical accuracy is now achievable for small radical reactions. For example, the experimental equilibrium constant for t- butyl radical addition to di-t-butyl thioketone at 25 8C was recently reproduced to within a factor of two. 60 In the case of propagation, calculations using dimer radicals as models of the propagating species have reproduced the experimental propagation rate coefficients for the polymerization of vinyl chloride and acrylonitrile to within chemical accuracy. 54 With respect to the RAFT process, the k ad, k b,andk values for a variety of R- and Z-groups of small compounds, as occurring in the pre-equilibrium, have been calculated by Coote and collaborators using high-level ab initio methods ,52,61 (A complete listing of the data from these various studies, updated to a consistent level of theory, is provided in ref. 62.) The simulations suggested that k b and K depend strongly on the nature of R and Z, and can range over 10 orders of magnitude at 60 8C. For Z ¼ Ph, the K for the RAFT reaction R þ S¼C(Ph)SCH 3 $ RSC (Ph)SCH 3 was, for example, Lmol 1 (30 8C) for R ¼ cumyl, L mol 1 (60 8C) for R ¼ benzyl, 18 and Lmol 1 (60 8C) for R ¼ CH 2 (COOCH 3 ). 18 The K values calculated for these pre-equilibrium type reactions involving small species generally agree with those expected for slow fragmentation (typically, 10 6 to 10 7 Lmol 1 ). 15,19 Very recently, calculations performed for R groups of different oligomeric chain length indicated a pronounced impact of the leaving group size on the obtained K values, which were found going from L mol 1 when R ¼ C(CH 3 ) 2 CN (chain length ¼ 0), over Lmol 1 when R ¼ CH(Ph)CH 2 C(CH 3 ) 2 CN (chain length ¼ 1), to L mol 1 when R ¼ CH(Ph)CH 2 CH(Ph)- CH 2 C(CH 3 ) 2 CN (chain length ¼ 2) in a simplified model reaction in which the nonparticipating R 0 -group is CH To date, it is not fully clear, which physical reason may be the cause of such a pronounced effect, making generalization to all RAFT systems difficult. The finding of a pronounced chain-length dependency of K, however, highlights the importance of considering the pre-equilibrium and the main equilibrium independently. Parameter Fitting of Polymerization Rates and Molecular Weight Distributions Based on Assumed Models The k ad and k b values may be estimated as adjusting parameters in fitting simultaneously the experimental rate of polymerization data and molecular weight distributions using computational methods based on a selected model. 6,10,12,15,19,28,30 The limitation of such an approach is that only model-dependent parameters are obtained. Such modeling was first applied to the styrene/cdb system at 60 and 80 8C: The assumption that k t,cross ¼ k t (irreversible cross-termination) 12 and k t,cross ¼ 0 (slow fragmentation), 15 respectively, resulted in largely different k b values, as described in the Introduction. It should however be stressed that using the same elementary reactions and kinetic parameters in different computational methods generally yields identical results, that is, variations in the computational methods cannot be the cause of the enormously differing k b values. Interestingly, Chong et al. 6 reported that the styrene polymerization kinetics at 60 8C, mediated by various dithiobenzoates, may appropriately be simulated by using the basic RAFT reaction scheme (see Scheme 1) and fragmentation rate coefficients that are typical for the crosstermination model, that is, k b ¼ s 1, 6 as these authors observed smaller rate retardation effects than

14 5822 J. POLYM. SCI. PART A: POLYM. CHEM.: VOL. 44 (2006) other groups. This observation signifies that the parameter fitting method described herein is extremely sensitive to the underlying kinetic data. Recently, a nonconstant polymerization rate in the initial phase of the CDB-mediated RAFT polymerization at 30 8C was observed experimentally (see Fig. 8) for the low monomer conversion regime. 19 The data was perfectly fitted by the slow fragmentation model, yielding K ¼ 10 7 L mol 1, assuming one K value for both the low-mass and polymeric RAFT species, 19 but could also be fitted equally well by the irreversible cross-termination model, assuming different K values for the pre-equilibrium and main equilibrium. 10 Simultaneously fitting a larger number of independent experimental probes provides access to an increased number of kinetic parameters. Based on this concept, the simultaneous fitting of the polymerization rate and molecular weight data via a Monte Carlo method was applied to the methyl acrylate/cdb system at 80 8C for extracting k ad and k b values both for the low-mass and high-mass (polymer) regions separately as well as for estimating k t,cross. 10 Combined Analysis of Polymerization Rate and ESR-Derived Intermediate Radical Concentration In case the addition fragmentation quasiequilibrium approximately holds over a relatively short timescale, K is given by eq 4. 11,16 K ¼ ½6Š ð4þ ½1Š½4Š K can thus be obtained by measuring both the propagating radical concentration, [1], and the intermediate radical concentration, [6], at a known concentration of the polymeric RAFT agent, [4]. The propagating radical concentration, for instance, may be calculated from the rate of polymerization in conjunction with independently obtained k p values, and the intermediate radical concentration can be measured directly by ESR spectroscopy, 4 which is a well-established method for quantifying radical concentrations. The major advantage of kinetic analysis via ESR relates to the fact that this experimental method under the assumption of a steady state provides direct access to K without any kinetic model assumptions. This approach to determine K was applied to dithiobenzoate-mediated styrene systems. 16,23,24 For the styrene/ pstdb system at 60 8C, 16 the radical concentrations were found to be stationary from an early stage of polymerization (see Fig. 4), that is, the addition fragmentation quasiequilibrium was quickly reached, and K was determined to be 55 L mol 1. By insertion of the independently determined k ad value, k b (¼ k ad /K) was estimated to be s 1, 13 suggesting that fragmentation is fast in the main equilibrium. In the styrene/cdb system, 23,24,40 K was found to significantly change when going from the lowmass region to the polymer region, demonstrating a different K for the cumyl and polystyryl leaving groups (see section Composite Slow Fragmentation/Irreversible Termination Model: Varying K for the Pre-equilibrium and Main Equilibrium). The intermediate radical concentration was also measured for dithiobenzoate-mediated acrylate systems, 4,10,22 and was found to be between 10 6 and 10 5 mol L 1 under the studied conditions. These values are significantly higher than the 10 7 mol L 1 obtained for the styrene systems 4,16,23,24 determined under similar experimental conditions, which implies a significantly larger K for acrylates than for styrene. The intermediate radical concentration was found to be equal or less than 10 5 mol L 1 in all ESR studies reported so far, 4,10,16,22 24,63,64 which is clearly below the 10 4 to 10 3 mol L 1 expected for slow fragmentation. 27 Because of the ESR instrumental sensitivity, there can be some uncertainties in the low concentration ranges, that is, at 10 8 to 10 7 mol L 1, but an error of orders of magnitude, that is, 10 7 to 10 5 mol L 1 versus 10 4 to 10 3 mol L 1 seems very unlikely. Time-Resolved ESR after Pulsed Laser Initiation After laser-pulse-induced formation of initiating radicals during a RAFT polymerization, the formation and decay of the intermediate RAFT radical can be monitored via ls-time-resolved ESR spectroscopy. 65 The build-up of the intermediate radical concentration with time reflects the addition of radicals to the dithioester groups, the decay kinetics of the intermediate species is governed by both the fragmentation rate of the intermediate RAFT radical and by irreversible termination reactions of any radical species present in the system. k ad and k b values can be determined by fitting intermediate radical concentration versus time traces to a kinetic scheme, which exclusively considers macroradicals, dithioester compounds, and intermediate radicals. The chain length of the participating species needs not to be considered in the modeling, as the size of the intermediate radical does not significantly change during one single-pulse experiment. 49 Whether or not cross-termination takes place has no effect on k ad and influences k b by less than one order of magnitude. As a single-pulse technique, the method may be applied at any time and monomer conversion, respectively, whereby chain-length dependent rate parameters may be deduced and the pre-equilibrium and the main equilibrium may be studied separately. Estimating k b is restricted to systems where the extent of self-termination is known. In the case that the rate coefficient of self-termination is included in the kinetic modeling as a fit parameter, the resulting k b values decrease with an increasing extent of self-termination.