A Journey Into Metal–Carbon Bond Homolysis Rinaldo Poli

A journey into metal–carbon bond homolysis Rinaldo Poli

To cite this version:

Rinaldo Poli. A journey into metal–carbon bond homolysis. Comptes Rendus Chimie, Elsevier Mas- son, 2021, 24 (1), pp.147-175. 10.5802/crchim.73. hal-03209704

HAL Id: hal-03209704 https://hal.archives-ouvertes.fr/hal-03209704 Submitted on 27 Apr 2021

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Comptes Rendus Chimie

Rinaldo Poli A journey into metal–carbon bond homolysis

Volume 24, issue 1 (2021), p. 147-175.

Published 26th April 2021

This article is licensed under the Creative Commons Attribution 4.0 International License. http://creativecommons.org/licenses/by/4.0/

Les Comptes Rendus. Chimie sont membres du Centre Mersenne pour l’édition scientiﬁque ouverte www.centre-mersenne.org Comptes Rendus Chimie 2021, 24, n 1, p. 147-175 https://doi.org/10.5802/crchim.73

Essay / Essai

A journey into metal–carbon bond homolysis

Rinaldo Poli a

a Laboratoire de Chimie de Coordination, 205 Route de Narbonne, 31077 Toulouse, France E-mail: [email protected]

Abstract. This article surveys the current knowledge in metal alkyl complexes with homolytically weak metal–carbon bonds, therefore prone to thermally produce alkyl radicals. It outlines the role of a metal complex as a moderator to control the radical reactivity (“persistent radical effect”). It describes the methods that have been used so far (as well as others that are potentially available) to investigate the metal–carbon bond cleavage thermodynamic and kinetic parameters, including their caveats and lim- itations. A number of systems scrutinized in the author’s own laboratory and in those of collaborators are presented and discussed. These investigations have combined metal complexes and alkyl radicals, with guidance and understanding provided by DFT calculations, to achieve higher performance in the controlled radical polymerization of challenging monomers (vinyl acetate, vinylidene fluoride) and in olefin radical cross-coupling, and have brought to light mechanistic questions of more general relevance. Keywords. Metal–carbon bond, Homolytic cleavage, Bond dissociation enthalpy, Persistent radical effect, Metal-mediated radical reactivity.

Manuscript received 2nd January 2021, revised 11th February 2021, accepted 15th February 2021.

1. Introduction An additional process leading to metal alkyl decomposition is bond homolysis, which is a one- Organometallic chemistry textbooks teach us that electron process generating an organic radical. As or- metal alkyl compounds are very reactive species, ganic radicals are themselves very reactive species often decomposing upon exposure to the atmo- and ultimately disappear by irreversible coupling sphere (oxygen, moisture) and by a few spontaneous and disproportionation processes, metal alkyl com- processes (β-H elimination, reductive elimination, plexes can be isolated or used as catalytic interme- etc.). Consequently, they can be maintained as sta- diates only if the metal–carbon bond has suﬃcient ble species, isolated and characterized only with spe- homolytic strength. Nevertheless, it has become in- cial precautions (inert atmosphere, choice of R group creasingly apparent that compounds with a weak and ancillary ligand, etc.). At the same time, this metal–carbon bond, hence prone to generate tran- high reactivity and the multitude of available reac- sient radical species, are extremely useful. Like for the tion pathways make them extremely useful and ver- above-mentioned two-electron reactivity, this one- satile in catalysis, where they are involved in elemen- electron reactivity can be harnessed to promote use- tary steps that generally feature 2-electron changes in ful and unique transformations. The importance of the metal coordination sphere (migratory insertion, metal–carbon bond homolysis has probably been bond metathesis, etc.). ﬁrst highlighted in biochemistry, where metal centers

ISSN (electronic) : 1878-1543 https://comptes-rendus.academie-sciences.fr/chimie/ 148 Rinaldo Poli

in enzymes and cofactors play a crucial role in the physiological regulation of radical reactivity, but has also subsequently been recognized in the areas of polymer chemistry and organic chemistry. In this article, I wish to highlight a few important principles and considerations and provide a few examples from our own work, which is mostly focused on metal- mediated radical polymerization. In this respect, although the standard abbreviation of a generic metal by organometallic chemists is M, I elected to use the alternative abbreviation Mt, to avoid confusion with the same abbreviation used within the polymer community for a generic monomer. A generic formal oxidation state will be indicated by the symbol x (e.g. x x 1 Mt , Mt + ). Again, in order to avoid confusion, one- Figure 1. Energetic profile of a metal–carbon electron ligands (in most cases a halogen) will be bond cleavage. identified by the symbol Y rather than by the more commonly used X. The symbol T will be used when a one-electron species, generally stable in the free 1 process is very small (estimated as 2 kcal mol− )[1– form, acts as a radical trap. A generic coordination < · x x 4]. Therefore, the experimentally more readily acces- sphere will be identified as L/ (e.g. L/Mt or Mt /L) ‡ sible activation enthalpy (∆Ha ) gives a close estimate and a generic radical as R. The considerations that I of the BDE, although this conclusion may be danger- develop in this article are also valid in principle for ous for systems where Mtx/L requires dissociation of metal-aryl bonds, though the vast majority of cur- solvent molecules, e.g. water, see further discussion rent applications involves alkyls. The Mt–R bond ho- in Section 4.3. molysis can be activated either thermally or photo- It is important to appreciate the relationship be- chemically. A redox stimulus may also promote bond tween these parameters and the “stability” or life- homolysis, but only via an alteration of the Mt–R time of the compound. A useful reference point is bond strength by the oxidation state change, while provided by [(CO)5Mn–CF3], which may be consid- the bond cleavage process itself remains either ther- ered as a typical “stable” organometallic compound. mally or photochemically activated. In this article, I Extensive heating of this compound in an inert sol- will only focus on the thermal activation method. vent at 100 °C or above and in the absence of a radical trapping species does not significantly de- 2. Energy profile and thermal stability compose it. However, as shown from recent work in my laboratory, the bond can be broken at mea- The key thermodynamic and kinetic parameters of surable rates in the presence of a radical trap (see the homolytic bond cleavage, which may be ex- the details in Section 4.6) [5]. The measured Mn– ‡ pressed on the enthalpy and/or free energy scales, C bond activation parameters are ∆Ha 53.8 1 ‡ = 1 1± are shown in Figure1. The homolytic cleavage pro- 3.5 kcal mol− and ∆Sa 66.0 9.5 cal mol− K− . · ‡ = ± · · cess generates two species from a single one, thus en- Assuming ∆Ga ∆G, the equilibrium constant can ≈ 18 14 tails a very positive entropy term (∆S), which is al- be estimated as 3.3 10− at 100 °C or 4.4 10− × × ready largely expressed at the transition state level at 150 °C, yielding initial radical concentrations of ‡ 9 7 (∆Sa ∆S; however, see further discussion in Sec- 2.9 10− and 2.1 10− M, respectively, from a ≈ ‡ × × tion 4.2 below). Consequently, ∆G ∆H and ∆Ga standard 1 M solution of the organometallic precur- ‡ ‡ ‡ ¿ ¿ 1/2 ∆H , whereas ∆G ∆H . In the absence of steric sor ([CF•] [(CO) Mn•] [K /(1M)] ). These con- a da ≈ da 3 = 5 = impediments, geometrical reorganization, the need centrations, though very small, are still suitable to of ligand dissociation, or a spin state change, the promote radical reactivity. For instance, the radical x ‡ ‡ ‡ L/Mt -radical coupling proceeds at diffusion-limited flux at 100 °C (where ∆Ga ∆Ha (373) ∆Sa 1 = − · ≈ rates, namely the barrier to the radical deactivation 30 kcal mol− ) from a 1 M solution of [Mn(CF )(CO) ] · 3 5

C. R. Chimie — 2021, 24, n 1, 147-175 Rinaldo Poli 149

Scheme 1. Generation and reactivity of radicals.

can be calculated from the Eyring equation as 2.3 3. The “persistent radical effect” 5 1 4 × 10− mol s− (t 3.2 10 s or 9 h), which is not · 1/2 = × unreasonably long for a chemical reaction. However, Radical reactions may be triggered by primary radi- heating [(CO)5Mn–CF3] alone in an inert solvent at cals (R0•) generated from a stable initiator (I) through 100 °C for 9 h will not lead to 50% decomposition, a suitable stimulus (thermal, irradiation, redox, etc.), not even to a small fraction of it. The reason is the Scheme1. These radicals then engage in a variety of greater efficiency of the back-trapping process from useful transformations, such as additions to unsat- the {[(CO)5Mn•],CF3•} caged fragment couple rela- urations, atom/group transfer and fragmentations, tive to cage escape and subsequent irreversible pro- generating new radicals (R0–S•,Z•,R•, collectively in- cesses leading to decomposition (coupling of two dicated as R•), which may undergo the same trans- [(CO)5Mn•] to yield [Mn2(CO)10] and of two CF3• to formations again (chain mechanism) and at some yield C2F6, plus other possible processes involving point eliminate a stable product (P) by a fragmenta- the solvent). In addition, the back-trapping process tion process. This useful reactivity always competes is kinetically first-order and the radical coupling pro- with the irreversible bimolecular radical termina- cesses are second order, hence disfavored at low con- tions by combination and disproportionation. How- centrations. Thus, even compounds with lower BDEs, ever, the presence of a reversible trapping species therefore giving greater radical fluxes and promoting (T), able to produce a stable but reactivatable dor- radical reactivity at higher rates, may be “thermally mant species T–R, reduces the impact of the termi- stable” in a practical sense, if kept away from radi- nations and improves the performance of the use- cal trapping species. The low BDE limit that allows ful radical reactivity. This is known as the “persis- isolation and manipulation of a metal alkyl com- tent radical effect” and the reversible formation of a plex under reasonably practical conditions appears metal–carbon bond (i.e. if T [L/Mtx]) is only one 1 = to be around 20 kcal mol− . Examples are certain of many ways in which this effect can be imple- · fragile alkylcobalamins and other related L/CoIII–R mented. The origin of this effect is kinetic: the use- compounds [6,7]. The point that I wish to stress in ful processes are all first-order in radical, whereas this article, however, is that even when the Mt–C both types of terminations (combination and dispro- bond strength is too small to allow isolation, the portionation may be combined into a single rate law organometallic species may still play a role as a with k k k ) are second-order. The activa- t = t,c + t,d transient in a stoichiometric or catalyzed reaction, tion/deactivation equilibrium lowers the active rad- improving the radical reaction selectivity, through ical concentration under steady-state reaction con- the so-called “persistent radical effect”. ditions. Hence, whereas the rate of the sought trans-

C. R. Chimie — 2021, 24, n 1, 147-175 150 Rinaldo Poli

formations scales linearly with the radical concentration, the terminations rate scales quadratically. Take an ideal but realistic example where, in the absence of reversible radical trap, the steady-state rad- 7 ical concentration is 10− M, the rate constants of the terminations and of the useful transformation 8 2 1 1 are 10 and 10 M− s− , respectively, and the sub- · strate concentration is 1 M. The useful chemistry 2 1 1 7 proceeds at a rate of (10 M− s− )(10− M)(1 M) 5 1 · = 10− M s− whereas the radicals disappear at a rate 8· 1 1 7 2 6 1 of (10 M− s− )(10− M) 10− M s− , namely · = · 10% of the generated radicals are lost before they can undergo any useful transformations. However, in the presence of a moderating equilibrium with, for 9 x 1 the sake of argument, K 10− and [L/Mt + –R] = = [L/Mtx], the steady-state radical concentration be- 9 7 1 comes 10− M, yielding rates of 10− M s− for the Scheme 2. Evolution of the radical and T con- 10 1 · useful transformation and 10− M s− for the termi- · centrations following terminations [9,15–17]. nations (only 0.1% of the radical are lost). The useful process has become 100 times slower, but the terminations have become 10,000 times slower. that the same effect can be provided by certain un- I wish to comment on the use of the “persis- stable species that slowly decompose themselves bi- tent radical effect” (PRE) terminology. This name was molecularly [13,14]. Therefore, the reversible radical coined after the initial investigations of radical reac- trap (T) that ensures the PRE does not need to be a tions where both R• and T are radicals, one of which radical and does not need to be persistent! It should (T) is persistent. This term was introduced by Daikh rather be identified as a moderating species/agent, and Finke (organometallic chemists) in a 1992 inves- since its role is just to moderate the steady-state rad- tigation of a radical isomerization process occurring ical concentration. I reluctantly continue to use the in a coenzyme B12 model complex, where the self- PRE terminology, though within quotes, but suggest coupling of benzyl radicals is suppressed by the pres- to rather use more appropriate and general termi- ence of the L/CoII “persistent radical” [8]. The PRE nologies such as “moderating effect” or “reversible terminology was then popularized by Fischer (who trapping effect”. actually inspired the study of Daikh and Finke) with In order to fully benefit from this moderating ef- specific application to controlled radical polymeriza- fect, if primary radicals are injected into solution tion [9,10], and has since been extensively used by the from a conventional initiator (as in Scheme1), the polymer chemists working in this area. The PRE ter- molar amount of the trapping species (T) must be minology has also creeped into the jargon of the or- at least as high as the total amount of the radicals ganic chemistry community working on radical reac- produced by the initiator. However, the reaction may tions [11,12]. There are two different ways to see the also be initiated by a labile T–R compound (e.g. an x 1 meaning of the PRE terminology. On one side, it can organometallic compound L/Mt + –R0 with a frag- be interpreted as a trick that allows to extend the life ile bond), providing itself the primary radical and of the R• radicals, hence rendering them more per- the moderating species T, in which case an indepen- sistent. This is a fine interpretation. It is, however, dent radical initiator (I) is not required. In this case, also prone to be interpreted (as it has been) as the the moderating equilibrium evolves in a predictable effect provided by the “persistent radical” T. This is way [9,15–17] as a consequence of the inevitable rad- not a correct interpretation! As we now know, the ical terminations (Scheme2). If T is stable, the R • dis- same effect can be provided by metal complexes with appearance entails the conversion of the equivalent any spin state, including diamagnetic ones. In ad- amount of T–R into T and the [T] monitoring pro- dition, work from my laboratory has demonstrated vides a convenient way to assess the fraction of lost

C. R. Chimie — 2021, 24, n 1, 147-175 Rinaldo Poli 151

radicals. Consequently, the T–R/T ratio decreases, in- BDEs. Another indirect method has consisted in the x 1 x ducing a [R•] decrease through the equilibrium ex- [L/Mt + –R] and [L/Mt ] measurement in the pres- pression. This evolution further increases the posi- ence of a steady-state concentration of continuously tive effect of the moderating equilibrium. generated and terminating R•, combined with the indirect [R•] knowledge from the steady-state condi- 1/2 tions (V V ), namely [R•] (k [IN]/2k ) ([IN] 4. Assessment of the thermodynamic and acti- i = t = d t = initiator concentration). For instance, the steady- vation parameters state concentration of styryl radicals, obtained from the thermal decomposition of AIBN in the presence 4.1. Thermodynamic parameters of excess styrene, has allowed the K determina- III The thermodynamic Bond Dissociation Enthalpy tion for the homolysis of [(TAP)Co –CH(CH3Ph)] (TAP tetra(p-anisyl)porphyrin) [23,24]. (BDE) may be obtained from constant pressure = calorimetric measurements (combustion calorime- The equilibrium constant (hence the BDFE) try, photoacoustic calorimetry, etc.), which generally can also be derived from the thermal decomposi- require the application of thermochemical cycles tion kinetics according to Scheme2(T L/Mtx) in = and a number of approximations and assumptions the absence of radical traps, monitoring either the x 1 x and results in the accumulation of experimental er- L/Mt + –R disappearance or the equivalent L/Mt rors [18,19]. Many BDEs determined by these meth- accumulation. Equations describing the time evo- ods have subsequently been reassessed (one exam- lution of [T] have been derived (Scheme3). Under ple will be provided below in Section 4.6) and this conditions where [T] [T–R]0, namely up to small ¿ line of research appears abandoned. Published BDE conversions (e.g. 10%), the T concentration grows ≈ values obtained by these methods should be taken linearly with t 1/3 and the equilibrium constant K can with extreme care. be extracted from the slope of the ([T]– T]0) versus | The equilibrium constant K gives access to the t 1/3 linear best fit, provided the radical termination Bond Dissociation Free Energy (BDFE) and its tem- rate constant kt is known [9,15,16]. For decompo- perature dependence yields the BDE, as well as the sitions proceeding to greater conversions, K can be entropy change, through van’t Hoff’s relationship. extracted from the slope of the F versus t linear best Direct equilibrium measurements, however, are im- fit [17]. Extraction of the BDE requires this analysis to possible because of the radical species instability. be repeated at different temperatures and knowledge Hence, indirect methods are necessary. In addition, of the kt temperature dependence. This method has the solvent must be innocent (i.e. not chemically been used to determine the equilibrium constants interact with L/Mtx and/or T), or the related interac- of other fragile bond homolyses such as the O–C tion energies must be known in order to correct the bond in alkoxylamines [17] and of atom transfer determined enthalpy value. Certain organometallic equilibria [25], but not yet of metal–carbon bonds compounds lead to measurable equilibria involving homolyses to the best of our knowledge. However, the metal–carbon bond homolysis to yield stable the method is simple and of potentially wide ap- end-products, which allows to calculate the bond plicability, provided T (i.e. L/Mtx) is stable. The T homolysis BDFE through thermochemical cycles. accumulation continuously slows down the decom- III For instance, various [(L)(dmgH)2Co –CH(CH3)Ph] position as discussed in the previous section, thus compounds (L pyridine, 4-substituted pyridine, relatively fragile bonds may be conveniently investi- = II x 1 imidazole) decompose to yield [(L)(dmgH)2Co ], gated. If L/Mt + –R is not sufficiently stable for isola- styrene and H2 selectively (no radical coupling) via tion, the method can still be applied if the compound II β-H atom transfer from Ph(CH3)CH• to the Co cen- can be generated in situ at a known concentration ter followed by bimolecular decomposition of the in a shorter timescale than that of its spontaneous III resulting [(L)(dmgH)2Co –H] intermediate [20–22]. decomposition. Combination of ∆H ◦ from the measured equilib- In a few special cases, K has also been derived by rium in the 10–37 °C temperature range with the combining the measured ka and kda values. This has heats of formation of styrene and of the Ph(CH3)CH• been applied to systems where R• CH3•, generated III =x radical has provided the sought Co –CH(CH3)Ph by pulse radiolysis, reacts with L/Mt to form unsta-

C. R. Chimie — 2021, 24, n 1, 147-175 152 Rinaldo Poli

Scheme 3. Time evolution of the T concentration during the T–R decomposition according to Scheme2[17].

Scheme 4. Kinetics approach for the measurement of the activation rate constant.

x 1 ble L/Mt + –CH3 transients, which subsequently de- (TEMPO) [6,22,29–33], H-atom donors (thiols [34], compose by rate-limiting homolysis [26,27]. silanes [5,35], stannanes [36], metal hydrides [37,38]), O2 [7,39,40] or other oxidizing agents (H2O2, aqueous 4.2. Bond cleavage barrier metal ions and complexes) [41–44] and also other L/Mtx complexes that form a stronger Mt–R bond The activation rate constant (ka) is accessible by measuring the rate of the organometallic precur- than the bond being broken [45,46]. For certain sys- sor disappearance under conditions in which the tems containing β-H atoms on R, the decomposition back recombination (radical deactivation) is re- could also be kinetically monitored in the absence moved from the kinetics scheme [28]. This approach of a trapping agent (see below) [20–22,34,47,48]. is easier to implement and has been much more This kinetic method can be applied to compounds widely used than the BDE or BDFE determination. that are suﬃciently stable to be isolated in a pure The key is to add a suitable trapping agent, able to form or that can be generated at known concentration in a shorter timescale than that of the trap- irreversibly capture one or both of the produced ‡ ping experiment. The lowest reported ∆Ha values fragments competitively with respect to the back 1 appear to be around 17 kcal mol− [22,36]. In a cou- reaction (i.e. saturation conditions), Scheme4. · ple of cases, the activation parameters could be ex- The trapping agent must also be inert with re- tracted from a temperature-dependent 1H NMR line spect to the organometallic precursor. The tem- ‡ broadening study, with no need of trapping agents perature dependence of ka then provides ∆Ha and ‡ or concentration-time measurements [23,49]. This ∆Sa from Eyring’s equation. Various radical traps have been used for this purpose, the choice being method may be more widely applicable to thermally dictated by the nature of the organometallic sys- “stable” compounds. Most investigations have dealt tem and the need to exclude side reactions such with complexes of 3d metals, which feature homolyt- as oxidation or ligand addition or exchange. Ex- ically weaker bonds than their heavier congeners. amples are (2,2,6,6-tetramethylpiperidin-1-yl)oxyl The ka measurement is complicated by the pres-

C. R. Chimie — 2021, 24, n 1, 147-175 Rinaldo Poli 153

Scheme 5. Kinetics scheme for the bond homolysis of a few L/CoIII–R compounds in the presence of β-H atom transfer.

ence of a β-H atom on the alkyl chain. In that case, Me Ph). However, steric bulk induces inversions be- < x 1 the decomposition may follow a trap-independent cause the crowded L/Mt + exerts a stronger steric pathway that involves β-H atom transfer from the pressure than the simple H atom. The facile ac- alkyl group to the metal center within the caged frag- tivation of the primary neopentyl group is a no- x 1 ment pair to generate the hydride complex L/Mt + – table example [36,46]. The investigation of series of H and an alkene, as pointed out for a few CoIII com- compounds with the same R, on the other hand, plexes (Scheme5)[20–22,34]. The fleeting Co III hy- illustrates the importance of ligand effects. More x 1 dride intermediate rapidly releases H2. In princi- electron-donating ligands stabilize L/Mt + –R better x ‡ ple, both the trap-free first-order (k0) and the trap- than L/Mt and thus tend to increase ∆Ha [6,21,22]. dependent saturation (k1/k 1,k2) pathways are initi- However, the bond strength has a more pronounced − ated by CoIII–C bond homolysis, but the need to en- and inverse dependence on the ligand cone angle, counter the trapping species in the latter affects the generally masking the basicity effect [6,36,47]. Lig- transition state structure, the activation enthalpy (in and addition to convert a 5- to a 6-coordinate system ‡ ‡ x 1 all cases, ∆H ∆H ), and particularly the activa- labilizes the Mt + –R bond (e.g. for the B system 0 < 1 12 tion entropy (∆S‡ ∆S‡). The k value could be mea- on going from the base-off to the base-on form), in 0 < 1 0 sured in the absence of trapping agent and the k1 line with the predominance of steric over electronic 1 value from the intercept of a plot of (kobs k0)− ver- effects. Linear free-energy relationships have been 1 − sus [Trap]− . For those derivatives where R does not proposed to understand the relative contributions of contain any β-H atom, k0 is zero and the decomposi- steric, bond polarity, and product stabilization fac- tion only follows the quenched pathway. tors to the homolytic strengths of various bonds [50– ‡ 52], though this method has not been applied so far, The reported activation enthalpies (∆Ha ) span x 1 1 to the best of my knowledge, to L/Mt + –R bonds. a rather wide range (17–54 kcal mol− ), being lim- · ited at the lower end by the ability to generate the Like the activation enthalpies, the reported ac- x 1 ‡ organometallic precursors L/Mt + –R and at the up- tivation entropies (∆Sa), surprisingly, also span 1 1 per end by the experimental conditions needed for a very wide range, up to 66 cal mol− K− for · · sufficiently high decomposition rates. For a series of [(CO)5Mn–CF3][5] and down to negative values, x 1 ‡ compounds with the same L/Mt + , the ∆Ha (and which is not consistent with a bond breaking pro- thus presumably the BDEs) qualitatively scales with cess. Halpern has commented on this unexpected the homolytic strength of the corresponding R–H behavior and highlighted a qualitative relationship bond (e.g. benzyl secondary alkyl primary alkyl between ∆S‡ and ∆H ‡, which was termed “compen- < < < a a

C. R. Chimie — 2021, 24, n 1, 147-175 154 Rinaldo Poli

ganization at the transition state level and can ac- count for negative activation entropies. This situation is closely related to the above-mentioned trap- free pathway when R contains β-H atoms, where the caged pair evolves by H-atom transfer within Scheme 6. New kinetics scheme for the the cage and indeed the resulting ∆Sa value is lower x 1 L/Mt + –R bond homolysis with consideration than for the trap-dependent pathway. Certain trap- of a trap-inclusive cage effect. ping species have a radical character (e.g. nitroxides, CoII complexes) and thus may favorably interact with the dissociative Mt R transition state be- ··· sation effect” [28], but a persuasive rationalization cause the C atom has developed a large spin density of this phenomenon could not be offered. I would at that level, as suggested by DFT calculations [54,55]. like to propose a possible way to interpret this be- This idea is also supported by investigations of en- x havior. The bond homolysis produces the {L/Mt ,R•} zymatic reactions catalyzed by adenosyl cobalamin fragment pair in a solvent cage before radical trap- (AdoCbl), where kinetic coupling between the cobalt- ping by T. All published activation parameters are carbon homolysis step and the subsequent radical- based on the kinetic analysis of Scheme4, in which substrate reaction was demonstrated by the pres- no consideration is given to the caged intermedi- ence of a deuterium kinetic isotope effect [56–58]. ate. One contribution pointed out that the caged If the trapping species does not intervene until af- pair needs to be separated (k2/k 2 equilibrium ter the radical cage escape, or at least until after − in Scheme6), but that analysis also assumed that the rate-determining transition state, a very positive the trapping agent does not intervene until after activation entropy is anticipated, as is indeed ob- radical escape from the cage (k3)[53]. That re- served for many bond homolyses. Note that the pres- vised kinetic scheme proposed a modulation of ence of a Mt T interaction at the transition state ··· the bond homolysis activation parameters by the level may also affect the measured activation en- parameters of the caged pair recombination and thalpy. In order to evaluate this effect, measurements cage escape: ∆H ‡ ∆H ‡ F [∆H ‡ ∆H ‡ ]; obs = 1 + c · 2 − 1 of the bond homolysis activation parameters for the ‡ ‡ ‡ ‡ − x 1 ∆Sobs ∆S1 Fc [∆S2 ∆S 1], where Fc (frac- same L/Mt + –R compound with different traps are = + · − − tional cage efficiency) k 1/(k 1 k2). When Fc 0 necessary. This test was indeed done for the bond − − = + = III 2 (fast caged pair separation relative to recombi- cleavage kinetics of several [(H2O)5Cr –R] + [41– nation) ∆H ‡ ∆H ‡ and ∆S‡ ∆S‡, whereas obs 1 obs 1 43] and [(H2O)(dmgBF2)2Co–CH2Ph] [44], yielding = = ‡ ‡ when F approaches 1 (and thus ∆H ∆H ), essentially indistinguishable activation parameters, c 2 À 1 ∆H ‡ ∆H ∆H ‡ and ∆S‡ ∆S ∆S‡. However,− but additional studies for other L/Mtx 1–R bonds obs = 1 + 2 obs = 1 + 2 + ‡ seem warranted. This “trap-inclusive cage effect”, even this scheme cannot rationalize negative ∆Sobs values.It seems to me quite plausible that the trapping however, does not predict any specific relationship ‡ ‡ agent may also be part of the cage walls, hence can between ∆Sa and ∆Ha . x 1 interact directly with the caged pair, introducing a Although most investigated L/Mt + –R bond ho- fourth step in the kinetic scheme (k4 in Scheme6). molyses are indeed consistent with an SH1 mech- The saturation kinetics analysis yields the 1/k1 value anism, an SH2 process (Scheme7) has been doc- by extrapolation of 1/kobs to 1/[T] 0 (i.e. to very umented for the methyl radical transfer in vari- = III II large [T]) rendering the assumption of no trap in the ous L/Co –CH3–L0/Co systems (L0 or L) with ‡ 1 = 6= ‡ cage rather absurd. This does not amount to say- ∆H as low as 7.5 kcal mol− and very negative ∆S 1 1· ing that the process is associative (such as an SH2 (ca. 20 cal mol− K− )[59]. The alkyl and benzyl − · · III process), for which the rate law would show a first- radical transfer from [(dmgH)2Co –R] to another or- II order dependence on [T] and the extrapolation of ganic radical R0•, to yield RR0 and [(dmgH)2Co ], also 1/k to 1/[T] 0 yields zero (no saturation). The adopts an S 2 mechanism [60], as are other metal– obs = H mechanism remains of SH1 type, but the transition carbon bond homolyses [61]. Thus, the rate law and state “feels” the presence of the trap in the cage particularly the presence of the saturation regime (Mt R T), which is expressed in the molecular or- should always be carefully checked. ··· ···

C. R. Chimie — 2021, 24, n 1, 147-175 Rinaldo Poli 155

4.3. Bond formation barrier

Direct measurement of the deactivation rate constant (kda) is difficult, as it requires either monitoring the disappearance of L/Mtx in the presence of a known concentration of radicals (which is possible Scheme 7. SH1 versus SH2 processes involving x 1 only in flash photolysis or pulse radiolysis experi- the Mt + –R bond homolysis. ments), or the measurement of the kda/ktrap com- petition and the separate knowledge of ktrap. These studies have generally been carried out only at one of events triggered and followed electrochemically, temperature to yield estimated kda values without x 1 where the L/Mt + –R compound is formed by trap- experimental error [3,43,62–69]. Most reported val- 7 8 1 1 ping in situ-generated radicals with in situ-generated ues are 10 and often 10 L mol− s− , i.e. close x > > · · L/Mt . The sequence of events and a typical elec- to the diffusion limit. The addition of the CH OH • 2 trochemical response are shown in Figure2[73–75]. radical, produced by radiolysis of N2O-saturated x 1 One-electron reduction of a stable L/Mt + –Y pre- aqueous solutions containing CH3OH, at low pH to x 1 II II cursor, where Mt + Cu and Y Cl or Br, at the complex [Co (nta)(H2O)2]− (nta nitrilotriacetate, = =x = EY potential (wave A) yields L/Mt and Y− at the N(CH COO ) ) is a rare example of a temperature- 2 − 3 electrode surface and is reversible (blue curve) in dependent investigation. The measured addition 8 1 1 the absence of R–Y substrate. The EY potential is af- rates in the 7–55 °C range are (0.97–4.1) 10 M− s− , ‡ 1× ·‡ fected by the rapid Y− dissociation equilibrium (KY, yielding ∆H 4.8 0.5 kcal mol− and ∆S = 1 1± · =‡ EC process) and is thus [Y−]-dependent. Addition 4.6 2 cal mol− K− [26]. The relatively high ∆H − ± · · of an alkyl halide substrate (R–Y) alters the voltam- value might be associated to the need to displace a metric response (red curve) as a consequence of the water molecule from the coordination sphere. For following events: the reduced L/Mtx complex acti- solvated cations, particularly in water, dissociation of vates R–Y by atom transfer (AT, rate constant ka,AT) the coordinated solvent seems indeed an important x 1 to generate R• and L/Mt + –Y, decreasing the inten- step during the Mtx 1–R bond formation process, + sity of the reverse wave B. This R–Y activation is re- as suggested by the activation volumes measured versible (KAT equilibrium), as exploited in the popu- by pressure-dependent kinetic studies [70,71]. The lar “atom transfer radical polymerization” (ATRP) [76, kda can also be estimated from the independent 77]. At this point, the produced R• can either (i) be knowledge of ka and K . In many contributions dis- x 1 x 1 trapped by L/Mt + –Y (deactivation by atom transfer, cussing the homolytic strength of a L/Mt + –R bond, ‡ rate constant kda,AT); (ii) be trapped (if desired) by an the BDEs was estimated from the ∆Ha data assum- ‡ added external trapping agent T; (iii) be trapped by ing a diffusion-limited recombination rate (∆Hda x x 1 L/Mt (deactivation to generate by L/Mt + –R, rate equal to the ∆H ‡ of the viscous flow, which is ca. 1 constant kda,OM); or (iv) spontaneously terminate by 2 kcal mol− for typical low-viscosity solvents) [72]. x 1 · coupling and/or disproportionation. The L/Mt + –R Indeed, ∆H ‡ measurements in higher-viscosity sol- a bond formation is also reversible (K equilibrium). vents such as ethylene glycol have yielded higher val- OM This bond formation occurs only if the bond is suf- ues [33]. The most precisely determined k values da ficiently strong (sufficiently small K ) and is evi- appear to be those obtained from electrochemical OM denced by the appearance of waves C and D at the E data (next section). R potential, which is more negative than EY because of the greater donating power of R− relative to Y−. The 4.4. Electrochemical simulations reversibility of this second electrochemical process depends on the follow-up events of the electrochem- x The simultaneous experimental determination of ically generated [L/Mt –R]− (e.g. equilibrated release x 1 the L/Mt + –R bond homolysis thermodynamic and of the reactive carbanion R−, KR). The simulation of kinetic parameters has been possible for certain the observed voltammogram depends on many inde- copper(II) systems from the simulation of cyclic pendent parameters (EY, ER, KY, ka,AT, kda,AT, ka,OM, voltammograms. The method is based on a cascade kda,OM, KR, [R–Y], scan rate), but the thorough explo-

C. R. Chimie — 2021, 24, n 1, 147-175 156 Rinaldo Poli

x 1 x 1 Figure 2. Left: scheme of the cascade processes leading to the generation of L/Mt + –R from L/Mt + –Y and R–Y. Right: cyclic voltammogram of one speciﬁc example (the lines are background-subtracted experimental data and the points correspond to the simulation) in the absence (blue) and presence (red) of R–Y 1 x 1 3 at a scan rate of 0.5 V s− in CH CN. The example used L/Mt + –Y [CuBr(TPMA)]+(1.0 10− M) made · 3 = × in situ from CuBr and tris(pyridylmethyl)amine (TPMA L) and R–Y bromoacetonitrile (Br–CH CN, 2 = = 2 BAN). The right image is reproduced with permission from Ref. [73]. Copyright 2019 American Chemical Society.

9 1 1 ration of the reaction space (variations of the scan in MeCN to (1.4 0.2) 10 M− s− for II ± × · rate and [R–Y]) and the independent determination [(TPMA)Cu –CH(CH3)CN]+ in DMF. This method, of certain parameters under simplified conditions only applied so far to the investigation of organocop- (for instance, using TEMPO as an irreversible radical per(II) species [73–75], can potentially be extended to x 1 x 1 trap under saturation conditions) has allowed an im- other L/Mt + –R systems with labile Mt + –R bonds. provement of the fit. The OM equilibrium parameter can be accurately In the more accurate investigation [73], the estimated only if ka,OM is sufficiently high to have solvent-dependent KOM was found to range from an impact on the CV shape within the timescale of 10 II x 1 (4.2 1.4) 10− for [(TPMA)Cu –CH2CN]+ (a the measurement, but at the same time the Mt + –R ± × 6 primary alkyl radical) in DMF to (2.7 1.4) 10− bond must be strong enough to allow the generation II ± × for [(TPMA)Cu –CH(CH3)COOCH3]+ (a secondary of observable amounts of the organometallic species alkyl radical) in MeCN, whereas tertiary radicals in situ. These conditions are associated to systems (e.g. •CMe2COOR) are not efficiently trapped for the that are generally not amenable to isolation as pure investigated systems. The corresponding BDFE range compounds, at least under standard laboratory 1 is 8.2–13.6 kcal mol− . The activation rate constant conditions. · 2 1 (ka,OM) spans a wide range from (4.6 0.8) 10− s− II ± × for [(TPMA)Cu –CH2CN]+ in MeCN to (6.4 4.5. Computational studies 2 1 II ± 2.4) 10 s− for [(TPMA)Cu –CH(CH )COOCH ]+ × 3 3 in DMSO and the deactivation rate constant In addition to all the above-mentioned experimen- 7 1 1 x 1 (kda,OM) is always 10 M− s− , ranging from tal methods of investigation, the L/Mt + –R bond ho- 7 1 > 1 · II (4.1 0.6) 10 M− s− for [(TPMA)Cu -CH CN]+ molysis, with particular focus on the thermodynamic ± × · 2

C. R. Chimie — 2021, 24, n 1, 147-175 Rinaldo Poli 157

investigations have proven invaluable to rationalize observed (sometimes unexpected) phenomena, providing useful insight and understanding, and to predict trends when exploring a series of closely related systems, assisting the design and experimental development of new systems capable of achieving a desired performance. A few examples of the contribution of DFT calculations for work carried out in my laboratory and in those of my collaborators will be presented in the remainder of this article.

4.6. Reassessing bond strengths: comparison of calorimetric, kinetic and computational data

As stated in Section 4.1, old thermochemical data Figure 3. Reported experimental and com- have occasionally been shown erroneous. A re- putational bond dissociation/activation en- I cent reassessment based on work carried out in thalpies for compounds [(CO)5Mn –RF] (RF = my laboratory has concerned the Mn–C BDE in CF3, CHF2). I I [(CO)5Mn –CF3] and [(CO)5Mn –CHF2]. Figure3 summarizes the evolution of the experimental and BDE parameter, has been investigated by theoreti- computational efforts to assess the bond strength I cal calculations, in most cases using a density func- of these two compounds. For [(CO)5Mn –CF3], the tional theory (DFT) approach. The computational BDE was first evaluated by Connor et al. in 1982 by error cannot be assessed in DFT methods, mak- microcalorimetric determinations of the enthalpies ing this approach of limited and doubtful value for of sublimation, thermal decomposition, bromi- the quantitative estimation of BDEs. Indeed, values nation and iodination of the compound, yielding spreading over a very large range have been ob- its enthalpy of formation through thermochemi- tained for the same compound depending on the cal cycles. This value could then be combined with selected functional. For instance, the intensively in- the already assessed Mn–Mn BDE in [Mn2(CO)10] 1 vestigated coenzyme B12 has afforded BDEs from (94 kJ mol− ), leading to the estimation, on the 1 · 29.5 kcal mol− (close to the experimentally accepted basis of additional assumptions, of the Mn–CF3 · 1 1 1 value) to 15 kcal mol− or less [78–83]. Pure func- BDE as 172 7 kJ mol− (41.1 1.7 kcal mol− )[87]. · ± · ± · tionals appear to afford better results than hybrid A re-evaluation in a 1990 review article by Mar- functionals [80,84,85] and it is crucial to appropri- tinho Simões and Beauchamp [19], based on a ately consider dispersion forces [80]. Complexes than new and perceived more precise Mn–Mn BDE in 1 can adopt two or more different spin configura- [Mn2(CO)10] (159 21 kJ mol− ), placed this BDE at 1 ± · 1 tions introduce additional difficulties because differ- 203 6 kJ mol− (48.5 1.4 kcal mol− ). However, an ± · ± · ent functionals introduce different relative stabiliza- independent photoionization mass spectrometric tion effects for spin isomers [86]. Thus, confidence in study cited in the same review article [19] and pub- the relevance of any computed value must be gained lished only in a 1981 Ph.D. thesis [88], which also from the extensive benchmarking of the computa- used thermochemical cycles and assumptions, gave 1 1 tional method against any available (and reliable) ex- a BDE of 182 11 kJ mol− (43.5 2.6 kcal mol− ). ± · ± · perimental value. The most reliable information con- The same method also provided a BDE of ‡ 1 1 cerns the enthalpic (BDE, ∆Ha ) parameters, not the 144 11 kJ mol− (34.4 2.6 kcal mol− ) for the Mn–C ± · I ± · free energies. This is related to the difficult transposi- bond in [(CO)5Mn –CHF2][19,88]. In 1993, how- tion of the computed gas phase entropic correction ever, Folga and Ziegler applied the DFT approach to the condensed phase. In spite of the uncertain- for the computation of BDEs in a number of Mt–H I I ties related to the absolute computational error, DFT and Mt–C bonds in [(CO)5Mn –R] and [(CO)4Co –R]

C. R. Chimie — 2021, 24, n 1, 147-175 158 Rinaldo Poli

compounds, including those with R CF and CHF , several R -H benchmarks and quite different than = 3 2 F using a local density approximation (LDA) with the that previously used by Folga and Ziegler. This func- optional addition of non-local exchange and corre- tional yielded BDEs for the [(CO)5Mn–RF] bonds 1 lation corrections (LDA/NL) [89]. The BDE values (55.1, CF ; 48.0, CHF ; 50.5, CH CF ; kcal mol− ) 3 2 2 3 · for these two systems were calculated as 53.5 and that are rather close to those estimated from the ki- 1 48.4 kcal mol− at the more accurate LDA/NL level, netically determined activation parameters and to · 1 respectively (or 72.6 and 66.0 kcal mol− at the LDA the LDA/NL values reported by Folga and Ziegler. · level) and these authors made the explicit suggestion It was particularly rewarding to observe the same that the previously published experimental values trend from the calculations and the experimental are too low. However, as mentioned in Section 4.5, kinetics, with a strojnger bond for the CF3 com- the DFT methods are not quantitatively reliable and pound, a weaker one for the CHF2 compound, and there may be questions about the suitability of the an intermediate strength for the CH2CF3 compound. chosen theory level. No further studies on these Incidentally, the DFT investigation was extended bonds have apparently appeared until our own re- to all F-substituted ethyl groups, yielding the results cent experimental and computational reinvestiga- shown in Figure4 (left). The investigation shows a tions [5,90]. bond strengthening upon introduction of α and β 1 The bonds in question are much stronger than F substituents. There is a 3 kcal mol− increase α α > · for any of the experimentally investigated com- from C H2 to C F2 with the greater difference be- ‡ ing associated to the introduction of the second α- pounds reported until then. The highest ∆Ha value available in the literature at that time was F substituent and a quantitatively even stronger and 1 apparently 39.5 1.0 kcal mol− for compound continuous BDE increase upon addition of β-F sub- 1 β β III ± 3 · stituents ( 7 kcal mol− from C H to C F ). An- [(H2O)5Cr –CH2-p-pyH] +, investigated in the 55– > · 3 3 other organometallic system, T [CoII(acac) ], yields 64 °C range [39]. We could find a suitable trap- = 2 ping agent (tris(trimethylsilyl)silane, TTMSS) and a qualitatively identical bond strengthening trend (Figure4 center), whereas the opposite trend (de- experimental conditions (70–100 °C in C6D6 solution, TTMSS/Mn 10) yielding saturation kinetics creasing BDE by both α-F and β-F substitution) oc- =I curs for T I. The reason for the different trends for the [(CO)5Mn –RF] disappearance (RF CF3, = = is related to the opposite bond polarity and to the CHF2 and CH2CF3). The resulting bond homolysis consequently opposite effect of the electronegative activation parameters are ∆Ha 53.8 3.5(CF3), = ± 1 F atoms on the energetic cost of the charge reor- 46.3 1.6(CHF2), 50.6 0.8(CH2CF3) kcal mol− ; ± ± · ganization that is associated to the homolytic bond and ∆Sa 66.0 9.5(CF3), 55.8 4.7(CHF2), = ± 1 1 ± cleavage [90]. I’ll come back to the cobalt system in 65.4 2.2(CH CF ) cal mol− K− . In spite of the ± 2 3 · · very strong bonds, the very high activation en- Section 5.3. tropies (higher than for any previously investi- x 1 gated L/Mt + –R bond homolysis) bring the ∆Ga values in a suitable range for kinetic monitor- 5. A few OMRP tales 5 1 ing, namely k in the (0.85–63) 10− s− range a × (half-lives between 23 h and 18 min). On the ba- I now wish to tell a few short stories on how homolyt- sis of the usual assumption of diffusion-limited ically weak metal–carbon bonds can aid, or play 1 recombination (∆H 2 kcal mol− ), the esti- havoc, in an area of strong interest and research ac- da ≈ · mated BDEs for the CF3 and CHF2 systems (51.8 tivity for the polymer chemistry community, namely 1 and 44.3 kcal mol− , respectively) are quite close to controlled radical polymerization (CRP). This poly- · those calculated by Folga and Ziegler at the LDA/NL merization strategy relies on the reduced impact of level and much higher than those previously ob- terminations relative to propagation, to the point tained by microcalorimetry and by photoionization of yielding quasi-living chain-growth. In terms of mass spectrometry. We have further investigated the the general radical reactivity shown in Scheme1, bond homolysis by the DFT approach [90] using a the useful chemistry is the repetitive addition of dispersion-corrected hybrid functional (BPW91*- the radical polymer chain-end to the monomer

D3), a functional that gave matching BDE values for (Pn• M Pn• 1). This leads to polymers of controlled + → +

C. R. Chimie — 2021, 24, n 1, 147-175 Rinaldo Poli 159

Figure 4. DFT-calculated T-CFnH2 nCFmH3 m BDEs (kcal/mol) (n 0, 1, 2; m 0, 1, 2, 3). T Mn(CO)5 − − = = = (left), Co(acac)2 (center), I (right). Reproduced with permission from Ref. [90]. Copyright 2018 Elsevier Science.

R0–Mn-T composition with chain-end functional- BDE, as already discussed in Section 4.2. Steric ity (R0 and T) approaching 100% and narrow mo- labilization of the metal–carbon bond makes the lar mass dispersities around the average degree of dormant species more easily reactivated, allowing polymerization (n), which is targeted on the basis CRP to be extended to less reactive monomers, i.e. of the initial molar monomer/initiator ratio. There associated to more reactive chain-end radicals. An are several ways to achieve this control, depending opportunity to test this idea came when Kevin M. on the nature of the selected controlling agent (T) Smith, a former post-doc of mine and now Pro- used in combination with a conventional radical fessor at UBC Okanagan, offered to test his half- II source (producing the primary R0• radical) or in a sandwich Cr complexes as moderators. He had ob- II Ar,Ar0 unimolecular R0–T initiator. Two different families tained a series or [CpCr (nacnac )] complexes Ar,Ar of methods can be distinguished on the basis of the (nacnac 0 ArNC(CH3)CHC(CH3)NAr0)[99] and = Pn–T dormant species activation, i.e. dissociative or had shown that they could be converted to stable III Ar,Ar associative, named respectively “reversible termina- [CpCr (nacnac 0 )(CH3)] complexes by oxidation” (or “reversible deactivation”) and “degenerative tion/alkylation, whereas derivatives with larger alkyl transfer” [91]. I’ll mostly focus on the dissociative ac- groups only led to decomposition products pre- III tivation method and on the use of a metal complex as sumably resulting from Cr –alkyl bond homolysis the radical concentration moderator (T L/Mtx). Re- and subsequent radical reactivity. Of particular in- x = terest was the comparison between the symmetric versible Pn• trapping by L/Mt yields an organometallic dormant species L/Mtx 1–P . For this reason, I (Ar Ar0 2,6-C6H3R2) complexes with R Me(Xyl) + n = = = have named this approach “organometallic radical or iPr (Dipp), which yield a quite different steric bulk polymerization” and gave it the OMRP acronym [92]. in the moderating species. In common usage, the method has later become Vinyl acetate (VAc), CH CHOCOCH , attracts 2 = 3 known as “organometallic-mediated radical poly- great attention because the corresponding polymer merization” [93–98], which is not in my mind a (PVAc) is the precursor of water-soluble poly(vinyl completely appropriate name, because the me- alcohol). Therefore, the incorporation of PVAc blocks diating agent is not necessarily organometallic; in tailored polymer chains opens the way to many only the dormant species is. However, the OMRP useful applications. The radical polymerization of acronym stuck. this monomer generates non-stabilized, relatively reactive PVAc–CH2CH•(OCOCH3) chain ends by 5.1. Steric control and ligand design the dominant regular (head-tail) monomer additions and even less stabilized, more reactive PVAc–

It has occurred to me that OMRP has a clear ad- CH(OCOCH3)CH2• chain ends by the less frequent vantage, relative to all other CRP strategies that use (ca. 1–2%) [100] inverted or head-head monomer ad- small atoms or groups as T (e.g. halogens, ONR2): the ditions. Earlier attempts to control the VAc polymer- steric parameter can be modulated by ligand engi- ization with capping halogens (i.e. by ATRP) [101] neering with a profound eﬀect on the metal–carbon or nitroxides [102] led to very slow conversions or

C. R. Chimie — 2021, 24, n 1, 147-175 160 Rinaldo Poli

to inhibition. It was thus of interest to test the half- tion slowdown for reasons that will be detailed in sandwich CrII complexes as moderators. the next section. It actually occurs by the associative III Polymerizations initiated by the labile V- activation method, whereas the [(acac)2Co -PVAc] 70 azo initiator (R –N N–R with R bond in the dormant chains is too strong to yield 0 = 0 0 = (CH ) C(OCH )CH C•(CH )(CN); t 10 h significant polymerization rates by reversible disso- 3 2 3 2 3 1/2 = at 30 °C) in the presence of the less bulky ciation, but dissociative activation can be promoted [CpCrII(nacnacXyl,Xyl)] system did not give any ev- by the addition of monodentate donor ligands [106]. idence of reversible chain trapping of polystyrene This phenomenon is related to the moderator stabi- II radical chains (same polymerization rate as the lization by coordination, [Co (acac)2(L)n](n 1,2), = metal-free control). Conversely, the polymerization and will not be further detailed here, since it does of VAc was much slower than the metal-free control, not involve a steric modulation of the CoIII–C BDE. but not fully inhibited. This indicates reversible CrIII– Of greater interest to the present discussion is the III C bond formation with the PVAc• chains, whereas steric modulation of the Co –PVAc BDE, in the ab- the PS chain-end radical does not form a suffi- sence of additional donor ligands, by operating on ciently strong bond. On the other hand, the bulkier the β-diketonate ligand scaffold. Indeed, the related [CpCrII(nacnacDipp,Dipp)] complex allowed a faster bis(2,2,6,6-hexamethylhepta-3,5-dionate) complex, VAc polymerization that exhibited the expected traits [Co{tBuC(O)CHC(O)tBu}2], induced a faster VAc of CRP, albeit with less than ideal control [103]. polymerization in the dissociative activation regime, Kevin was later able to obtain a well-defined CrIII while at the same time and for the same reason de- alkyl compound with a bulky neopentyl group, creasing the positive acceleration effect of added III Xyl,Xyl [CpCr (nacnac )(CH2tBu)], which proved an donor ligands [107]. excellent single-molecule initiator for a slow but rel- It is also of interest to point out that the VAc atively well-controlled VAc polymerization [54,104]. polymerization has also been controlled by por- Interestingly, an analogous alkylchromium(III) phyrin [108–110] and Schiff-base [111–114] cobalt(II) complex with the bulkier nacnacDipp,Dipp could complexes. These polymerizations only occur by de- not be obtained. Computational studies of the generative exchange, by photoinduced bond cleav- III Ar,Ar [CpCr (nacnac )-CH(CH3)OCOCH3] system, age, or by addition of donor ligands, whereas the dis- where the alkyl group models a poly(vinyl acetate) sociative thermal activation of the dormant chains is chain, confirmed the significant steric labilization completely inhibited in spite of the high ligand steric for the CrIII–C bond homolysis (∆E 25.9, 20.6 and bulk in some cases (e.g. tetramesitylporphyrin). This 1 = 18.6 kcal mol− for Ar Ph, Xyl and Dipp, respec- suggests that the O4 coordination sphere in the · = III tively) [54]. Even though clearly illustrating the prin- bis(β-diketonate) systems labilizes the Co –C bond ciple of ligand-based steric labilization in OMRP, relative to the N4 and N2O2 coordination spheres this study remains of purely academic interest for of porphyrin and Schiff base ligands. Indeed, fur- two reasons. The first one is the elaborate synthesis, ther playing with electron delocalization in the con- fragility (air sensitivity) and toxicity of the moder- jugated O,O-bidentate ligand by switching to the ating agent, which remains as chain end group in 9-oxyphenalenone (OPN) ligand system, the CoIII– the macromolecular product. The second reason is PVAc bond was weakened to such a point where related to a slowdown phenomenon, which results the presence of competitive catalytic chain transfer from the inverted monomer additions. I’ll return to (CCT) could be highlighted for the first time in VAc this phenomenon in the next section. polymerization [115] (see Section 5.4 for further de- Another steric labilization phenomenon was tails on the CCT phenomenon). highlighted in my laboratory in the same period for a cobalt bis(β-diketonate) system. Work by Antoine 5.2. Effect of the inverted monomer additions for Debuigne et al. had previously shown the excellent VAc performance of a simple and commercially available II II compound, [Co (acac)2], as moderating agent for How could [Co (acac)2] sustain a well-controlled the VAc radical polymerization [105]. This polymer- PVAc• chain growth without any slowdown or loss ization is not negatively affected by a polymeriza- of control? The primary PVAc–CH(OCOCH3)CH2•

C. R. Chimie — 2021, 24, n 1, 147-175 Rinaldo Poli 161

chain ends that are occasionally obtained (ca. 1– 2%) by inverted monomer addition are expected to yield stronger, hence less easily reactivatable, PVAc– x 1 CH(OCOCH3)CH2–Mt + /L bonds relative to the bond established by the regular secondary chain- x 1 end radical, PVAc–CH2CH(OCOCH3)–Mt + /L. Such a continuous accumulation of less easily reactivatable dormant chains is anticipated to in- duce a decrease of polymerization rate in a dissociative activation mode, as indeed observed for L/Mtx [CpCrII(nacnacXyl,Xyl)] (previous sec- = tion) [104], and a loss of control in the degenerative transfer activation mode. There are only two pos- II sible explanations: either [Co (acac)2] is somehow able to reduce the frequency of the head-head additions, or the CoIII–C bond to the primary radical in the dormant species is not significantly stronger (or is weaker) than the bond obtained after trapping the regular secondary chain-end radical. The first possibility would imply a monomer addition to the II caged {PVAc•,Co (acac)2} fragment pair with an in- fluence of the cobalt complex on the relative addition barriers. In a collaborative effort with Antoine De- buigne [116], the first option was explored by a thorough NMR investigation of the recovered PVAc– III Co (acac)2 macromolecules (both in-chain and Figure 5. Above: relative enthalpies and chain-end monomer configurations), revealing that optimized geometries of the species impli- the inverted monomer frequency is undistinguish- cated in the deactivation process of the tail able from that of free radical polymerization. Thus, (left) and head (right) PVAc radical models II either propagation occurs on the uncaged radical, or by [Co (acac)2] (reproduced with permis- II the [Co (acac)2] proximity has no significant effect sion from Ref. [116]; copyright 2013 American on the relative monomer addition barriers. The sec- Chemical Society). Below: activation enthalpies 1 ond option was tested by DFT calculations, using five (in kcal mol− ) for the tail and head dormant · different functionals and the CH3(CH3COO)CH• and species, and their difference, obtained with (CH3COO)CH2CH2• radicals to model the regular- different functionals. and inverted-monomer PVAc chain ends. This is a case where the type of functional has a dramatic effect on the BDE, particularly because a spin state enthalpies obtained with all functionals varies in a change occurs during the bond cleavage process: very narrow range. Furthermore, these differences III the diamagnetic [(acac)2Co –CH(OOCCH3)CH3] are very close to zero, suggesting that the two dor- III and [(acac)2Co –CH2CH2(OOCCH3)] species mant species should be reactivated at nearly equiv- II yield the organic radical (S 1/2) and a spin alent rates. The feature making this [Co (acac)2] II = quartet (S 3/2)[Co (acac)2] complex. The moderating agent “special” is its coordinative unsat- III = II [(acac)2Co –CH(OOCCH3)CH3] BDE was calcu- uration. Bond formation between [Co (acac)2] and 1 lated as low as 9.3 kcal mol− with the hybrid B3LYP the radical yields a diamagnetic 5-coordinate com- · 1 functional and as high as 34.2 kcal mol− with the plex with a square pyramidal geometry and an axial · diffusion-corrected M06L functional, see Figure5. alkyl group. This 16-electron species can be further However, the difference between the two activation stabilized by saturation of the vacant coordination

C. R. Chimie — 2021, 24, n 1, 147-175 162 Rinaldo Poli

site. In the absence of added donor ligands, chela- compounds (Figure4, right) [90], showing that the tion by the ester carbonyl function of the Co-bonded associative exchange of PVDF–CF2CH2–I with the monomer unit prevails for entropic reasons over dominant PVDF–CH2CF2• radical chains is endoer- coordination of an external monomer. Thus, the reg- gic (non-degenerate) because the PVDF–CF2CH2–I ular (secondary) chain-end model radical (Figure5, bond is stronger than the PVDF–CH2CF2–I bond. above right) forms a better stabilizing 5-membered Bruno Améduri attracted my attention to the VDF chelate ring, whereas the inverted (primary) chain- CRP problem. With his group at the ICG Montpel- end model radical (Figure5, above left), which yields lier, he had already shown that the VDF polymer- as expected a homolytically stronger bond, forms ization using another degenerate transfer method a poorer stabilizing 6-membered chelate ring. The of control, which uses a xanthate (R0–SC(S)–OEt) opposite trends of the CoIII–C BDE and chelate as reversible chain transfer agent, yields a similar ring stabilization effects re-equilibrate the needed loss of control beyond relatively small targeted de- enthalpy cost for the reactivation of the two dor- grees of polymerization [126]. In a first collabora- mant species. The chelated nature of the PVAc tive effort, by performing again DFT calculations chain end in the dormant species had been ex- on model compounds, we showed that the tail dor- perimentally established by a previous NMR and mant chains, PVDF–CF2CH2–SC(S)OEt, have a ho- III IR investigation of a small [(acac)2Co –VAcn–R0] molytically stronger C–S bond than the head ones, oligomer mixture (n 4) [117], which can be iso- PVDF–CH CF –SC(S)OEt [127]. Thus, the behavior ∼ 2 2 lated from the V-70-initiated VAc polymerization of the VDF degenerate transfer polymerization with II in the presence of a large [Co (acac)2] excess. Inci- iodoalkanes and xanthate transfer agents is quali- dentally, this product is an excellent unimolecular tatively the same. The kinetic model indicated that initiator for the polymerization of VAc and other control was lost after all chains are trapped in the monomers [117–120]. tail isomeric form but also suggested that these dormant species are not dead. They can be reacti- 5.3. Effect of the inverted monomer additions for vated, through too slowly to ensure reasonable con- VDF trol. Indeed, the 100% tail PVDF–SC(S)OEt product could be reactivated by chain extension with VAc, Vinylidene fluoride (VDF), CH CF , is another yielding well-defined PVDF-b-PVAc diblock copoly- 2 = 2 less activated monomer of great interest for a vari- mers [128]. The DFT study could rationalize this phe- ety of high-tech applications and polymerizes only nomenon. A more interesting question, however, was by the radical mechanism [121]. Its polymerization whether the two types of dormant chains, PVDF– with a high level of control has long been sought CH2CF2–T and PVDF–CF2CH2–T, could equally well but is hampered by the high inverted monomer be reactivated through an OMRP approach. The addition probability. The chain growth mainly in- L/Mtx moderator choice was naturally oriented to- II volves regular head-tail additions to yield the head ward [Co (acac)2] because of its proven ability to yield homolytically weak bonds with the PVAc radi- PVDF–CH2CF2• radical, while the minor (ca. 4–5%) cal chains. head-head additions lead to the tail PVDF–CF2CH2• radical. Both radicals are highly reactive and yield As already shown in Figure4 (center), our strong PVDF-T bonds in the dormant species. DFT calculations led to the prediction of a bond Iodine-transfer polymerization (working through strengthening upon introduction of F atoms on the degenerative transfer principle) was the first the alkyl ligand, at both the α and β positions, like methods to produce controlled PVDF chains, though for the [(CO)5Mn-Et] system (Figure4, left), and with limited control [122]. The accumulation of the opposite to the trend observed for I-Et (Figure4, less easily reactivatable tail dormant species, PVDF– left) and EtOC(S)S–Et [90]. In particular, the cost CF2CH2–I, was assumed to be responsible for this of the reactivation of the tail and head dormant limitation and higher degrees of polymerization chains is predicted to be quite similar, even slightly 1 could only be obtained using chemical tricks for in favor of the tail species, 26.0 kcal mol− for the · 1 this species reactivation [123–125]. This hypothesis [(acac) Co–CH CHF ] model, versus 27.4 kcal mol− 2 2 2 · was later validated by DFT calculations on model for the head [(acac)2Co–CF2CH3] model. Encouraged

C. R. Chimie — 2021, 24, n 1, 147-175 Rinaldo Poli 163

by these results and with the help of Debuigne’s III oligomeric [(acac)2Co –VAc 4–R0] initiator, we pro- ∼ II ceeded to experimentally test the [Co (acac)2]- mediated OMRP of VDF and we were very pleased with the result: the VDF polymerization proceeded with an unprecedented level of control up to high degrees of polymerization ( 100; 4 inverted monomer > > additions) with a linear Mn versus conversion plot and low M /M values ( 1.3) [129]. Subsequently, w n < we could show similar control for the polymeriza- II tion initiated by the combination of [Co (acac)2] and a conventional radical initiator, increasing the practical value of this polymerization process [130].

5.4. Catalytic chain transfer (CCT): β-H elimination or homolysis/transfer?

Catalytic chain transfer entails transfer of a chain- Scheme 8. [L/FeII] complexes used as ATRP end β-H atom to the L/Mtx catalyst in a first step, catalysts for styrene polymerization and pro- yielding a dead chain with an unsaturated ω end posed interplay of CCT via the OMRP dormant (macromonomer) and a hydride intermediate x 1 species. L/Mt + –H. In the second step, this hydride complex transfers the H atom to the monomer, generating a new growing chain with an H α chain end. β-H atom transfer from the radical chain to the metal A chain transfer catalyst operates through the same center, which was the established dogma until then, oxidation state and coordination number changes but proposed the radical trapping/β-H elimination as an OMRP moderator and an ATRP catalyst. Thus, sequence to rationalize their observations. the same molecule can promote all three processes, A few years later, Michael P. Shaver (who also as was first observed in my laboratory working with coauthored the previous contribution with Gib- a half-sandwich MoIII system [131–134]. Therefore, son) and his coworkers demonstrated that an ap- this is an unwanted phenomenon in CRP, although parently related system supported by a diamino- it is of importance in industry, if L/Mtx has a very bis(phenolate) ligand (B in Scheme8) is able to con- high CCT activity, to obtain controlled molar mass trol the polymerization of styrene by both ATRP and macromonomers [135]. OMRP mechanisms, without any significant CCT In 2006–2007 [136–139], Gibson et al. showed that contribution. These polymerizations were initiated II III [Fe Cl2(α-diimine)] compounds (A in Scheme8), in using the stable [L/Fe –Cl] complex and a conven- combination with organohalide initiators, catalyze tional radical initiator (reverse ATRP conditions), the ATRP of styrene, although the process was af- generating the [L/FeII] catalyst in situ [140,141]. On fected by variable degrees of CCT depending on the basis of this result, if CCT indeed requires β-H the ligand substituents. These authors proposed that elimination from the OMRP dormant species, it is the CCT pathway was favored by an increased “car- unclear why A, which is a poorer trapping agent for bophilicity” of the ATRP catalyst and that this corre- the polystyrene radical chain, would lead to CCT II lated with the spin state of the [Fe Cl3(α-diimine)] whereas B, which more favorably leads to the OMRP atom transfer product. In their proposed scheme, dormant species, does not. the increased carbophilicity would promote the di- In collaboration with Shaver, I have therefore car- rect chain trapping by L/Mtx to generate the OMRP ried out a DFT study of these systems. In a first dormant species, which would then lead to the hy- contribution, we could rationalize the better per- dride intermediate by β-H elimination. These au- formance of the B system in ATRP/OMRP when thors did mention that CCT may also occur by direct R1,R2 Cl, Cl relative to the tert-butyl-substituted =

C. R. Chimie — 2021, 24, n 1, 147-175 164 Rinaldo Poli

x 1 Scheme 9. Two pathways for alkene elimination from L/Mt + –R and for the reverse alkene addition to x 1 L/Mt + –H.

ligand system. This is a consequence of the induc- catalysts render the background CCT process less sig- tive electron withdrawing effect of the phenolato nificant. In conclusion, the β-H elimination hypoth- Cl substituents, leading to minor but determining esis for these organoiron(III) systems has to be aban- energy differences that make both the ATRP and doned and the homolysis/β-H transfer dogma holds the OMRP trapping processes more favorable [55]. for CCT. In a subsequent contribution, we could show that I wish to further comment on the alkene elim- III x 1 the organometallic intermediate L/Fe –PS, mod- ination from L/Mt + –R compounds (also named elled as L/FeIII–CHMePh in the calculations, has “dehydrometallation”) and its microscopic reverse, x 1 a stronger bond for B because the strain of the the alkene insertion into the L/Mt + –H bond. Sev- tetradentate diaminobis(phenolate) ligand raises the eral contributions (that I shall not cite) on metal- relative energy of the L/FeII system, whereas A can promoted or catalyzed radical reactions using 3d relax to the preferred tetrahedral coordination envi- transition metals (mostly Cr, Mn, Fe, Co, Ni and ronment. The establishment of a FeII CHMePh in- Cu) invoke the ubiquitous β-H elimination (path ··· teractions for A provides an insignificant stabiliza- a in Scheme9) as the dehydrometallation mech- 1 1 x 1 tion (1.0 kcal mol− , versus 13–15 kcal mol− for the anism. However, the L/Mt –R bonds for those · · + BDE in B depending on the L/ substituents). There- metals, at least in the oxidation states presumed in fore, the pathway leading to CCT proceeds directly by the proposed mechanisms, are homolytically quite H-atom transfer. This pathway occurs entirely along weak and more likely follow the homolysis and β-H the spin quartet surface, since the ground state of the atom transfer pathway (path b), like the organo- hydride product (S 3/2) is equivalent to the anti- iron(III) system above. Path a is typically followed by = ferromagnetic combination of the [L/FeII] complex complexes with homolytically strong metal–carbon and styryl radical spin states (2 and 1/2) [142]. Fur- bonds. Furthermore, path a requires a cis-vacant − thermore, we discovered that, contrary to the ear- coordination site, whereas path b does not have this lier proposition [136], there is no spin state control requirement. It only requires a homolytically weak of CCT: the L/FeII systems are always spin quintets, Mt–C bond. Thus, all β-H elimination claims for III x 1 the L/Fe –Cl systems are always spin sextets and 3d-metal L/Mt + –R intermediates in radical pro- the L/FeIII–CHMePh systems are always spin quar- cesses should be reconsidered. Conversely, the tets, independent on the ligand system (Cl2-diimine reverse path b process (named “hydrogen atom or diaminobis(phenolate)) and on the ligand sub- transfer” or HAT) is consciously and universally in- stitution pattern. The different aptitude of the α- voked by those practicing radical organic chem- diimine systems with different ligands to promote istry as a way to trigger radical transformations. CCT has a simpler explanation. It can be traced to This chemistry is initiated by reactive metal hy- electronic effects that significantly alter the ATRP ac- dride intermediates with homolytically weak Mt–H x 1 tivation barrier, whereas the H-atom transfer barrier bonds, often made in situ from a stable L/Mt + –Y is essentially unaffected. Thus, the more active ATRP precursor (halide, alkoxide, acetylacetonate, etc.)

C. R. Chimie — 2021, 24, n 1, 147-175 Rinaldo Poli 165

and a reducing hydride (silane, stannane, borane, OMRP conditions, but the results were not excit- etc.) [143]. However, the produced radical is then ing and were not published. Meantime, Kris and his proposed to pursue its useful chemical transforma- students kept increasing the ATRP catalytic activ- tions, in most cases without consideration of the ity of L/CuI complexes through an increase of the possible L/Mtx intervention as a moderating agent. moderating equilibrium constant by ligand engineer- I’ll return to this point in Section6. Metal hydride ing. This mostly involved the tris(pyridylmethyl)- complexes with homolytically strong Mt–H bonds, amine (TPMA) ligand family (Scheme 10)[149]. A 1 2 on the other hand, if able to provide facile access to a donor power increase (TPMA TPMA∗ TPMA∗ 3 NMe2 < < cis-vacant coordination site, are more likely to adopt TPMA∗ TPMA ) exerts a greater stabilizing < < path a (coordination/insertion) as is well-established effect on L/CuII–Y than on L/CuI, as evidenced by in metal-catalyzed olefin polymerization as the ini- the redox potentials, resulting in a KATRP increase by tial step after a chain transfer event by β-H elimina- several orders of magnitude. The current champion I NMe2 tion. As a short summary: the transformation shown on the activity scale, [Cu (TPMA )]+, was recently in Scheme9, in both directions, prefers path a for synthesized in a collaborative effort [150]. x 1 homolytically strong Mt–R/Mt–H bonds and path b As already pointed out in Section 4.2, a L/Mt + –R for homolytically weak Mt–R/Mt–H bonds. bond strengthening by a greater ligand donor power had been demonstrated by previous investigations 5.5. Catalyzed radical termination (CRT) on organocobalt(III) compounds. The same phenomenon may therefore be anticipated for L/CuII– In an ATRP process, the L/Mtx ATRP catalyst may also R. However, the lower polarity of CuII–R relative to II form metal–carbon bonds with the growing radical Cu –Y bonds further suggests that KOMRP should be chain, positively contributing to the polymerization less affected than KATRP. Indeed, this was demon- control by providing an additional moderating effect. strated (see Figure6) by the already cited electro- As already stated above, this OMRP/ATRP interplay chemical study [73] (Section 4.4). A deeper investiga- was first demonstrated for a MoIII system in my lab- tion of the interaction between growing poly(n-butyl oratory [131], but was also later shown to occur for acrylate) radical chains and L/CuI, using the very ac- II II 3 Os [144,145] and for Fe [55,140,141]. tive TPMA∗ ligand, revealed that the observed poly- For a number of years, I have had a most fruit- merization rate decrease results from a new phenom- ful collaboration with Krzysztof Matyjaszewski (Kris enon, not previously witnessed for any other OMRP for his friends), who co-discovered ATRP [146] and system, namely the catalytic action of the L/CuI com- is one of the main players in this area, mainly work- plex in radical termination. Since then, this phe- ing with the L/CuI–L/CuII–Y system. The question nomenon has also been highlighted for a FeII cata- of whether organocopper(II) species are present in lyst [151]. L/CuI-catalyzed ATRP has intrigued me since the A subsequent study with different ligands has beginning of our collaboration, because the natu- shown that the CRT activity, like the ATRP activity, ral homolytic weakness of CuII–C bonds, as sug- scales with the ligand donor power and involves the gested by the paucity of stable alkylcopper(II) com- formation of L/CuII–R, which is then capable of pro- pounds [147], lends hope for a controlled polymer- moting the interaction between the trapped radical ization of less activated monomers by the OMRP and a second radical [152]. The intermediacy of the approach using L/CuI as chain trapping agent. In- organometallic complex is also consistent with the deed, Kris had already shown in 1998, through the absence of CRT for the polymerization of methacry- observation of a retardation effect on the polymer- lates, since the tertiary polymethacrylate chain-end ization rate, that L/CuI complexes interact with prop- radical does not form as strong a bond with the L/CuI agating poly(methyl acrylate) chains while L/CuII moderator as the secondary polyacrylate chain-end complexes do not, but the produced polymers did radical. To this day, however, the intimate mecha- not have the expected characteristics of a controlled nism of this radical termination process is not fully process and the nature of this interaction was un- elucidated. We do not even know, as yet, what frac- clear [148]. We have tested a few polymerizations tions of coupling and disproportionation products of VAc in the presence of L/CuI complexes under are produced by the CRT process, since the product

C. R. Chimie — 2021, 24, n 1, 147-175 166 Rinaldo Poli

Scheme 10. Ligands of the TPMA family used in ATRP and in the study of CRT.

variants (Scheme 11a), a L/Mtx(Y) complex is con- x verted to an L/Mt (R0) intermediate by interaction with the nucleophilic reagent R0–Z with elimination of Y–Z (e.g. MgClY in Kumada-type coupling). This intermediate then activates R–Y by either a radical pathway (atom transfer, AT) to gen- x 1 erate L/Mt + (R0)(Y) and R• followed by radical re- x 2 bound to yield L/Mt + (R)(R0)(Y), or by standard 2-electron oxidative addition (indicated in blue). The cycle is then completed by the reductive elimina- Figure 6. Comparison of ATRP (red) and OMRP tion of the cross-coupled product RR0. Alternatively, (blue) equilibrium constants for different L the activation of the nucleophilic and electrophilic systems, involving the activation of methyl 2- reagents may occur in the reverse order as shown bromoproprionate in DMF. Reproduced with in Scheme 11b. In the latter case, yet another pos- x 1 permission from Ref. [73]. Copyright 2019 sibility is that the L/Mt + –Y intermediate obtained American Chemical Society. by AT is first alkylated by R0Z and then R• adds to x 1 the resulting L/Mt + (R0) (variant indicated in red). Alkyl electrophiles seem to prefer the AT/rebound distribution is skewed by other competing phenom- radical pathway whereas aryl electrophiles undergo ena (e.g. conventional radical termination, reductive 2-electron oxidative addition. Either way, the typi- radical terminations) [153–155]. Further investiga- cally proposed final step is the RR0 reductive elimi- tions aimed at determining this product distribution nation, which is also represented in Scheme12 (path are ongoing. The possibility to use a L/CuI modera- a). Note that the same intermediate, under favor- tor for the OMRP of less activated monomers also re- able circumstances, may also evolve by β-H elimi- mains an open question. nation/reductive R-H elimination (path a0) to yield I now wish to draw a parallel between the CRT disproportionation products. phenomenon and the C–C bond forming step in- It is possible, however, to envisage an alter- volved in a family of powerful radical cross cou- native way in which the C–C bond forming step pling processes [156–159]. Several Fe-, Co-, Ni- and takes place. Since metal–carbon bonds are homolyt- Cu-catalyzed cross-couplings between an alkyl or ically weak for a 3d metal, it is conceivable that the x 2 aryl halide, R–Y, and a nucleophilic coupling partner L/Mt + (R)(R0) intermediate, even if it does form, R0–Z involve radical intermediates and many dif- may preferentially proceed by Mt–R bond homolysis, ferent catalytic cycles have been proposed, though followed by radical rebound on R0 (Scheme12, path rarely with sufficient supporting evidence. The most b). Indeed, a direct radical rebound to the metal- consensual cycles (though not the only ones) in- bonded aryl group, without formation of a dialkyl I II III 0 I II I II III x 2 volve Fe /Fe /Fe , Co /Co /Co , Ni /Ni /Ni , or derivative in the Mt + oxidation state, has been CuI/CuII/CuIII species. In one of many possible proposed for a few Fe-catalyzed Kumada radical

C. R. Chimie — 2021, 24, n 1, 147-175 Rinaldo Poli 167

Scheme 11. A few of many proposed catalytic cycles for radical cross-coupling processes.

excluded on the basis of the high rate constant, e.g. 8 1 1 (6 3) 10 M− s− for CH , whereas addition to the ± × · 3 metal would require dissociative replacement of an 1 NH3 ligand, which is known to be slow (1.8 s− for the III 2 trans labilized ligand in [(NH3)5Co (CH3)] +)[165]. Incidentally, a direct bond formation similar to that shown in Scheme 12 (path b) can also oc- Scheme 12. Radical rebound to the metal cen- cur when R0 H. This has been proposed for the = ter versus the metal-bonded R0 ligand in cross- [Co2(CO)8]-catalyzed hydrogenation of anthracene coupling. and derivatives [166]. In this reaction, [Co2(CO)8] and H2 yield [HCo(CO)4], which transfers the H atom to the substrate to yield [Co(CO)4•] and a stabilized cross-couplings involving R–Y and ArMgBr, namely anthacenyl radical. The latter is too stabilized and II does not form a metal–carbon bond. Rather, it re- with [Fe Cl2{1,2-[(3,5-R2C6H3)2P]2C6H4}] (R II = acts with a second [HCo(CO) ] molecule to yield tBu,SiMe3)[160,161], [Fe Cl2(IPr)2][162] and 4 II [Fe Ph (IPr Me ) ][163] (IPr Me 1, 3- the hydrogenated product and a second [Co(CO)4•], 2 2 2 2 2 2 = diisopropyl-4,5-dimethylimidazol-2-ylidene; IPr 1, which combines with the first one to regenerate = 3-diisopropylimidazol-2-ylidene) pre-catalysts. For [Co2(CO)8]. II II these systems, a cycle involving L(Y)/Fe , L(Ar)Fe If the alkyl radical R• rebounds on a metal-bonded III and L(Ar)Fe –Y species (without involvement of alkyl (rather than aryl) group R0, the possible abstrac- IV an unlikely Fe species) has been proposed. Of tion of a R0 β-H atom (path b0), leading to the dispro- ( H) great relevance to this alternative pathway for C– portionation products (R–H and R0 − ), also seems C bond formation, Meyerstein et al. demonstrated possible. This radical pathway is an alternative to x 2 the occurrence of path b for the reaction between the 2-electron path a0 from the LMt + (R)(R0) in- II 2 [(NH3)5Co (H2O)] + and radiolytically produced termediate, leading to the same disproportionation R• (R CH3, CH2COO−): a first radical adds to products. In this respect, a seminal investigation = III 2 generate a [(NH3)5Co –R] + transient, which then by Kochi et al. on the effect of the metal oxida- quenches a second R• to produce RR [164]. This tion state for the reductive elimination of dialkyl- pathway dominates relatively to the direct bimolec- iron complexes merits to be highlighted and com- ular coupling and to the heterolytic bond cleavage, mented [167]. Compounds [(bipy) FeIIR ] (R Et, 2 2 = which yields the alternative RH termination prod- nPr, among others) were thermally decomposed III 3 uct and [(NH3)5Co (H2O)] +. The possible addi- and the same study was carried out for their 1- tion of the second radical to the CoIII center was electron and 2-electron oxidation products, leading

C. R. Chimie — 2021, 24, n 1, 147-175 168 Rinaldo Poli

( H) to variable distributions of R–H, R − and R–R. The the possible contribution of reversible metal–carbon neutral FeII complexes decompose at 50 °C and yield bond formation. On the other hand, the involvement the disproportionation products selectively, with of organometallic intermediates is well-recognized 0.1% of R–R, which clearly suggests a 2-electron for systems leading to sufficiently strong Mt–C < β-H elimination/R-H reductive elimination path- bonds, allowing their isolation or spectroscopic way (a0 in Scheme 12). The 1-electron oxidation observation, notably organocobalt(III) species [168]. III products [Fe (bipy)2R2]+ are more thermally fragile, However, even bonds that are so weak as to hamper but sufficiently stable to be isolated and character- any spectroscopic detection (not to mention their ized. They decay by a first-order rate law with k in the isolation) may contribute to improve a radical reac- 1 1 (2.0–4.1) 10− s− range at 30 °C, leading to mixtures tion selectivity by moderating the radical concen- × of disproportionation and coupling products, the lat- tration and consequently reducing the impact of the ter being dominant (75–90%). The investigation ele- bimolecular terminations. gantly demonstrated that the majority of these prod- I could give a small contribution to this area ucts originate from an in-cage radical recombina- thanks to the solicitation of Patrick L. Holland. tion, with the small fraction of cage escape being He has previously investigated, in collaboration III III assessed by a [Fe (bipy)2Et2]+/[Fe (bipy)2(nPr)2]+ with Phil S. Baran, the mechanism of a versatile crossover experiment, which leads to the detection [Fe(acac)3]-catalyzed radical cross-coupling pro- of small amounts of pentane. The authors proposed cess, the scope of which had already been demon- that the first step of this process is bond homol- strated by Baran [169]. The reaction couples an II ysis to produce a {[(bipy)2Fe R]+,R•} caged pair, electron-rich alkene, R1CH CR2R3 (including = and stated that “this homolysis may be followed with heteroatom functionalities) with an electron- in rapid succession by the cleavage of the second poor alkene, R4CH CR5-EWG (EWG electron- = = alkyl-iron bond”. However, no further comments withdrawing group), in the presence of a silane or additional experiments were offered to interpret as hydride donor and ethanol (solvent) as proton the observed product distribution, except for not- 1 2 3 4 5 donor to produce R CH2CR R –CHR CR (EWG)H III ing that the [(bipy)2Fe Et2]+ decomposition yields in high yields and chemoselectivity. The Baran- a similar product distribution to the photolysis of Holland collaborative efforts led them to propose azoethane, EtN NEt. Clearly, after homolysis of = that the process starts by generation of an ac- the first bond, the residual bond in [(bipy) FeIIR] III 2 + tive [Fe H(acac)2] species, which is able to acti- should be homolytically stronger, thus a rebound of vate the donor alkene by HAT with generation of the caged radical onto the FeII-bonded alkyl group II 1 2 3 [Fe (acac)2] and a tertiary radical, R CH2C•R R . as shown in paths b and b0 of Scheme 12 is an al- The latter then adds to the acceptor alkene to yield ternative to be considered. In that case, the minor 1 2 3 4 5 R CH2CR R –CHR C•R (EWG) and final quench- disproportionation product may indeed result from ing occurs by capturing a proton from ethanol and a direct β-H atom abstraction. The second oxida- II an electron from [Fe (acac)2]. This final step pro- tion process, generating [(bipy) FeIVR ]2 , is elec- III 2 2 + duces [Fe (OEt)(acac)2], which was proposed to trochemically irreversible. Both chemical and elec- be the cycle resting state. Indeed, this compound trochemical oxidations selectively yielded R–R with was independently prepared and characterized as a ( H) only traces or undetectable amounts of R − , which diethoxo-bridged dimer and found to be catalytically demonstrates a clean 2-electron reductive elimi- III competent. The reaction of [Fe (OEt)(acac)2] with nation with no involvement of radicals for the FeIV III PhSiH3 then regenerates [Fe H(acac)2] to start the intermediate. next cycle [170]. However, a few points concerning this mechanism and the observed high selectivity re- 6. Metal complex moderators in organic radi- mained open: (i) the lack of experimental evidence of III cal reactions the [Fe H(acac)2] intermediate; (ii) the preference of the nucleophilic radical for addition to the acceptor In previous sections, I have commented on the lack alkene over reductive quenching; (iii) the preference of the systematic consideration, by those who prac- of the electrophilic radical for reductive quenching tice metal-mediated/catalyzed radical reactions, of over addition to a second acceptor alkene molecule

C. R. Chimie — 2021, 24, n 1, 147-175 Rinaldo Poli 169

(oligomerization); (iv) the low impact of bimolecular radical terminations; (v) the intimate mechanism of the reductive quenching step (stepwise outer-sphere electron transfer, OSET, or concerted proton-coupled electron transfer, PCET). With additional and intertwined experimental and computational mechanistic investigations, the full cycle could be elucidated (Figure7)[171]. The first part of the free energy profile, calculated using isobutene and methyl acrylate as model donor and acceptor alkenes, is shown in Figure8. The computed free energy difference between TS1 (turnover-determining transition state, TDTS) and [Fe(acac)2(OEt)]2 (resting state) gave the cycle free 1 energy span as 24.9 kcal mol− . This value is not · far from the number experimentally provided by Figure 7. Mechanism for Fe-catalyzed inter- 1 new kinetics experiments (22.8 0.2 kcal mol− , molecular alkene cross-coupling, supported by ± · from the turnover frequency through the Eyring re- experiments and computations. Reproduced 1 lationship). A 3.3 kcal mol− fraction of this amount with permission from Ref. [171]. Copyright · is the cost of splitting the dimer into monomeric 2019 American Chemical Society. III [Fe (OEt)(acac)2] (sextet ground state). The unob- III served [Fe H(acac)2] intermediate is nearly isoen- ergetic in all possible spin states (doublet, quar- tures with a wider variety of solvents, silane donors tet and sextet), the quartet being the ground state, and Fe starting materials, which delayed the revised while dimerization is unfavorable. H atom transfer manuscript submission by a few months, but no III from [Fe H(acac)2] to isobutene has a very small convincing signals could be found. Fortunately, the 1 barrier of 4.7 kcal mol− through TS2, whereas the reviewer and the editor were finally persuaded by our · back reaction has a much higher activation barrier arguments. The reason for the very high reactivity 1 III of 16.3 kcal mol− . It was hard to convince a re- of this [Fe H(acac)2] complex is obvious when con- · viewer, who demanded us to experimentally prove sidering that the calculated Fe–H homolytic BDE is the existence of this hydride intermediate as a con- only 17 kcal/mol, whereas the BDE of typical stable dition for our manuscript acceptance, that with such (through reactive) metal hydride compounds are in 1 an energy profile the detection of this compound the 55–75 kcal mol− range [172,173]. · would be impossible: the rate constant for forma- Of greater relevance to this article topic, the pro- II tion of the hydride complex from the ethoxide dimer duced {[Fe (acac)2,tBu•} caged pair is predicted ‡ 1 6 1 1 III (∆G 24.9 kcal mol− ) is k 3.53 10− s− M− ; to collapse to an organometallic [Fe (tBu)(acac)2] 1 = · 1 = × · ‡ 1 adduct with a quartet ground state, although the sta- the reverse step (∆G 1 16.3 kcal mol− ) occurs 1 − 1 = · with k 1 7.07 s− M− and the forward HAT step bilization provided by the “bond” is barely significant ‡ − = · 1 9 1 1 (∆G 4.7 kcal mol− ) with k 2.2 10 s− M− . (see Figure9). Still, a slight “persistent radical e ffect” 2 = · 2 = × · Using these rate constants and the known reagent (PRE) may be expected: from this ∆G and the catalyst concentrations, the upper limit on the hydride in- concentration used in the experiments, ca. 50% of termediate concentration is estimated from the the radicals are predicted to be protected as dormant 17 steady-state approximation as 6 10− M. This is organometallic species. This improves the efficiency × of course a rough approximation, using the stan- of the subsequent addition to the acceptor olefin, dard free energy values calculated at room temper- which occurs through a relatively small activation ature. It is conceivable that the situation may im- barrier (TS3). With all the caveats outlined in Sec- prove somewhat by lowering the temperature and tion 4.5, this value can only be taken as indicative, at Holland’s student Dongyoung Kim at Yale redoubled best. Calculations with a dispersion-corrected pure his efforts to detect this species at lower tempera- functional (BP86-D3) yielded a much stronger BDE

C. R. Chimie — 2021, 24, n 1, 147-175 170 Rinaldo Poli

III Figure 8. DFT-calculated energy proﬁle of the catalytic cycle in Figure7 from the [Fe (acac)2(OEt)]2 resting state to the Me3C• radical.

Figure 9. DFT-calculated energy proﬁle of the catalytic cycle in Figure7: PRE versus cross-coupling (TS3).

C. R. Chimie — 2021, 24, n 1, 147-175 Rinaldo Poli 171

1 II (25.1 kcal mol− ). However, the BPW91*-D3 func- moderating action by the [Fe (acac) ] compound · 2 tional selected for this investigation has given, in our generated in the HAT activation to reversibly trap the hands, more credible results for a number of other initial nucleophilic radical. I personally feel that a previously investigated systems of light transition similar moderating action is present for a great many metals (including those highlighted in the previous organic radical reactions involving transition metals, sections), as well as satisfactory agreement with three where such a phenomenon has been overlooked. different benchmarks during the present investigation: the first one has already been highlighted above 7. Conclusion (calculated cycle span versus experimental TOF). The second benchmark was the electrochemical reduc- I have presented a general analysis of the radical con- III II tion potential of the [Fe (acac)3]/[Fe (acac)3]− cou- centration moderation through the reversible forma- ple versus the ferrocene/ferricenium standard, mea- tion of homolytically weak metal–carbon bonds 1 sured as 0.48 V (∆G 11.1 kcal mol− from the and described how the thermodynamic and ki- − = + · 1 Nernst equation) and calculated as 9.6 kcal mol− . + · netic parameters of the bond homolysis equilib- The third benchmark was the EtOH binding equi- rium can be experimentally assessed. I have also II II librium to [Fe (acac)2] to yield [Fe (acac)2(EtOH)2] mentioned the useful contribution, as well as the II with no observable [Fe (acac)2(EtOH)] intermedi- caveats, of the computational analysis of the metal– 1 ate, measured as K 3.2 0.2 M− (∆G 0.69 carbon bond homolysis. This survey highlights sev- 1 = ± = − 1± 0.95 kcal mol− ) and calculated as 0.6 kcal mol− , eral opportunities for future investigations. One in- · II − · while the relative G of the [Fe (acac)2(EtOH)] inter- teresting avenue is the measurement of the homol- 1 mediate was calculated as 1.3 kcal mol− , in agree- + · ysis equilibrium constant by the kinetic approach ment with its non-observation. described in Schemes2 and3, which is potentially Finally, the OSET pathway for the prod- applicable to the very fragile bonds of compounds uct quenching step, involving sequential that cannot be isolated. The measurement of the ac- 1 tBuCH2CH•(COOCH3) reduction to enolate by tivation rate constant from the H NMR line broad- II [Fe (acac)2] and protonation, was found too en- ening, used so far only in a few cases, also appears ergetically costly. On the other hand, the PCET as a method of wider applicability for compounds pathway involving the proton of one of the co- with relatively strong bonds. It may also be useful II ordinated ethanol ligands in [Fe (acac)2(EtOH)2] to develop linear free-energy relationships for the 1 occurs with a low barrier of 9.4 kcal mol− (TS4). metal-alkyl BDE to analyze the relative importance · The DFT exploration also unveiled an unexpected of electronic, steric and resonance stabilization fac- alternative and competitive pathway (barrier of tors, similar to those already developed for alkyl- 1 9.9 kcal mol− ) involving the tBuCH2CH•(COOCH3) halogen and other alkyl-heteroatom bonds, though · II addition to [Fe (acac)2(EtOH)] through the O such relationships would be metal-specific. In ad- atom to yield the enolate [FeIII(acac) {O-C(OMe) dition, the weird low (even negative in some cases) 2 = CHCH2tBu}(EtOH)] followed by intramolecular pro- bond cleavage activation entropies determined in ton transfer, whereas the FeIII–C bond formation previous studies may be worthy of a re-evaluation for this radical is unfavorable. The direct quench- on the basis of a new kinetic scheme that considers ing of the tBu• radical by CPET is also energetically the direct trapping of the solvent caged fragment competitive (and is indeed experimentally observed pair (k4 in Scheme6), for instance by performing in the absence of acceptor olefin) but the addition these investigations again using different trapping to methyl acrylate is favored by the concentration agents. The electrochemical study of systems where bias, whereas the barrier to radical propagation with radicals can be generated in situ and trapped by methyl acrylate to make oligomers is slightly higher electroactive L/Mtx metal complexes to generate x 1 than for CPET. In conclusion, the high chemoselec- electroactive L/Mt + –R species, applied so far only tivity for this versatile radical reaction involves a del- to the investigation of organocopper(II) systems as icate balance of the relative barriers and relative con- shown in Section 4.4, might allow the investigation of centrations of the reagents involved in the various other compounds with quite thermally fragile metal– steps, and might benefit, in addition, from a small carbon bonds. Finally, our recent use of TTMSS as

C. R. Chimie — 2021, 24, n 1, 147-175 172 Rinaldo Poli

a radical trap for the successful determination of stabilization provided by the metal–carbon bond for- the [(CO)5Mn-RF] bond activation parameters (the mation is very low. strongest bonds ever measured by this kinetic approach), as shown in Section 4.6, could be extended Acknowledgments to the investigation of other strong metal–carbon bonds. My modest contribution to the area of metal–carbon In the second part of this article I have described bond homolysis has been inspired by stimulating a few investigations where the synergistic approach discussions with several scientists, mostly but not of DFT calculations and experiments carried out exclusively polymer chemists, from whom I have in my laboratory and in those of my collaborators learned a lot of fascinating science. My first and have provided insight into the contribution of re- strongest acknowledgement goes to them. They are, versible metal–carbon bond homolysis to the areas of in chronological order: Roger Spitz, Yves Gnanou, organometallic radical polymerization (OMRP) and Jérôme Claverie, Krzysztof Matyjaszewski, David M. metal-mediated organic transformations. In partic- Haddleton, the Liège CERM group (Robert Jérôme, ular, we have shown that L/Mtx moderating species Antoine Debuigne, Christophe Detrembleur), Prze- have the advantage of an Mt–C BDE steric modula- myslaw Kubisa, Kevin M. Smith, W. Stephen McNeil, tion through ligand engineering. We have also shown the Lyon C2P2 group (Bernadette Charleux, Franck II how the simple [Co (acac)2] moderator, for different D’Agosto, Muriel Lansalot), the Toulouse P3R team and unrelated reasons linked to monomer chelation (Matthias Destarac, Stéphane Mazières), Michael and bond polarity effects, re-equilibrates the BDEs of P. Shaver, the Montpellier IAM team (Bruno Amé- the head and tail dormant species obtained in the duri, Vincent Ladmiral), Patrick L. Holland and Yong CRP of two challenging monomers, VAc and VDF. Wang. A very close second and heartfelt thanks goes In that respect, homolytically weak metal–carbon to all the students and post-docs in my group who bonds make more robust mediating systems for the have practically and intellectually contributed to CRP of challenging asymmetric monomers [174]. We this area. They are, again in chronological order: have argued how reversible metal–carbon bond ho- Erwan Le Grognec, François Stoffelbach, Sébastien molysis may be the preferred pathway for dehy- Maria, José Mata, Ulrich Baisch, Yohan Champouret, drometallation of organometallic intermediates in Zhigang Xue, Aurélie Morin, Andrés Cardozo, Si organic radical transformations and confirmed that Chen, Ahmad Joumaa, M. S. Wahidur Rahaman, catalytic chain transfer polymerization does not re- Roberto Morales-Cerrada, Ekaterina V. Bellan, Lu- quire the formation of metal–carbon bonds followed cas Thevenin, Andrii Karpus, Hui Wang, Ramakr- by β-H elimination. We have also investigated the ishna Gandikota and Maxime Michelas. I would also L/CuI-catalyzed radical termination (CRT), which like to acknowledge the contribution of the visit- may feature the same elementary steps as certain ing students and post-docs from the laboratories 3d-metal catalyzed radical cross-couplings. The in- of my collaborators (Tomislav Pintauer, Wade A. timate mechanism of this process and the nature of Braunecker, Santosh Kumar K. S., Thomas G. Ribelli, the terminated chains (coupling or disproportiona- Daniel L. Coward, Marco Fantin, Jirong Wang, Julian tion) remains to be fully elucidated. It is also of in- Sobieski) and the permanent coworkers in my group terest to understand the origin of the CRT activity who have been involved in this topic (Eric Manoury, and its dependence on the radical nature, in order to Christophe Fliedel, Florence Gayet). This research engineer new L/CuI ATRP catalysts that do not pro- effort has enjoyed continuous support from the Cen- mote this unwanted process. Another specific goal is tre National de la Recherche Scientifique (CNRS) the development of new efficient OMRP moderators through the recurrent funding of my hosting insti- for less activated monomers based on L/CuI systems tute (LCC, UPR 8241) and through two specific grants and other metals. Finally, I have described recent for my collaboration with K. Matyjaszewski (PICS work on the mechanism of a Fe-catalyzed selective No. 06782, 2015–2017 and LIA No. 1240, 2018–2021), and versatile alkene radical cross-coupling, pointing from the Institut Universitaire de France throughout out that the reversible formation of organometallic the period of my active membership (2007–2017), intermediates needs to be considered, even when the from the Agence Nationale de la Recherche through

C. R. Chimie — 2021, 24, n 1, 147-175 Rinaldo Poli 173

several grants (“Novel-CRP”, No. NT05-2_42140, [23] B. B. Wayland, A. A. Gridnev, S. D. Ittel, M. Fryd, Inorg. Chem., 2006–2008, shared with Y. Gnanou; “OMRP”, No. 1994, 33, 3830-3833. NT10-BLAN-710 1, 2010–2013; “BIPHASNANOCAT”, [24] D. C. Woska, Z. D. Xie, A. A. Gridnev, S. D. Ittel, M. Fryd, B. B. Wayland, J. Am. Chem. Soc., 1996, 118, 9102-9109. n◦ ANR-11-BS07-025-01, 2011–2015, shared with B. [25] W. Tang, N. V.Tsarevsky, K. Matyjaszewski, J. Am. Chem. Soc., Charleux; “FLUPOL”, No. ANR-14-CE07-0012, 2015– 2006, 128, 1598-1604. 2018, shared with B. Améduri; “RAFTSWITCH”, No. [26] D. Meyerstein, H. A. Schwarz, J. Chem. Soc., Faraday Trans I, ANR-16-CE29-0014-02, 2017–2020, shared with S. 1988, 84, 2933-2949. Mazières; “POLYSWITCH”, No. ANR-19-CE07-0031- [27] H. Cohen, D. Meyerstein, Inorg. Chem., 1988, 27, 3429-3431. [28] J. Halpern, Polyhedron, 1988, 7, 1483-1490. 01, 2020–2022), and from the European Commission [29] J. P.Collman, L. McElweewhite, P.J. Brothers, E. Rose, J. Am. through a Marie Curie IIF grant (“METMED-CRP”, Chem. Soc., 1986, 108, 1332-1333. No. 235249. 2009–2011, fellowship of Z. Xue). I am [30] R. G. Finke, B. L. Smith, B. J. Mayer, A. A. Molinero, Inorg. also grateful to the CALMIP mesocenter of the Uni- Chem., 1983, 22, 3677-3679. versity of Toulouse for the continuous allocation of [31] R. G. Finke, B. P.Hay, Inorg. Chem., 1984, 23, 3041-3043. [32] B. P.Hay, R. G. Finke, J. Am. Chem. Soc., 1986, 108, 4820-4829. computational resources to this project. [33] M. K. Geno, J. Halpern, J. Chem. Soc., Chem. Commun., 1987, 1052-1053. References [34] T.-T. Tsou, M. Loots, J. Halpern, J. Am. Chem. Soc., 1982, 104, 623-624. [1] J. Halpern, J. P.Maher, J. Am. Chem. Soc., 1965, 87, 5361-5366. [35] R. Morales-Cerrada, C. Fliedel, F.Gayet, V. Ladmiral, B. Amé- [2] P.B. Chock, J. Halpern, J. Am. Chem. Soc., 1969, 91, 582-588. duri, R. Poli, Eur. J. Inorg. Chem., 2019, 4228-4233. [3] J. F.Endicott, G. J. Ferraudi, J. Am. Chem. Soc., 1977, 99, 243- [36] C. G. Riordan, J. Halpern, Inorg. Chim. Acta, 1996, 243, 19-24. 245. [37] M. J. Nappa, R. Santi, S. P. Diefenbach, J. Halpern, J. Am. [4] J. Halpern, Acc. Chem. Res., 1982, 15, 238-244. Chem. Soc., 1982, 104, 619-621. [5] R. Morales-Cerrada, C. Fliedel, J.-C. Daran, F. Gayet, V. Lad- [38] M. J. Nappa, R. Santi, J. Halpern, Organometallics, 1985, 4, miral, B. Améduri, R. Poli, Chem. Eur. J., 2019, 25, 296-308. 34-41. [6] M. K. Geno, J. Halpern, J. Am. Chem. Soc., 1987, 109, 1238- [39] R. G. Coombes, M. D. Johnson, J. Chem. Soc. A, 1966, 177-182. 1240. [40] A. R. Schmidt, T. W. Swaddle, J. Chem. Soc. A, 1970, 1927- [7] G. N. Schrauzer, J. H. Grate, J. Am. Chem. Soc., 1981, 103, 541- 1932. 546. [41] R. S. Nohr, J. H. Espenson, J. Am. Chem. Soc., 1975, 97, 3392- [8] B. E. Daikh, R. G. Finke, J. Am. Chem. Soc., 1992, 114, 2938- 3396. 2943. [42] G. W. Kirker, A. Bakac, J. H. Espenson, J. Am. Chem. Soc., [9] H. Fischer, Macromolecules, 1997, 30, 5666-5672. 1982, 104, 1249-1255. [10] H. Fischer, Chem. Rev., 2001, 101, 3581-3610. [43] J. H. Espenson, P. Connolly, D. Meyerstein, H. Cohen, Inorg. [11] A. Studer, Chem. Eur. J., 2001, 7, 1159-1164. Chem., 1983, 22, 1009-1013. [12] D. Leifert, A. Studer, Angew. Chem. Int. Ed., 2020, 59, 74-108. [44] A. Bakac, J. H. Espenson, J. Am. Chem. Soc., 1984, 106, 5197- [13] S. Maria, F. Stoﬀelbach, J. Mata, J.-C. Daran, P. Richard, 5202. R. Poli, J. Am. Chem. Soc., 2005, 127, 5946-5956. [45] J. Halpern, S. H. Kim, T. W. Leung, J. Am. Chem. Soc., 1984, [14] Y. Zhang, K. Schröder, Y. Kwak, P. Krys, A. N. Morin, T. Pin- 106, 8317-8319. tauer, R. Poli, K. Matyjaszewski, Macromolecules, 2013, 46, [46] S. H. Kim, H. L. Chen, N. Feilchenfeld, J. Halpern, J. Am. 5512-5519. Chem. Soc., 1988, 110, 3120-3126. [15] K. Ohno, Y. Tsujii, T. Miyamoto, T. Fukuda, M. Goto, [47] F. T. T. Ng, G. L. Rempel, J. Halpern, Inorg. Chim. Acta, 1983, K. Kobayashi, T. Akaike, Macromolecules, 1998, 31, 1064- 77, L165-L166. 1069. [48] M. H. Schoﬁeld, J. Halpern, Inorg. Chim. Acta, 2003, 345, [16] H. Fischer, J. Polym. Sci., A: Polym. Chem., 1999, 37, 1885- 353-358. 1901. [49] D. C. Woska, B. B. Wayland, Inorg. Chim. Acta, 1998, 270, 197- [17] W. Tang, T. Fukuda, K. Matyjaszewski, Macromolecules, 2006, 201. 39, 4332-4337. [50] D. Bertin, D. Gigmes, S. R. A. Marque, P. Tordo, Macro- [18] H. A. Skinner, Experimental Thermochemistry, vol. II, Inter- molecules, 2005, 38, 2638-2650. science, New York, 1962. [51] J. L. Hodgson, C. Y. Lin, M. L. Coote, S. R. A. Marque, K. Maty- [19] J. A. Martinho Simões, J. L. Beauchamp, Chem. Rev., 1990, 90, jaszewski, Macromolecules, 2010, 43, 3728-3743. 629-688. [52] C. Y. Lin, S. R. A. Marque, K. Matyjaszewski, M. L. Coote, [20] J. Halpern, F. T. T. Ng, G. L. Rempel, J. Am. Chem. Soc., 1979, Macromolecules, 2011, 44, 7568-7583. 101, 7124-7126. [53] T. Koenig, R. G. Finke, J. Am. Chem. Soc., 1988, 110, 2657- [21] F. T. T. Ng, G. L. Rempel, J. Halpern, J. Am. Chem. Soc., 1982, 2658. 104, 621-623. [54] Y. Champouret, K. C. MacLeod, K. M. Smith, R. Poli, [22] F. T. T. Ng, G. L. Rempel, C. Mancuso, J. Halpern, Organometallics, 2010, 29, 3125-3132. Organometallics, 1990, 9, 2762-2772. [55] R. Poli, M. P.Shaver, Inorg. Chem., 2014, 53, 7580-7590.

C. R. Chimie — 2021, 24, n 1, 147-175 174 Rinaldo Poli

[56] E. N. G. Marsh, D. P. Ballou, Biochemistry, 1998, 37, 11864- saied, S. Suradi, R. Carson, G. Altakhin, H. A. Skinner, 11872. Organometallics, 1982, 1, 1166-1174. [57] R. Padmakumar, R. Padmakumar, R. Banerjee, Biochemistry, [88] A. E. Stevens, Fundamental Studies of Reactive Intermediates 1997, 36, 3713-3718. in Organometallic Chemistry, California Institute of Tech- [58] S. Chowdhury, R. Banerjee, J. Am. Chem. Soc., 2000, 122, nology, Pasadena, California, USA, 1981. 5417-5418. [89] E. Folga, T. Ziegler, J. Am. Chem. Soc., 1993, 115, 5169-5176. [59] J. F. Endicott, K. P. Balakrishnan, C. L. Wong, J. Am. Chem. [90] R. Poli, S. M. W. Rahaman, V. Ladmiral, B. Améduri, J. Soc., 1980, 102, 5519-5526. Organomet. Chem., 2018, 864, 12-18. [60] R. C. McHatton, J. H. Espenson, A. Bakac, J. Am. Chem. Soc., [91] R. Poli, Chem. Eur. J., 2015, 21, 6988-7001. 1982, 104, 3531-3533. [92] R. Poli, Angew. Chem. Int. Ed., 2006, 45, 5058-5070. [61] M. D. Johnson, Acc. Chem. Res., 1983, 16, 343-349. [93] K. M. Smith, W. S. McNeil, A. S. Abd-El-Aziz, Macromol. [62] H. Cohen, D. Meyerstein, Inorg. Chem., 1974, 13, 2434-2443. Chem. Phys., 2010, 211, 10-16. [63] R. J. Blau, J. H. Espenson, A. Bakac, Inorg. Chem., 1984, 23, [94] M. Hurtgen, C. Detrembleur, C. Jerome, A. Debuigne, Polym. 3526-3528. Rev., 2011, 51, 188-213. [64] W. A. Mulac, D. Meyerstein, J. Am. Chem. Soc., 1982, 104, [95] L. E. N. Allan, M. R. Perry, M. P. Shaver, Prog. Polym. Sci., 4124-4128. 2012, 37, 127-156. [65] H. Elroi, D. Meyerstein, J. Am. Chem. Soc., 1978, 100, 5540- [96] R. Poli, “Organometallic Mediated Radical Polymerization”, 5548. in Polymer Science: A Comprehensive Reference (K. Maty- [66] T. S. Roche, J. F.Endicott, Inorg. Chem., 1974, 13, 1575-1580. jaszewski, M. Möller, eds.), vol. 3, Elsevier, 2012, 351-375. [67] A. M. Tait, M. Z. Hoﬀman, E. Hayon, Int. J. Radiat. Phys. Ch., [97] R. Poli, “Organometallic mediated radical polymerization”, 1976, 8, 691-696. in Reference Module in Materials Science and Materials Engi- [68] J. F.Endicott, T. L. Netzel, J. Am. Chem. Soc., 1979, 101, 4000- neering, Elsevier, 2016. 4002. [98] A. Debuigne, C. Jerome, C. Detrembleur, Polymer, 2017, 115, [69] C. Y. Mok, J. F. Endicott, J. Am. Chem. Soc., 1978, 100, 123- 285-307. 129. [99] J. C. Doherty, K. H. D. Ballem, B. O. Patrick, K. M. Smith, [70] R. Van Eldik, H. Cohen, D. Meyerstein, Angew. Chem. Int. Ed., Organometallics, 2004, 23, 1487-1489. 1991, 30, 1158-1160. [100] P.J. Flory, F.S. Leutner, J. Polym. Sci., 1948, 3, 880-890. [71] R. Van Eldik, H. Cohen, D. Meyerstein, Inorg. Chem., 1994, [101] J. Xia, H.-j. Paik, K. Matyjaszewski, Macromolecules, 1999, 32, 33, 1566-1568. 8310-8314. [72] H. H. Schuh, H. Fischer, Helv. Chim. Acta, 1978, 61, 2130- [102] M. Y. Zaremski, A. V. Plutalova, E. S. Garina, M. B. Lachinov, 2164. V. B. Golubev, Macromolecules, 1999, 32, 6359-6362. [73] M. Fantin, F. Lorandi, T. G. Ribelli, C. Fliedel, L. Thevenin, [103] Y. Champouret, U. Baisch, R. Poli, L. Tang, J. L. Conway, K. M. A. A. Isse, R. Poli, K. Matyjaszewski, Macromolecules, 2019, Smith, Angew. Chem. Int. Ed., 2008, 47, 6069-6072. 52, 4079-4090. [104] Y. Champouret, K. C. MacLeod, U. Baisch, B. O. Patrick, K. M. [74] T. J. Zerk, P.V. Bernhardt, Inorg. Chem., 2017, 56, 5784-5792. Smith, R. Poli, Organometallics, 2010, 29, 167-176. [75] T. J. Zerk, L. R. Gahan, E. H. Krenske, P.V. Bernhardt, Polym. [105] A. Debuigne, J. R. Caille, R. Jérôme, Angew. Chem. Int. Ed., Chem., 2019, 10, 1460-1470. 2005, 44, 1101-1104. [76] K. Matyjaszewski, J. H. Xia, Chem. Rev., 2001, 101, 2921-2990. [106] S. Maria, H. Kaneyoshi, K. Matyjaszewski, R. Poli, Chem. Eur. [77] M. Kamigaito, T. Ando, M. Sawamoto, Chem. Rev., 2001, 101, J., 2007, 13, 2480-2492. 3689-3745. [107] K. S. Santhosh Kumar, Y. Gnanou, Y. Champouret, J.-C. [78] T. Andruniow, M. Z. Zgierski, P. M. Kozlowski, J. Am. Chem. Daran, R. Poli, Chem. Eur. J., 2009, 15, 4874-4885. Soc., 2001, 123, 2679-2680. [108] C. H. Peng, J. Scricco, S. Li, M. Fryd, B. B. Wayland, Macro- [79] N. Dölker, F. Maseras, A. Lledós, J. Phys. Chem. B, 2001, 105, molecules, 2008, 41, 2368-2373. 7564-7571. [109] S. Li, B. de Bruin, C. H. Peng, M. Fryd, B. B. Wayland, J. Am. [80] K. P.Jensen, U. Ryde, J. Phys. Chem. A, 2003, 107, 7539-7545. Chem. Soc., 2008, 130, 13373-13381. [81] P. E. M. Siegbahn, M. R. A. Blomberg, S. L. Chen, J. Chem. [110] C. S. Hsu, T. Y. Yang, C. H. Peng, Polym. Chem., 2014, 5, 3867- Theory Comput., 2010, 6, 2040-2044. 3875. [82] J. Kuta, S. Patchkovskii, M. Z. Zgierski, P. M. Kozlowski, J. [111] C.-M. Liao, C.-C. Hsu, F.-S. Wang, B. B. Wayland, C.-H. Peng, Comput. Chem., 2006, 27, 1429-1437. Polym. Chem., 2013, 4, 3098-3104. [83] S. L. Chen, M. R. A. Blomberg, P. E. M. Siegbahn, J. Phys. [112] Y. Zhao, M. Yu, S. Zhang, Z. Wu, Y. Liu, C.-H. Peng, X. Fu, Chem. B, 2011, 115, 4066-4077. Chem. Sci., 2015, 6, 2979-2988. [84] K. P. Jensen, B. O. Roos, U. Ryde, J. Chem. Phys., 2007, 126, [113] Y. G. Zhao, S. L. Zhang, Z. Q. Wu, X. Liu, X. Y. Zhao, C. H. Peng, article no. 014103. X. F.Fu, Macromolecules, 2015, 48, 5132-5139. [85] P. M. Kozlowski, M. Kumar, P. Piecuch, W. Li, N. P. Bauman, [114] Y. Zhao, Y. Wang, X. Zhou, Z. Xue, X. Wang, X. Xie, R. Poli, J. A. Hansen, P.Lodowski, M. Jaworska, J. Chem. Theory Com- Angew. Chem. Int. Ed., 2019, 58, 14311-14318. put., 2012, 8, 1870-1894. [115] E. V. Bellan, L. Thevenin, F. Gayet, C. Fliedel, R. Poli, ACS [86] M. Reiher, Inorg. Chem., 2002, 41, 6928-6935. Macro Lett., 2017, 6, 959-962. [87] J. A. Connor, M. T. Zafaranimoattar, J. Bickerton, N. I. El- [116] A. N. Morin, C. Detrembleur, C. Jérôme, P. D. Tullio, R. Poli, A. Debuigne, Macromolecules, 2013, 46, 4303-4312.

C. R. Chimie — 2021, 24, n 1, 147-175 Rinaldo Poli 175

[117] A. Debuigne, Y. Champouret, R. Jérôme, R. Poli, C. Detrem- [145] W. A. Braunecker, W. C. Brown, B. Morelli, W. Tang, R. Poli, bleur, Chem. Eur. J., 2008, 14, 4046-4059. K. Matyjaszewski, Macromolecules, 2007, 40, 8576-8585. [118] A. Debuigne, C. Michaux, C. Jérôme, R. Jérôme, R. Poli, [146] J.-S. Wang, K. Matyjaszewski, J. Am. Chem. Soc., 1995, 117, C. Detrembleur, Chem. Eur. J., 2008, 14, 7623-7637. 5614-5615. [119] A. Debuigne, A. N. Morin, A. Kermagoret, Y. Piette, C. De- [147] T. G. Ribelli, K. Matyjaszewski, R. Poli, J. Coord. Chem., 2018, trembleur, C. Jérôme, R. Poli, Chem. Eur. J., 2012, 18, 12834- 71, 1641-1668. 12844. [148] K. Matyjaszewski, B. E. Woodworth, Macromolecules, 1998, [120] Y. Piette, A. Debuigne, C. Jérôme, V. Bodart, R. Poli, C. De- 31, 4718-4723. trembleur, Polym. Chem., 2012, 3, 2880-2891. [149] A. Kaur, T. G. Ribelli, K. Schroeder, K. Matyjaszewski, T. Pin- [121] B. Améduri, Chem. Rev., 2009, 109, 6632-6686. tauer, Inorg. Chem., 2015, 54, 1474-1486. [122] C. Boyer, D. Valade, L. Sauguet, B. Améduri, B. Boutevin, [150] T. G. Ribelli, M. Fantin, J.-C. Daran, K. F. Augustine, R. Poli, Macromolecules, 2005, 38, 10353-10362. K. Matyjaszewski, J. Am. Chem. Soc., 2018, 140, 1525-1534. [123] A. D. Asandei, O. I. Adebolu, C. P.Simpson, J. Am. Chem. Soc., [151] H. Schroeder, M. Buback, Macromolecules, 2014, 47, 6645- 2012, 134, 6080-6083. 6651. [124] A. D. Asandei, O. I. Adebolu, C. P.Simpson, J.-S. Kim, Angew. [152] T. G. Ribelli, S. M. W. Rahaman, J.-C. Daran, P.Krys, K. Maty- Chem. Int. Ed., 2013, 52, 10027-10030. jaszewski, R. Poli, Macromolecules, 2016, 49, 7749-7757. [125] A. D. Asandei, Chem. Rev., 2016, 116, 2244-2274. [153] T. G. Ribelli, K. F. Augustine, M. Fantin, P. Krys, R. Poli, [126] M. Guerre, B. Campagne, O. Gimello, K. Parra, B. Améduri, K. Matyjaszewski, Macromolecules, 2017, 50, 7920-7929. V. Ladmiral, Macromolecules, 2015, 48, 7810-7822. [154] L. Thevenin, C. Fliedel, M. Fantin, T. G. Ribelli, K. Maty- [127] M. Guerre, S. M. W. Rahaman, B. Améduri, R. Poli, V. Ladmi- jaszewski, R. Poli, Inorg. Chem., 2019, 58, 6445-6457. ral, Macromolecules, 2016, 49, 5386-5396. [155] L. Thevenin, C. Fliedel, K. Matyjaszewski, R. Poli, Eur. J. In- [128] M. Guerre, S. M. W. Rahaman, B. Améduri, R. Poli, V. Ladmi- org. Chem., 2019, 4489-4499. ral, Polym. Chem., 2016, 7, 6918-6933. [156] T. Parchomyk, K. Koszinowski, Synthesis-Stuttgart, 2017, 49, [129] S. Banerjee, V. Ladmiral, A. Debuigne, C. Detrembleur, 3269-3280. R. Poli, B. Améduri, Angew. Chem. Int. Ed., 2018, 57, 2934- [157] G. Cahiez, A. Moyeux, Chem. Rev., 2010, 110, 1435-1462. 2937. [158] J. B. Diccianni, T. Diao, Trends Chem., 2019, 1, 830-844. [130] P. G. Falireas, V. Ladmiral, A. Debuigne, C. Detrembleur, [159] C. Sambiagio, S. P. Marsden, A. J. Blacker, P. C. McGowan, R. Poli, B. Améduri, Macromolecules, 2019, 52, 1266-1276. Chem. Soc. Rev., 2014, 43, 3525-3550. [131] E. Le Grognec, J. Claverie, R. Poli, J. Am. Chem. Soc., 2001, [160] T.Hatakeyama, T.Hashimoto, Y. Kondo, Y. Fujiwara, H. Seike, 123, 9513-9524. H. Takaya, Y. Tamada, T. Ono, M. Nakamura, J. Am. Chem. [132] F.Stoffelbach, J. Claverie, R. Poli, C. R. Acad. Sci. Paris C, 2002, Soc., 2010, 132, 10674-10676. 5, 37-42. [161] S. L. Daifuku, M. H. Al-Afyouni, B. E. R. Snyder, J. L. Knee- [133] F. Stoffelbach, R. Poli, P. Richard, J. Organometal. Chem., bone, M. L. Neidig, J. Am. Chem. Soc., 2014, 136, 9132-9143. 2002, 663, 269-276. [162] J. A. Przyojski, K. P. Veggeberg, H. D. Arman, Z. J. Tonzetich, [134] F. Stoffelbach, R. Poli, S. Maria, P. Richard, J. Organomet. ACS Catal., 2015, 5, 5938-5946. Chem., 2007, 692, 3133-3143. [163] Y. S. Liu, J. Xiao, L. Wang, Y. Song, L. Deng, Organometallics, [135] A. A. Gridnev, S. D. Ittel, Chem. Rev., 2001, 101, 3611-3659. 2015, 34, 599-605. [136] M. P.Shaver, L. E. N. Allan, H. S. Rzepa, V. C. Gibson, Angew. [164] N. Shaham, A. Masarwa, Y. Matana, H. Cohen, D. Meyerstein, Chem. Int. Ed., 2006, 45, 1241-1244. Eur. J. Inorg. Chem., 2002, 87-92. [137] L. E. N. Allan, M. P.Shaver, A. J. P.White, V. C. Gibson, Inorg. [165] C. Ducker-Benfer, M. S. A. Hamza, C. Eckhardt, R. van Eldik, Chem., 2007, 46, 8963-8970. Eur. J. Inorg. Chem., 2000, 1563-1569. [138] R. K. O’Reilly, M. P.Shaver, V. C. Gibson, A. J. P.White, Macro- [166] H. M. Feder, J. Halpern, J. Am. Chem. Soc., 1975, 97, 7186- molecules, 2007, 40, 7441-7452. 7188. [139] M. P. Shaver, L. E. N. Allan, V. C. Gibson, Organometallics, [167] W. Lau, J. C. Huffman, J. K. Kochi, Organometallics, 1982, 1, 2007, 26, 4725-4730. 155-169. [140] L. E. N. Allan, J. P. MacDonald, A. M. Reckling, C. M. Kozak, [168] G. Pattenden, Chem. Soc. Rev., 1988, 17, 361-382. M. P.Shaver, Macromol. Rapid Commun., 2012, 33, 414-418. [169] J. C. Lo, Y. Yabe, P. S. Baran, J. Am. Chem. Soc., 2014, 136, [141] L. E. N. Allan, J. P. MacDonald, G. S. Nichol, M. P. Shaver, 1304-1307. Macromolecules, 2014, 47, 1249-1257. [170] J. C. Lo, D. Kim, C. M. Pan, J. T. Edwards, Y. Yabe, J. H. Gui, [142] R. Poli, M. P.Shaver, Chem. Eur. J., 2014, 20, 17530-17540. T. Qin, S. Gutierrez, J. Giacoboni, M. W. Smith, P.L. Holland, [143] S. W. M. Crossley, C. Obradors, R. M. Martinez, R. A. Shenvi, P.S. Baran, J. Am. Chem. Soc., 2017, 139, 2484-2503. Chem. Rev., 2016, 116, 8912-9000. [171] D. Kim, S. M. W. Rahaman, B. Q. Mercado, R. Poli, P. L. [144] W. A. Braunecker, Y. Itami, K. Matyjaszewski, Macro- Holland, J. Am. Chem. Soc., 2019, 141, 7473-7485. molecules, 2005, 38, 9402-9404. [172] D. C. Eisenberg, J. R. Norton, Isr. J. Chem., 1991, 31, 55-66. [173] D. Wang, R. J. Angelici, J. Am. Chem. Soc., 1996, 118, 935-942. [174] C. Fliedel, R. Poli, J. Organomet. Chem., 2019, 880, 241-252.

C. R. Chimie — 2021, 24, n 1, 147-175