Is the Electrophilicity of the Metal Nitrene the Sole Predictor of Metal-Mediated Nitrene Transfer to Olefins? Secondary Contributing Factors as Revealed by a Library of High-Spin Co(II) Reagents

Abstract

Recent research has highlighted the key role played by the electron affinity of the active metal-nitrene/imido oxidant as the driving force in nitrene additions to olefins to afford valuable aziridines. The present work showcases a library of Co(II) reagents that, unlike the previously examined Mn(II) and Fe(II) analogues, demonstrate reactivity trends in olefin aziridinations that cannot be solely explained by the electron affinity criterion. A family of Co(II) catalysts (17 members) has been synthesized with the assistance of a trisphenylamido-amine scaffold decorated by various alkyl, aryl, and acyl groups attached to the equatorial amidos. Single-crystal X-ray diffraction analysis, cyclic voltammetry and EPR data reveal that the high-spin Co(II) sites (S = 3/2) feature a minimal [N₃N] coordination and span a range of 1.4 V in redox potentials. Surprisingly, the Co(II)-mediated aziridination of styrene demonstrates reactivity patterns that deviate from those anticipated by the relevant electrophilicities of the putative metal nitrenes. The representative L⁴Co catalyst (−COCMe₃ arm) is operating faster than the L⁸Co analogue (−COCF₃ arm), in spite of diminished metal-nitrene electrophilicity. Mechanistic data (Hammett plots, KIE, stereocontrol studies) reveal that although both reagents follow a two-step reactivity path (turnover-limiting metal-nitrene addition to the C_b atom of styrene, followed by product-determining ring-closure), the L⁴Co catalyst is associated with lower energy barriers in both steps. DFT calculations indicate that the putative [L⁴Co]NTs and [L⁸Co]NTs species are electronically distinct, inasmuch as the former exhibits a single-electron oxidized ligand arm. In addition, DFT calculations suggest that including London dispersion corrections for L⁴Co (due to the polarizability of the tert-Bu substituent) can provide significant stabilization of the turnover-limiting transition state. This study highlights how small ligand modifications can generate stereoelectronic variants that in certain cases are even capable of overriding the preponderance of the metal-nitrene electrophilicity as a driving force.

Graphical Abstract

INTRODUCTION

The role of aziridines¹ as intermediates and end products of synthetic and biological chemistry is hard to overstate. Not only do aziridines afford avenues for further structural development by taking advantage of the energetic content and stereochemical disposition of their strained three-atom ring (ring opening, expansion, or rearrangement),² but also they constitute valuable functionalities in the framework of several natural products possessing antibiotic or antineoplastic activities.³ In addition to the central role exercised by aziridines as fine chemicals and pharmaceutical agents,⁴ their contribution to the chemistry of materials has been increasingly recognized,⁵ especially as key entities for the development and postmodification of polymeric scaffolds.

Synthetic protocols for the generation of aziridines abound, but largely rely on three major methodologies. The cyclization of 1,2-amino precursors constitutes a traditional approach that has been more recently complemented by addition of either C₁ sources to imines or electrophilic N₁ donors to alkenes.⁶ The latter “C₂ + N₁” addition is extensively implemented due to its operational simplicity and availability of a wide range of suitable substrates and catalysts. The N₁ donors encompass a variety of nitrene/nitrenoid precursor oxidants such as iminoiodanes (ArI=NR),⁷ haloamines (RNNaX, X = Cl, Br),⁸ O/N-substituted hydroxylamines and N-tosyloxycarbamates (RN-(X)–OR′, X = H, leaving group)⁹ or atom-economical organic azides (RN₃).¹⁰ As opposed to oxo-transfer chemistry, the corresponding nitrene/nitrenoid transfer relies significantly on the choice of the attendant R group to control the electrophilicity of the active moiety and provide activated (R = SO₂R, CO₂R, COR, carbamoyl, sulfamoyl) or nonactivated aziridines (R = H, alkyl, aryl, silyl) with differential reactivity.¹¹

A wide range of catalysts has been explored to influence reactivity and selectivity outcomes in nitrene transfer to alkenes, including several organocatalytic¹² and metal-mediated processes.¹³ In the latter case, the presumptive and rather elusive metal-nitrene (M=NR) active species are entities with rich and variable stereoelectronic attributes, inherent and/or ligand induced, whose operation vis-à-vis olefinic substrates is a matter of intense investigation. The variety of transition metals employed, both from the first-row (Mn, Fe, Co, Ni, Cu)^14–22 and from the heavier platinum-group^23–25 and coinage elements,^26,27 coupled with a range of ancillary ligand frameworks (e.g., porphyrinoids, salens, bis-oxazolines, tetracarboxylate paddlewheels, trispyrazolyl-borates/methanes, polypyridines) is a testament to the vigorous activity in this field and that of the closely related C–H bond amination reactions.²⁸

Among the late 3d transition elements, the case of cobalt is most intriguing, inasmuch as isolable or even putative Co=NR units have been invoked with a variety of oxidation states (from II to V), electronic ground-state spins (S = 0, 1/2, 1, 3/2, 2), and coordination numbers (from 2 to 5).²⁹ The most common configuration is that of diamagnetic Co(III) imidos (S = 0),³⁰ mostly supported by C₃ or C₂ symmetric ligands. In a handful of cases, open-shell spin-states were observed for Co(III) imidos, as for instance with (trispyrazolylborato)Co^III(NAd) (S = 1, at T > 280 K),³¹ (dipyrrin)Co^III(NR) (S = 1 for R = Mes; S = 0 or 0 → 2 transition, for R = ^tBu, 1-Ad, other alkyls),³² [(hmds)₂Co^III(N^tBu)]⁻ (S = 1; hmds = N(SiMe₃)₂),³³ and possibly bimetallic Zr(μ-NMes)Co^III(NMes) (S = 0 → 2 transition, near room temperature).³⁴ None of these compounds have been reported to mediate nitrene-transfer to alkenes. Observable reactivity includes (i) nitrene-transfer to carbon monoxide;^30g,31a,35 (ii) insertion of nitrene into ligand-derived carbene residues;^30e (iii) formal hydrogen-atom abstraction from a ^tBu or Mes ligand moiety by open-shell Co=NR, presumably generating an amido Co–NHR unit and a carbon centered radical; the latter can then recombine with the amido,^31a dimerize,^31b or generate a Co–C bond;^31b,32a (iv) intramolecular C–H bond insertion into alkyl azides (source of imido), mediated by (dipyrrin)Co^III(NR), to generate substituted N-heterocycles;^32b,c and (v) a rare instance of intermolecular hydrogen-atom abstraction from C–H bonds of various substrates with BDE_C–H ≤ 92 kcal mol⁻¹ by [(hmds)₂Co^III(N^tBu)]⁻,³³ leading to the corresponding Co(II) amido; the amido can then react with another equivalent of substrate (C–H) to perform either proton transfer (frequently with the concomitant formation of Co^II–C organometallics) or formal hydrogen-atom abstraction via stepwise proton/electron transfer or direct HAT, giving rise to Co(I) and substrate dehydrogenation product. In several instances noted above, the carbophilic character of cobalt is notable as a product-determining factor.

Cobalt(II) imidos are more recent additions to the repertoire of cobalt reagents, and encompass both high-spin (S = 3/2)³⁶ and low-spin (S = 1/2)³⁷ cases as two- and four-coordinate compounds, respectively. The high-spin examples have been reported to perform nitrene-transfer to ethylene to afford RN=CH—CH₃, presumably due to a [2π + 2π] activation mode. Similarly, certain C(sp)–H and Si–H bonds are activated not via H atom abstraction, but by means of [2π + 2σ] interactions.^36b On the other hand, the low-spin Co(II) imidos are unreactive versus alkenes, although they engage in nitrene-transfer and/or nitrene-exchange with O/S with respect to substrates such as CO, PMe₃, PhCHO, and CS₂.³⁷ Finally, two examples of high-valent Co(IV) and Co(V) bis-imido complexes ([IMes]Co(NDipp)₂]^0/+), possessing low-spin ground states of S = 1/2 and 0, respectively, proved to be rather unreactive.³⁸ The open-shell Co(IV) congener is the only one that exhibits intramolecular nitrene C–H insertion into the o-Me group of the Mes residue, possibly via an ortho-cobaltation intermediate (Co–C).²⁹

Whereas the catalytic formation of new C–N bonds by means of the isolable cobalt imidos noted above is only rarely observed, the advent of a library of Co^II(Por) complexes that give rise to Co^III–nitrenoid radicals [(Por)Co^III–^•NR] or [(Por^•)Co^III–(^•NR)₂] has provided numerous instances of highly effective catalytic systems for the stereo-, chemo-, and site-selective aziridination of alkenes and amination of C–H bonds.¹⁹ Starting with Co(TPP), and electron-deficient analogues, several generations of Co^II(Por) reagents with richly decorated porphyrins have been introduced in the past two decades to facilitate the activation of various organic azides, leading to the generation of well characterized low-spin (S = 1/2) Co^III–^•NR moieties, with spin density largely localized on the N atom.^19f,39 These relatively long-lived Co^III–nitrene-radical intermediates owe their stability to hydrogen-bonding interactions of the nitrene moieties with porphyrin-appended amido residues (−NHCOR*), which can further introduce and metal-orient chiral auxiliaries via their R* functionality in D₂-symmetric overall geometries. Detailed theoretical and experimental studies^19f,39 have established that the mode of operation of [(Por)Co^III–^•NR] metalloradicals vis-à-vis C=C or C–H bonds consists of a two-step process: initial formation of a new N–C bond with alkenes and relocation of the spin density on the distal carbon atom (Co^III–N(R)–C–^•C–) (or formation of a Co^III–NHR amido and a substrate-bound radical via hydrogen-atom abstraction from a C–H bond), followed by an essentially barrierless collapse of the carbon-centered radical with the N atom to generate the product of aziridination (or amination) along with Co^II(Por).

More recently, the structurally related [Co^III(TAML^red)]⁻ and [Co^III(TAML^sq)] compounds, featuring the tetraamido macrocyclic ligand TAML in its intact reduced form TAML^red and one-electron oxidized variant TAML^sq (sometimes denoted as TAML^+•), have been shown to give rise to [Co^III(TAML^q)(^•NR)₂]⁻ (S = 1) and [Co^III(TAML^q)(^•NR)] (S = 1/2), respectively (TAML^q = doubly oxidized, diamagnetic ligand; Co^III site is low-spin, S = 0).⁴⁰ These cobalt nitrenes have emerged as capable catalysts for the aziridination of largely styrene substrates by imidoiodinanes (PhINNs, PhINTs, PhINTces).⁴¹ Their mode of operation is considered to be unique, inasmuch as the turnover-limiting, initial N–C bond formation with styrenes features an asynchronous transition state, encompassing a partial electron-transfer to form a styrenyl radical cation, in turn undergoing a nucleophilic attack by the nitrene lone pair (see below for more details). This initial charge-transfer has also emerged as a central component of the operation of iron(IV) imido species developed by Latour and co-workers for alkene aziridinations, further underscoring the importance of the electron affinity of the metal nitrene as a commonly encountered driving force.⁴²

Finally, of great interest are recently reported Co(II) organoazides,⁴³ which either thermally or photolytically can extrude N₂ to give rise to nitrenoids best described as iminyl [Co^III–^•NR] units (R = aryl, alkyl). Although the electronics of these fleeting intermediates are not yet known, crystal structures of such species (R = alkyl) have been determined following N₂ expulsion from single crystals of metal azides in the solid state. For R = aryl, the cobalt(II) organoazide can promote nitrene-transfer by means of unisolable [Co^III–^•NR], both intramolecularly ([3 + 2] annulation) and intermolecularly (C–H allylic amination or styrene aziridination in modest yields). The reactivity of the Co(II) aliphatic azides is more complex and includes (i) α-H atom abstraction via the incipient [Co^III–^•NCH₂R] to generate the imine (RCH=NH), if strong δ-C–H bonds (sec, prim) are present; (ii) δ-H atom abstraction and amination of relatively weaker δ-C–H bonds (benzylic, tertiary) by the cobalt alkyl azide itself (initial N₂ extrusion is not needed), leading to substituted pyrrolidines; and (iii) intramolecular 1,3-dipolar cycloaddition of cobalt-bound CH₂=CH(CH₂)₄N₃ to afford 1,2,3-dihydrotriazole.

In the present work, we examine a library of high-spin Co(II) reagents (S = 3/2), supported by a modular trisphenylamido-amine ligand framework, giving rise to a weak equatorial field. Previous DFT calculations⁴⁴ on one member of this library of reagents indicated that exposure to a nitrene source (PhI = NTs) generates a Co(III)–nitrene radical (Co^III–^•NTs) with a high-spin ground state (S = 5/2). The corresponding doublet and quartet [Co]NTs states lie slightly higher than the sextet by ΔG values of 0.4 and 1.5 kcal mol⁻¹, respectively. The computed spin density for the S = 5/2 state places ~1.1 unpaired electron on the nitrene N atom, and ~3.3 unpaired electrons on Co, with the remaining spin density being distributed to other atoms. Similarly to the low-spin porphyrin-supported Co^III–nitrene-radical (S = 1/2) noted above, the high-spin congener is capable of performing alkene aziridinations in a two-step process (successive formation of two N–C bonds). Remarkably, the computed transition-state barrier (ΔG^‡ = 23.4 kcal mol⁻¹ vs Co^III–^•NTs/styrene) for the rate-determining, initial C_β…NTs bond-forming step of the high-spin system is very similar to that reported for the corresponding low-spin Co(Por) (ΔG^‡ = 24.1 kcal mol⁻¹) or Co(AmidoPor) (ΔG^‡ = 22.8 kcal mol⁻¹) with respect to Co^III–^•NSO₂Ph/styrene.^19f The present work significantly enlarges the scope of high-spin Co(II) compounds as nitrene-transfer reagents, and provides insights in their operational characteristics, not only vis-à-vis the reported low-spin (Por)Co(II) paradigms, but also in comparison with the previously examined libraries of Mn(II) and Fe(II) reagents, supported by the same trisphenylamido-amine ligand framework.⁴⁴ Whereas the nitrene-transfer reactivity of the Mn(II) reagents in alkene aziridinations largely depends on the electrophilicity of the presumptive Mn^III–^•NR (S = 3/2) moiety, underscoring the role of the electron affinity of the metal nitrene as a dominant factor, the reactivity of the corresponding, more reactive, Co(II) reagents is affected by additional subtle electronic and steric factors. These most likely arise from the tighter disposition of the reaction cavity, resulting in ligand-coordination flexibility, electronic rearrangement, and secondary stabilizing interactions. In this publication we show that even an otherwise small change in ligand substitution can have a significant effect on nitrene-transfer reactivity in aziridination reactions, occasionally overriding the preponderance of the metal-nitrene electrophilicity as a driving force.

RESULTS AND DISCUSSION

Synthesis and Characterization of New Ligands and Co(II) Complexes.

The family of trisphenylamido-amine ligands (L¹H₃–L¹⁷H₃) employed in this study is shown in Figure 1. The majority of these ligands (L¹H₃–L¹⁵H₃) have been used and reported in previous studies.^44–48 They are all derivatives of the common 2,2′,2″-triaminotriphenylamine framework,⁴⁵ featuring carbonaceous arm substituents (alkyl, aryl acyl). Ligand L¹⁶H₃ is prepared by methylation of deprotonated (KH) 2,2′,2″-triaminotriphenylamine by MeI in THF, and ligand L¹⁷H₃ is derived via condensation of the same triamine with the corresponding chiral acyl chloride in the presence of Et₃N in dichloromethane. The solid-state structures of these two new ligands (Figure S1) are indicative of their favorable preorganization for metalation, in a cavity that is buttressed by alkyl and acyl arms, respectively.

Figure 1.

Ligands employed in the present study.

Co^II complexes were synthesized with all ligands, by reacting the deprotonated (KH) ligand with anhydrous beads of CoCl₂ in THF (alkyl and aryl armed ligands) or N,N-dimethylacetamide (DMA) (acyl armed ligands). A subset of Co^II compounds, namely L³Co (3), L⁵Co (5), L⁸Co (8a), L⁹Co (9), L¹⁰Co (10), and L¹³Co (13), has been previously reported in a study that examined the use of these catalysts in controlled radical polymerization of olefins.⁴⁸ In addition, L⁸Co (8b) has been explored in conjunction with the L⁸ Mn and L⁸Fe congeners, toward establishing metal-dependent trends in catalytic nitrene transfer to olefins.⁴⁴ Figure 2 depicts representations of the minimal coordination site of each Co^II site, derived from single-crystal X-ray diffraction data. In all cases the ligand coordinates in a trigonal pyramidal [N_3(amido)N_amine] mode, exhibiting various degrees of distortion, although in two instances (7, 17) the coordination to the axial N_amine residue can be best described as a long contact. Moreover, compound 11 features a noncoordinating amido residue that has been protonated. The dominant four-coordinate [N₃N] pattern is retained as the sole ligand field of seven Co^II compounds (4, 5, 8a, 8c, 9, 10, 12). Additional elements of metal coordination, essentially located trans to the axial N_amine residue, are observed with all other compounds, and include solvent moieties, especially for compounds crystallized from MeCN (3, 6, 13) and THF (16), as well as carbonyl units (−C(R)=O–Co^II) deriving from acyl residues belonging to the ligand (7, 14, 17). In a single case (15), a six-coordinate Co^II site arises from the presence of two ligand-derived ether residues in the metal coordination sphere, in addition to the usual [N₃N] framework. Importantly, most structures are polymeric, largely due to an intricate network of intermolecular interactions generated by K⁺ ions. Mononuclear (1, 2, 5, 12, 13) or oligonuclear (8c, 11, 17) compounds (molecular or ionic) are only encountered in a handful of cases. A more detailed description of the structural features of the new Co^II compounds is provided below.

Figure 2.

Minimal coordination of Co^II compounds with ligands L¹–L¹⁷ explored in this study.

Co^II Compounds with Acyl-Armed Ligands.

The seven new compounds (4, 6, 7, 8c, 14, 15, 17; Figure 3) that belong in this category are polymeric, with the exception of [K(DMA)₃(L⁸)Co^II]₂ (8c) and [K₂(THF)₂K₂(L¹⁷)₄Co^II₄] (17) that feature a dimeric and tetrameric molecular unit, respectively. The catalytically important {[K(L⁴)Co^II]·Diethyl Ether}_n (4) exhibits higher symmetry than all other compounds, consisting of a rigorous 3-fold axis along the Co–N_amine direction as well as through K⁺ ions relating three different molecules in the crystal lattice. This compound is characterized by an exclusive four-coordinate [N₃N] ligand field and an open metalated cavity fortified by the three −CO^tBu arms. The carbonyl residues are positioned exo with respect to the cavity and are further engaged in contacts with K⁺ ions, inasmuch as each potassium is coordinated by three oxygen (carbonyl) atoms belonging to different molecules, and is also involved in K⁺–arene π contacts. These structural features are largely retained in the structure of [K(NCMe)(L⁶)Co^II–NCMe]·2MeCN (6), but important deviations also apply, mostly because of the presence of a coordinated MeCN molecule in a trans position versus the N_amine (N_amine–Co–N_MeCN = 178.75(14)°) and the lack of a strict 3-fold crystallographic symmetry. Otherwise, each K⁺ ion is still coordinated by three carbonyl residues belonging to different molecules in addition to a single MeCN, in lieu of any K⁺–arene contacts.

Figure 3.

ORTEP diagrams (from left to right) of {[K(L⁴)Co^II]·Diethyl Ether}_n (4), [K(MeCN)(L⁶)Co^II–NCMe]·2MeCN (6), [K₂(DMA)₄]-[[K(L⁷)₂Co^II₂]₂·2DMA (7), [K₂(DMA)₃(L⁸)Co^II]₂ (8c), [K(DMA)(L¹⁴)Co^II]·DMA (14), [K(THF)₃(L¹⁵)Co^II]·THF (15), and [K(THF)K(L¹⁷)₂Co^II₂]₂·3Pentane (17), drawn with 40% thermal ellipsoids. Hydrogen atoms are omitted for clarity. Selective interatomic distances (Å) and angles (deg): 4, Co(1)–N(1) = 2.151(11), Co(1)–N(2) = 1.986(6), Co(1)–[N(2), N(2), N(2)] = 0.32(2) (distance of Co from mean plane) Å, N(2)–Co(1)–N(2) = 117.48(11), N(2)–Co(1)–N(1) = 80.8(2); 6, Co(1)–N(1) = 2.235(3), Co(1)–N(2) = 2.046(4), Co(1)–N(3) = 2.049(4), Co(1)–N(4) = 2.053(4), Co(1)–N(5) = 2.063(4), Co(1)–[N(2), N(3), N(4)] = 0.48(2) (distance of Co from mean plane), N(2)–Co(1)–N(4) = 115.18(15), N(2)–Co(1)–N(3) = 117.17(15), N(3)–Co(1)–N(4) = 111.69(15), N(2)–Co(1)–N(1) = 76.70(14), N(4)–Co(1)–N(1) = 76.49(14), N(3)–Co(1)–N(1) = 76.15(14), N(3)–Co(1)–N(5) = 103.24(15), N(2)–Co(1)–N(5) = 104.55(14), N(4)–Co(1)–N(5) = 102.81(15), N(1)–Co(1)–N(5) = 178.75(14); 7, Co(1)–N(1) = 2.463(4), Co(1)–N(2) = 2.053(4), Co(1)–N(3) = 2.054(5), Co(1)–N(4) = 2.054(4), Co(1)–O(6) = 1.975(4), Co(1)–[N(2), N(3), N(4)] = 0.61(2) (distance of Co from mean plane), Co(2)–N(5) = 2.517(5), Co(2)–N(6) = 2.081(5), Co(2)–N(7) = 2.033(5), Co(2)–N(8) = 2.069(5), Co(2)–O(3) = 1.984(4), Co(2)–[N(6), N(7), N(8)] = 0.67(2) (distance of Co from mean plane), N(2)–Co(1)–N(4) = 109.51(18), N(2)–Co(1)–N(3) = 113.66(17), N(3)–Co(1)–N(4) = 111.52(18), N(2)–Co(1)– N(1) = 73.14(16), N(4)–Co(1)–N(1) = 73.24(16), N(3)–Co(1)–N(1) = 71.83(16), N(3)–Co(1)–O(6) = 103.19(17), N(2)–Co(1)–O(6) = 106.85(17), N(4)–Co(1)–O(6) = 111.92(16), N(1)–Co(1)–O(6) = 174.16(15), N(6)–Co(2)–N(7) = 111.14(19), N(6)–Co(2)–N(8) = 112.01(19), N(7)–Co(2)–N(8) = 106.34(19), N(6)–Co(2)–N(5) = 70.48(19), N(7)–Co(2)–N(5) = 70.69(16), N(8)–Co(2)–N(5) = 71.42(17), N(6)–Co(2)–O(3) = 105.77(19), N(7)–Co(2)–O(3) = 105.15(18), N(8)–Co(2)–O(3) = 116.27(17), N(5)–Co(2)–O(3) = 172.25(16); 8c, Co(1)–N(1) = 2.1422(13), Co(1)–N(2) = 1.9941(14), Co(1)–N(3) = 1.9899(14), Co(1)–N(4) = 1.9872(13), Co(1)–[N(2), N(3), N(4)] = 0.306(4) (distance of Co from mean plane), K(1)–O(1) = 2.7298(14), N(2)–Co(1)–N(4) = 114.79(6), N(2)–Co(1)–N(3) = 118.04(6), N(3)–Co(1)–N(4) = 120.18(6), N(2)–Co(1)–N(1) = 80.83(5), N(4)–Co(1)–N(1) = 81.54(5), N(3)–Co(1)–N(1) = 81.06(5); 14, Co(1)–N(1) = 2.224(3), Co(1)–N(2) = 2.018(4), Co(1)–N(3) = 2.031(4), Co(1)–N(4) = 2.021(4), Co(1)–O(3) = 2.162(3), Co(1)–[N(2), N(3), N(4)] = 0.42(1) (distance of Co from mean plane), N(2)–Co(1)–N(4) = 110.16(14), N(2)–Co(1)–N(3) = 121.11(15), N(3)–Co(1)–N(4) = 115.82(15), N(2)–Co(1)–N(1) = 76.31(14), N(4)–Co(1)–N(1) = 78.82(13), N(3)–Co(1)–N(1) = 78.58(14), N(3)–Co(1)–O(3) = 106.07(14), N(2)–Co(1)–O(3) = 91.35(13), N(4)–Co(1)–O(3) = 108.68(13), N(1)–Co(1)–O(3) = 167.33(12); 15, Co(1)–N(1) = 2.337(17), Co(1)–N(2) = 2.041(17), Co(1)–N(3) = 2.048(18), Co(1)–N(4) = 2.033(17), Co(1)–O(4) = 2.405(13), Co(1)–O(6) = 2.302(14), Co(1)–[N(2), N(3), N(4)] = 0.53(2) (distance of Co from mean plane), N(2)–Co(1)–N(4) = 103.6(7), N(2)–Co(1)–N(3) = 118.1(7), N(3)–Co(1)–N(4) = 117.9(7), N(2)–Co(1)–N(1) = 74.3(6), N(4)–Co(1)–N(1) = 76.4(7), N(3)–Co(1)–N(1) = 73.5(6), N(3)–Co(1)–O(6) = 107.6(6), N(2)–Co(1)–O(6) = 127.7(6), N(4)–Co(1)–O(6) = 73.7(6), N(1)–Co(1)–O(6) = 146.4(6), N(3)–Co(1)–O(4) = 72.9(6), N(2)–Co(1)–O(4) = 89.5(6), N(4)–Co(1)–O(4) = 154.0(6), N(1)–Co(1)–O(4) = 129.3(6), O(4)–Co(1)–O(6) = 80.5(5); 17, Co(1)–N(1) = 2.428(5), Co(1)–N(2) = 2.055(5), Co(1)–N(3) = 2.038(5), Co(1)–N(4) = 2.047(5), Co(1)–O(6) = 1.990(4), Co(1)–[N(2), N(3), N(4)] = 0.58(1) (distance of Co from mean plane), Co(2)–N(5) = 2.375(5), Co(2)–N(6) = 2.049(5), Co(2)–N(7) = 2.026(5), Co(2)–N(8) = 2.100(5), Co(2)–O(3) = 2.027(4), Co(2)–[N(6), N(7), N(8)] = 0.56(1) (distance of Co from mean plane), N(2)–Co(1)–N(4) = 110.2(2), N(2)–Co(1)– N(3) = 118.9(2), N(3)–Co(1)–N(4) = 107.2(2), N(2)–Co(1)–N(1) = 73.94(18), N(4)–Co(1)–N(1) = 74.25(18), N(3)–Co(1)–N(1) = 72.09(18), N(3)–Co(1)–O(6) = 109.47(18), N(2)–Co(1)–O(6) = 99.64(18), N(4)–Co(1)–O(6) = 110.20(18), N(1)–Co(1)–O(6) = 172.96(17), N(6)–Co(2)–N(7) = 112.3(2), N(6)–Co(2)–N(8) = 113.4(2), N(7)–Co(2)–N(8) = 112.71(19), N(6)–Co(2)–N(5) = 74.71(18), N(7)–Co(2)–N(5) = 74.52(19), N(8)–Co(2)–N(5) = 73.18(17), N(6)–Co(2)–O(3) = 97.07(18), N(7)–Co(2)–O(3) = 112.18(19), N(8)–Co(2)–O(3) = 108.06(17), N(5)–Co(2)–O(3) = 171.18(17).

The structure of [K₂(DMA)₄][K(L⁷)₂Co^II₂]₂·2DMA (7) is organized in a much more complex manner, featuring a “one-dimensional” array of a repeating unit, [Co(1)/Co(2)–K(1)–Co(3)/Co(4)–K(2)]_n, connected to identical arrays via lateral links provided by a DMA solvated K(3)/K(4) dimer (K₂(DMA)₄). Within the repeating unit, the Co(II) sites are arranged in two similar dimers linked via K⁺ contacts. The dimeric unit is composed of two slightly different Co(II) centers, each featuring the usual [N₃N] ligand coordination, but with a long axial Co–N_amine interaction (av. 2.49 Å). In addition, each Co(II) is coordinated by an oxygen atom (carbonyl) positioned trans with respect to the axial Co–N_amine direction (av. N_amine–Co–O_C=O = 173.2°, Co–O_C=O = 1.98 Å). Importantly, the oxygen atom (carbonyl) coordinated to each Co^II center belongs to the ligand surrounding the partner Co^II site, hence giving rise to a dimer. The K⁺ ions interconnecting the dimers in a pseudo 1-D array are coordinated by two carbonyl and, more weakly, two N_amido residues all belonging to one dimer, and by only a single carbonyl moiety (O(10)) belonging to the adjacent dimer. A much more simplified version of this structure is adopted by [K(DMA)₃(L⁸)Co^II]₂ (8c), exhibiting a dimeric structure comprised of two inversion symmetry related [N₃N]Co^II units connected via carbonyl O atoms of acyl residues to a central K₂(μ-DMA)₂(DMA)₄ core.

Compound [K(DMA)(L¹⁴)Co^II]·DMA (14) retains the usual [N₃N] trigonal-pyramidal coordination but exhibits an additional unique feature, inasmuch as one of the chiral arms generates a seven-member loop by positioning the ester carbonyl in the coordination sphere of Co^II, trans to the axial N_amine residue (N_amine–Co–O_C=O = 167.33(12)°, Co–O_C=O = 2.162(3) Å). The polymeric nature of the compound arises again due to identical K⁺ ions, coordinated by one DMA, forming contacts with oxygen residues (amidato carbonyls, MeC(=O)O–) belonging to three different Co^II sites in the crystal lattice. Compound [K(THF)₃(L¹⁵)Co^II]·THF (15) is the only six-coordinate species observed, inasmuch as two acyl arms generate five-membered loops that place two ether residues (ROMe) in the coordination sphere of the [N₃N]Co^II site (Co–O = 2.302(14), 2.405(13) Å; N_amine–Co–O = 146.4(6), 129.3(6)°). Identical K⁺ ions are coordinated by three THF molecules and two acyl residues, each located in neighboring molecules, thus giving rise to pseudo 1-D polymeric structures.

Finally, the structure of [K₂(THF)₂][K(L¹⁷)₂Co^II₂]₂ (17) is very similar to that observed for 7 with respect to the formation of two interconnecting dimers, but the K⁺ ions are organized differently, to afford a molecular (tetranuclear) rather than a polymeric complex. First, two K⁺ ions link the two dimers in 17, by employing the same contact pattern noted for the single K⁺ ion connecting the two dimers in 7. Second, the remaining two K⁺ ions in 17 are terminated by THF molecules, and thus do not provide connections that could generate an 1-D array of repeating tetranuclear units as in 7. Otherwise, the coordination and arrangement of the Co^II sites in 17 and 7 is very similar, with somewhat more pronounced contact for the N_amine residue (av. Co–N_amine = 2.40 Å), and concomitant weaker attachment of the oxygen (carbonyl) moiety (av. Co–O_C=O = 2.00 Å), along the axial coordination of Co^II sites in 17 versus that of 7.

Co^II Compounds with Alkyl-Armed Ligands.

The new methyl-substituted compound 16 and the previously reported isopropyl congener (9)⁴⁸ are the only members of the alkyl-armed category of Co(II) compounds explored in this study. Compound 16 (Figure 4) demonstrates the familiar [N₃N] coordination, but unlike the more bulky isopropyl analogue, it exhibits a five-coordinate Co^II site due to the presence of a coordinated THF molecule trans to the N_amine residue (N_amine–Co–O_THF = 175.75(13)°). The electron-rich alkyl substitution dictates a stronger equatorial ligand field (av. Co–N_amido = 2.000 Å) by comparison to all other five-coordinate Co^II sites investigated in this study. The polymeric nature of the compound arises by means of a repeating −[Co(1)–K(1)]–sequence, which features K(1) ions engaging in K–N_amido and K⁺–arene contacts with both ligands of adjacent Co(1) sites.

Figure 4.

ORTEP diagram of [K(L¹⁶)Co^II–THF]·0.5Pentane (16) drawn with 40% thermal ellipsoids. Hydrogen atoms are omitted for clarity. Selective interatomic distances (Å) and angles (deg): Co(1)–N(1) = 2.226(3), Co(1)–N(2) = 2.007(4), Co(1)–N(3) = 1.990(4), Co(1)–N(4) = 2.004(4), Co(1)–O(1) = 2.207(3), Co(1)–[N(2), N(3), N(4)] = 0.37(2) (distance of Co from mean plane), N(2)–Co(1)–N(4) = 112.70(15), N(2)–Co(1)–N(3) = 114.04(15), N(3)–Co(1)–N(4) = 122.94(15), N(2)–Co(1)–N(1) = 79.40(13), N(4)–Co(1)–N(1) = 78.61(14), N(3)–Co(1)–N(1) = 79.64(14), N(3)–Co(1)–O(1) = 100.45(14), N(2)–Co(1)–O(1) = 104.32(13), N(4)–Co(1)–O(1) = 97.91(14), N(1)–Co(1)–O(1) = 175.75(13).

Co^II Compounds with Aryl-Armed Ligands.

Among the seven aryl-supported Co(II) reagents shown in Figure 2 (1, 2, 11, 12 are new; 3, 5, 13 have been previously reported⁴⁸), only those crystallized from MeCN solutions (3, 13) feature a solvent molecule coordinated to the Co(II) center. All others, crystallized from THF solutions, exhibit four-coordinate [N₃N]Co(II) sites devoid of any axial THF residues, in sharp contrast to analogous five-coordinate Mn and Fe reagents previously reported.

Among the four new Co(II) complexes (Figure 5), (L¹)Co (1) provides nice green crystals from concentrated THF solutions, but single-crystal specimens (albeit of low quality) were only obtained in the presence of the exceptional K⁺ binder 2.2.2-cryptand. The resulting ionic complex [K(2.2.2-cryptand)][(L¹)Co^II]·3THF (1) exhibits a distorted [N₃N] coordination with an equatorial ligand field (av. Co–N = 1.927 Å) that is equal or stronger than that demonstrated by similar four-coordinate Co(II) sites supported by aryl substituents (2, 5, 12), presumably due to the electron-rich character of the 4-^tBu-substituted phenyl arm. Similarly, the 3,5-^tBu₂ disubstituted compound (L²)Co (2) proved to be isolable only in the presence of 2.2.2-cryptand, in the form of green crystals of [K(2.2.2-cryptand)][(L²)Co^II]·1.5Pentane (2) of marginal quality. Its structure is almost identical with that of 1, with a similarly strong equatorial field (av. Co–N = 1.927 Å). The corresponding 3,5-Me₂ disubstituted compound [K(THF)₃(L¹²)Co^II]·THF (12) is monomeric and geometrically analogous to 2, with a weaker equatorial field (av. Co–N = 1.956 Å), but, as opposed to 1 and 2, can be isolated without the assistance of 2.2.2-cryptand. The K⁺ ion in 12 is supported by three THF molecules and a host of contacts with aromatic moieties and N atom residues.

Figure 5.

ORTEP diagrams (from left to right) of [K(2.2.2-cryptand)][(L¹)Co^II]·3THF (1), [K(2.2.2-cryptand)][(L²)Co^II]·1.5Pentane (2), [K(THF)(L¹¹H)Co^II–OH]₂ (11), and [K(THF)₃(L¹²)Co^II]·THF (12) drawn with 40% thermal ellipsoids. Hydrogen atoms are omitted for clarity. Selective interatomic distances (Å) and angles (deg): 1, Co(1)–N(1) = 2.102(18), Co(1)–N(2) = 1.946(18), Co(1)–N(3) = 1.92(2), Co(1)–N(4) = 1.915(19), Co(1)–[N(2), N(3), N(4)] = 0.17(2) (distance of Co from mean plane), N(2)–Co(1)–N(4) = 122.7(8), N(2)–Co(1)–N(3) = 118.0(8), N(3)–Co(1)–N(4) = 117.0(8), N(2)–Co(1)–N(1) = 84.0(8), N(4)–Co(1)–N(1) = 85.1(8), N(3)–Co(1)–N(1) = 85.6(8); 2, Co(1)–N(1) = 2.085(8), Co(1)–N(2) = 1.921(7), Co(1)–N(3) = 1.933(8), Co(1)–N(4) = 1.926(7), Co(1)–[N(2), N(3), N(4)] = 0.14(1) (distance of Co from mean plane), N(2)–Co(1)–N(4) = 120.4(3), N(2)–Co(1)–N(3) = 118.2(3), N(3)–Co(1)–N(4) = 119.8(3), N(2)–Co(1)–N(1) = 85.9(3), N(4)–Co(1)–N(1) = 86.0(3), N(3)–Co(1)–N(1) = 85.5(3); 11, Co(1)–N(1) = 2.260(5), Co(1)–N(2) = 1.956(5), Co(1)–N(3) = 1.939(5), Co(1)–O(1) = 1.947(4), K(1)–O(1) = 2.675(5), K(1)–O(2) = 2.677(5), Co(1)–[N(2), N(3), N(4)] = 0.62(2) (distance of Co from mean plane), N(2)–Co(1)–N(3) = 118.7(2), N(2)–Co(1)–N(1) = 81.6(2), N(3)–Co(1)–N(1) = 80.1(2), N(1)–Co(1)–O(1) = 139.41(18), N(2)–Co(1)–O(1) = 108.5(2), N(3)–Co(1)–O(1) = 122.9(2), Co(1)–O(1)–K(1) = 113.83(18), O(1)–K(1)–O(2) = 116.95(15); 12, Co(1)–N(1) = 2.121(7), Co(1)–N(2) = 1.950(8), Co(1)–N(3) = 1.964(8), Co(1)–N(4) = 1.955(7), Co(1)–[N(2), N(3), N(4)] = 0.20(1) (distance of Co from mean plane), N(2)–Co(1)–N(4) = 119.2(3), N(2)–Co(1)–N(3) = 121.5(3), N(3)–Co(1)–N(4) = 116.2(3), N(2)–Co(1)–N(1) = 84.3(3), N(4)–Co(1)–N(1) = 84.0(2), N(3)–Co(1)–N(1) = 84.0(3).

Although both the 3,5-Me₂ and 2,6-Me₂ disubstituted compounds 12 and 13, respectively, can be isolated and characterized, the corresponding 2,4,6-Me₃ trisubstituted species (L¹¹)Co^II has proven to be difficult to synthesize, apparently due to extreme sensitivity to even traces of water. In contrast, the analogous [K(THF)₃(L¹¹)Mn^II–THF] can be readily prepared.⁴⁴ Indeed, after initial formation of a deep green species in the reaction of deprotonated L¹¹H₃ and CoCl₂ in THF, the color soon fades and ligand can be recovered intact along with separation of blue Co(OH)₂. In one instance, under scrupulous water exclusion, a few green crystals of [K(THF)(L¹¹H)Co^II–OH)]₂ (11) have been isolated, amounting to a species that can be viewed as the formal product of water addition to (L¹¹)Co^II. Indeed, 11 features protonation and dissociation of one nitrogen residue from the equatorial field, with concomitant formation of a Co^II–OH moiety. The hydroxide is further coordinated by two K(THF)⁺ ions in an overall dimeric structure that connects two inversion-related (L¹¹H)Co^II–OH monomers by means of a K₂(OH)₂ rhomb.

Structures Featuring Ligand Rearrangement.

In two instances, we have isolated a few compounds that exhibit a characteristic oxidative ligand rearrangement in the presence of traces of dioxygen or one-electron oxidants. Similarly reorganized compounds, featuring electron-donor substituents, have been previously studied in our lab and attributed to the formation of an incipient aminyl radical.^45b,49 Indeed, the electron-rich (L²)Co^II (2) is highly sensitive to oxidative rearrangement, and provided two crystallographically characterized species, [K(THF)₃(L²_re)Co^II–THF] (2b) and [(L²_re,ox)-Co^II–THF]·0.5 Pentane (2c) (Scheme 1 and Figure S2; bonds broken in 2 and formed in 2b and 2c are shown in red), formed in comparable amounts. Compound 2c is not only ligand-rearranged, but also one-electron oxidized, as noted by relevant metrical parameters associated with the phenylene ring between atoms N₁ and N₂ (Figure S2). A possible overall stoichiometry for this reaction can be written as 2[(L²)Co^II]⁻ → [(L²_re)Co^II]⁻+ [(L²_re,ox) Co^II] + e⁻.

Scheme 1.

Compound (L¹⁶)Co^II (16) is also highly sensitive to the same type of ligand rearrangement in the presence of traces of dioxygen, affording the isolable dimer [K(THF)₂(L¹⁶_re)Co^II]₂ (16b, Figure S3), which is equivalent to 2b in terms of ligand reorganization. In this case, we were not able to isolate any other compound that might provide evidence for the location of the oxidizing equivalent(s).

Electrochemistry.

Ten Co(II) compounds possessing aryl (L³Co, L⁵Co, L¹³Co), alkyl (L⁹Co), and acyl arms (L⁴Co, L⁶Co, L⁷Co, L⁸Co, L¹⁰Co, L¹⁷Co) were selected as representative examples for examination by cyclic voltammetry. Electro-chemical data for a handful of these examples have been previously reported.⁴⁸ Figure 6 provides a collective presentation of the corresponding waves (first oxidation event), and Table S2 summarizes relevant electrochemical data (potentials are reported versus the ferrocenium/ferrocene (F_c⁺/F_c) couple). All aryl- and alkyl-armed Co(II) compounds examined by cyclic voltammetry feature semireversible waves at negative potentials, ranging from −0.665 (L¹³Co) to −0.090 V (L³Co), in accordance with the electron-rich nature of the corresponding substituents. Specifically for the alkyl-armed L⁹Co, the two closely spaced, semireversible waves observed (−0.654, −0.500 V), may represent the two slightly different Co(II) sites in the crystal structure. Given the almost identical wave currents for all these aryl- and alkyl-armed Co(II) compounds (3.0 M), and their anodic shifts with respect to the analogous Mn(II)/Mn(III) and Fe(II)/Fe(III) couples^44,46a by approximately 0.65 and 0.25 V, respectively, we assign the corresponding semireversible waves to essentially metal-centered Co(II)/Co(III) cycles.

Figure 6.

Cyclic voltammograms of compounds [K(L³)Co^II–NCMe] (3) and [K(NCMe)₃(L¹³)Co^II–NCMe] (13) in MeCN/(ⁿBu₄N)PF₆, [K(L⁴)Co^II]·Et₂O (4), [K(THF)₆][(L⁵)Co^II]·1.5THF (5), [K(MeCN)(L⁶)Co^II–NCMe]·2MeCN (6), [K₂(DMA)₄][K(L⁷)₂Co^II₂]₂·2DMA (7), [K(THF)₂(L⁸)Co^II] (8a), [K₂(L⁹)₂Co^II₂] (9), and [K(THF)K(L¹⁷)₂Co^II₂]₂·3Pentane (17) in DMF/(ⁿBu₄N)PF₆, and [K₂(DMA)₃(L¹⁰)₂Co^II₂]·0.5Et₂O (10) in DMA/(ⁿBu₄N)PF₆, with a Au disk electrode (1.6 mm in diameter); scan rate 0.1 V/s.

In sharp contrast, the acyl-armed Co(II) compounds examined demonstrate irreversible anodic ways with variable i_p,a values (mostly large versus the aryl/alkyl-substituted congeners), suggesting significant ligand-centered contributions. In addition, the first anodic wave shown in Figure 6 overlaps with a subsequent oxidation wave (not shown), rendering any attempts to garner further information from exhaustive electrolysis futile. Nevertheless, all initial anodic waves for the acyl-armed Co(II) sites are shifted to positive potentials (0.032 (L¹⁷Co) to 0.719 V (L⁸Co)), reflecting the effect of the electron-withdrawing acyl arm on the Co(II) center, albeit in an order not always consistent with the electronic nature of each individual acyl group. The most significant deviation is observed for L¹⁰Co (E_p,a = 0.559 V), whose anodic shift versus that of L¹⁰ Mn (E_p,a = −0.108 V; tentatively assigned to Mn(II)/Mn(III))⁴⁴ betrays very little, if any, metal-centered contributions to the anodic wave. The intervention of ligand-centered oxidation events does not permit any secure identification of the solution structure of species such as L⁷Co and L¹⁰Co, which exhibit single anodic waves even in the presence of dinuclear units in their polymeric solid-state structures. On the other hand, the dual anodic-wave feature for L¹⁷Co (solid-state tetramer) may signify a dinuclear structure in solution. Finally, the catalytically important L⁸Co (−COCF₃ arm) possesses the indisputably most electrochemically stable Co(II) site encountered in this series of Co(II) compounds, in agreement with previous findings for the L⁸ Mn and L⁸Fe analogues.^44,46a

EPR Spectroscopy.

EPR spectra of selected Co(II) compounds were recorded from frozen DMF solutions of 3, 4, 5, 6, 7, 8a, 9, 10, 13, 14, 15, and 17. In all cases, the samples give rise to signals that are consistent with isolated Co^II (S = 3/2) species.⁵⁰ Spectra recorded at 10 K are shown in Figure 7. The spectra can be interpreted within the framework of the spin Hamiltonian:⁵¹

$${\widehat{H}}_{\text{Zfs}}=D\left[{\widehat{S}}_z^2-\frac{S\left(S+1\right)}{3}+\frac{E}{D}\left({\widehat{S}}_x^2-{\widehat{S}}_y^2\right)\right]+\widehat{S}\cdot A\cdot \widehat{I}+\mathit{\beta }B\cdot g_0\cdot \widehat{S}$$ (1)

Figure 7.

Experimental (black line) and theoretical (red-dashed line) EPR spectra from frozen DMF solutions of complexes 3, 4, 5, 6, 7, 8a, 9, 10, 13, 14, 15, and 17. For 3, 4, 6, 7, 8a, 10, and 17, the subspectra from sites 1 (green), and 2 (blue) are also shown, as described in the text. A spike signal at ca. 3400 G in the spectra of 5, 9, and 10 is due to an impurity in the cavity. EPR conditions: Temperature, 10 K; modulated amplitude 10 Gpp; microwave power, 2.0 mW; microwave frequency, 9.4 GHz.

In eq 1, D and E are the zero field splitting (zfs) parameters. A is the hyperfine tensor relevant to the hyperfine interactions of the electronic (S = 3/2) and nuclear (I = 7/2 for ⁵⁹Co) spins, and g₀ is the intrinsic g-tensor of the Co(II) ion. For simplicity we assume that the principle axes of the tensors are parallel to each other. Often, due to large spin orbit coupling effects for Co^II (S = 3/2), the zfs parameter |D| is quite large, whereas A and g₀ are characterized by significant anisotropy. Under the influence of zfs, the 4-fold degeneracy is partially lifted in zero magnetic field, yielding two Kramers’ doublets, |±1/2〉 and |±3/2〉 separated by 2|D|. The |±1/2〉 (or |±3/2〉) doublet is the ground state for positive (or negative) D. For |D| ≫ hν (≈ 0.31 cm⁻¹ at X-band), each Kramers’ doublet can be described by an effective S_eff = 1/2, giving rise to an EPR spectrum characterized by an anisotropic g_eff tensor.

The EPR signals shown in Figure 7 are consistent with the | ±1/2〉 doublet. No detailed temperature dependence of the spectra was pursued in the present work. However, spectra recorded at 4.2 and/or 20 K (not shown) indicate that the intensity of the signals (scaled as Intensity × Temperature) decreases as temperature increases. This suggests that the |±1/2〉 doublet is the ground state, implying a positive value for D. For |D| ≫ hν, the spectra observed do not depend on D but rather on the rhombic zfs parameter E/D, the values of the intrinsic g₀-tensor,⁵² and the hyperfine term. With the exception of the spectrum for complex 15, the g₀-tensor was assumed axial (g_0x = g_0y = g_0⊥ ≠ g_0z = g_∥). In several cases (3, 4, 6, 7, 8a, 10, 17), the spectra indicated the presence of two Co(II) species, characterized by a different value for the parameter E/D. Assuming a common value for |D|, the simulations determine the relative abundance of each Co(II) site. The specific line shape of the spectra results from a combination of factors, including distributions on the parameter E/D (E/D strain), unresolved hyperfine interactions, and residual line-broadening mechanisms.⁵³ The EPR parameters for all samples, as well as the relative ratios of the two species when applicable, are presented in Table S3. Because an unequivocal deconvolution of the line broadening mechanisms is not feasible, the quoted values for the relevant parameters are indicative.

With the exception of the six-coordinate 15, all other complexes (four- or five-coordinate) exhibit rhombic parameters (E/D) that lie in the interval [0, 0.19], indicating different degrees of rhombicity. All complexes (except 15) demonstrate a valley-shaped signal at g ~ 2.0, which corresponds to the g_∥ component of the g_eff-tensor. This signal is broadened due to hyperfine interactions, and in some cases (7, 9, 17) the hyperfine lines are well resolved. The simulations indicate that A_∥ values in all cases are in the range of 200–300 MHz and that the hyperfine term has to be taken into account in order to reproduce the low field region of the spectra, corresponding to the perpendicular components of the tensors in eq 1. Due to restrictions in the determination of the line-broadening mechanisms, it is not possible to unambiguously evaluate the magnitudes of the A_x and A_y. Therefore, an average value, A_⊥ = (A_x + A_y)/2, is quoted in Table S3. This value ranges in the 30–90 MHz range. Complex 15 exhibits a relatively sharp peak at g = 6.24 and an extremely broad derivative feature with an apparent zero crossing at ca. g = 2.07. This behavior indicates a rhombic system with E/D = 0.33. From this point of view, complex 15 exhibits unique EPR properties in this series, most likely due to its special, six-coordinate ligand field, featuring two O residues in addition to the familiar [N₃N] coordination. The origin of dual EPR-active species in DMF solutions of some compounds cannot be ascertained at the present time, especially since structural data from crystals derived from DMF solutions are not available. Possible sources are small deviations in the coordination of Co(II) sites in polymeric species (as noted for 7, crystallized from DMA solutions), and also potential differentiation arising from partial DMF coordination to Co(II) and even K⁺ sites.

In summary, the EPR studies from frozen DMF solutions of the complexes indicate that all complexes feature a Co(II) (S = 3/2) ion with a large and positive zfs term, D, and variable degree of rhombicity. The existence of more than one species in some cases suggest that the coordination environment of the Co ion can be quite flexible in solution.

Catalytic Aziridinations of Olefins.

Styrene.

The 17 Co^II compounds shown in Figure 2 were first explored as catalysts (5 mol %) for the aziridination of styrene (8.0 equiv) by PhINTs (1.0 equiv) in chlorobenzene at room temperature (Table 1). The high-yielding L⁴Co^II was first investigated as catalyst in several solvents (MeCN, 50%; 2,2,2-Trifluoroethanol, 62%; PhCF₃, 61%; CH₂Cl₂, 61%; Benzene, 53%; PhCl, 70%) and found to afford superior levels of aziridine in chlorobenzene (similar solvent-screening results were previously obtained with Mn^II reagents⁴⁴). The yield drops significantly (25%) if equimolar amounts of styrene and PhINTs are employed in PhCl, whereas a 2-fold excess of styrene over PhINTs improves the yield to 46%. Increasing the amount of the L⁴Co catalyst to 10 mol % suppresses the yield to 53%. The styrene aziridination yields obtained with various Co^II catalysts (Table 1) vary significantly (18–70%) as a function of the ligand employed. As opposed to the corresponding L^xMn^II reagents (x = 1–15)⁴⁴ that reveal a dominant relationship between increasing styrene aziridination yields with anodically shifting Mn^II/Mn^III redox potentials (and, by extension, with increasing electrophilicity of the putative Mn^III–^•NR), the Co^II catalysts provide yields that indicate a more complex pattern of electronic and steric contributions. For example, although the comparatively electron-deficient, acyl-armed reagents remain, on average, more productive than aryl- or alkyl-armed congeners, a comparison between the −COCMe₃ (L⁴) and −COCF₃ (L⁸) supported Co^II reagents furnishes essentially the same aziridination yields, in spite of the fact that the Co(II)/Co(III) couple for L⁸Co is anodically shifted by ~500 mV versus L⁴Co. In addition, L⁸Co is potentially less congested than L⁴Co. In contrast, the Mn^II and Fe^II congeners are associated with significantly diverging yields in favor of the more electron-deficient L⁸M reagents (L⁴Mn, 12%; L⁸Mn, 75%; L⁴Fe, 45%; L⁸Fe, 73%), correlating with an anodic potential shift of approximately 600 mV for both the L⁸Mn and L⁸Fe catalysts with respect to their L⁴M analogues. More importantly, as indicated below, L⁴Co (4) is faster than L⁸Co (8b, 8c) in mediating styrene aziridination. The closely related L⁷Co (−COⁱPr arm) is in principle slightly less electron rich and sterically congested than the L⁴Co congener, but affords lower yields than L⁴Co in a slower reaction (the opposite is true for the corresponding Mn(II) catalysts).⁴⁴ Two other acyl-substituted Co^II catalysts (L¹⁴Co, L¹⁷Co), exhibiting significantly lower yields, are indicative of how the metal-nitrene electron-affinity bias may be overridden by other electronic or steric factors in the fairly restricted reaction cavity of Co reagents. The corresponding L¹⁴ Mn catalyst, by contrast, is among the most productive Mn^II aziridination reagents examined (yield: 67%).⁴⁴ As expected, the aryl-substituted ligands generate Co^II reagents that are poor mediators of styrene aziridination. These sites are oxidatively and even hydrolytically sensitive and tend to generate thermodynamic sinks. Nevertheless, the muted role of the electron-affinity criterion can still be discerned in the series of the electron-rich, aryl-armed reagents (1, 2, 3, 5, 12, 13), inasmuch as the highest-yielding L¹³Co (13) is the most electron rich member of the group. Finally, the alkyl-substituted L⁹Co and L¹⁶Co reagents are only modestly productive, as anticipated for electron-rich Co^II sites, but again the more electron rich, isopropyl-substituted L⁹Co provides higher yields than the methyl-substituted congener L¹⁶Co. However, the latter undergoes facile oxidative ligand rearrangement that may compromise its structural integrity.

Table 1.

Yields of Styrene Aziridination Mediated by Co^II Reagents 1–17^a

compound	yield (%)
[K(2.2.2-cryptand)][(L1)CoII]·3THF (1)	18
[K(2.2.2-cryptand)][(L2)CoII]·1.5Pentane (2)	18
[K(L3)CoII−NCMe] (3)	36
[K(L4)CoII]·Diethyl Ether (4)	70
[K(THF)6](L5)CoII]·1.5THF (5)	32
[K(NCMe)(L6)CoII−NCMe]·MeCN·0.5H2O (6)	50
[K2(DMA)4][K(L7)2CoII2]2·2DMA (7)	59
[K(NCMe)(L8)CoII−NCMe] (8b)	69
[K(DMA)3(L8)CoII]2 (8c)	68
[K2(L9)2CoII2]n (9)	38
[K2(DMA)3(L10)2CoII2]·0.5Et2O (10)	49
[K(THF)(L11H)CoII−OH)]2 (11)	–
[K(THF)3(L12)CoII]·THF (12)	25
[K(NCMe)3(L13)CoII−NCMe] (13)	38
[K(DMA)(L14)CoII]·DMA (14)	36
[K(THF)3(L15)CoII]·THF (15)	45
[K(L16)CoII−THF] (16)	25
[K(THF)K(L17)2CoII2]2 (17)	35

^aConditions: Catalyst, 0.0125 mmol (5 mol %); PhINTs, 0.25 mmol; styrene, 2.0 mmol; MS 5 Å, 20 mg; PhCl, 0.200 g; 30 °C; 12 h.

Alkene Aziridinations by [K(L⁴)Co^II]·Diethyl Ether (4).

The highest yielding L⁴Co (4) catalyst was subsequently investigated as a nitrene-transfer (NTs) mediator with a panel of aromatic, cyclo (Table 2), and aliphatic alkenes (Table S4). Styrenes substituted at the para position with both electron-donor and electron-acceptor groups were examined first (entries 1–9) under the conditions noted in the previous section. Both methylene chloride and chlorobenzene were employed (as well as acetonitrile in a few instances), invariably resulting in better aziridination yields in chlorobenzene. In the majority of cases, yields above 70% were recorded irrespective of the electron-rich or poor character of the substituent, with the exception of two moderately yielding cases involving strong electron-withdrawing groups (p-CF₃, p-NO₂). Moreover, the product of 4-MeO-styrene aziridination is known to be unstable,⁵⁴ and is thus associated with modest yields. Overall, the aziridination yields for these para-substituted styrenes, save for the parent styrene, are comparable to those previously reported for L⁸Co. Ortho-substitution (entry 10) affects aziridination yields, presumably interfering with nitrene-transfer both sterically and electronically (due to the orthogonal orientation of the aromatic versus the olefinic plane of the substrate⁵⁵). Similar steric inhibition is also observed for α-substituted styrenes (methyl, phenyl; entries 11 and 12), especially for the bulkier α-phenyl-styrene. An allylic amination product is also obtained in both cases, ascribed to aziridine-ring opening,⁵⁶ which is more pronounced for the α-phenyl-substituted product.⁵⁴ Steric hindrance is also evident in the aziridination of β-substituted styrenes (entries 13–16), especially for the bulky cis- and trans-stilbene (entries 15, 16). In agreement with previous observations in the application of L⁸Co,⁴⁴ the cis congeners are more productive (entries 14, 15), with significant loss of stereochemical integrity. For all these encumbered substrates, L⁸Co is on average more productive than L⁴Co, presumably reflecting the somewhat more voluminous reaction cavity of L⁸Co. Allylic or benzylic aminations compete successfully with aziridinations (entries 17–22), unless cis (entry 18) and/or electron-rich (entry 21) olefins are involved. Beyond cycloalkenes, other terminal or internal aliphatic olefins exhibit low aziridination yields (Table S4), in accordance with previous results pertaining to the application of L⁸Co in aziridinations of aromatic and aliphatic olefins.⁴⁴ A competitive styrene versus 1-hexene (1.0 mmol each) aziridination by PhINTs (0.25 mmol) catalyzed by L⁴Co (5 mol %) in chlorobenzene provided a ratio of 25:1 in favor of the styrene aziridination product (combined aziridine yield: 72%), not unlike the L⁸Co catalyst (28:1, yield 73%).⁴⁴

Table 2.

Yields of Aziridination/Amination of Olefins by [K(L⁴)Co^II]·Et₂O (4)^a

^aConditions: 4, 0.0125 mmol (5 mol %); PhINTs, 0.25 mmol; olefin, 2.0 mmol; MS 5 Å, 20 mg; solvent (MeCN/CH₂Cl₂/PhCl) 0.200 g; 30 °C; 12 h.

^bIn MeCN.

Mechanistic Studies.

Comparative Reaction Profile.

The formation of the product of styrene aziridination was monitored as a function of time (Figure 8) for catalysts L⁴Co (4), L⁷Co (7), and L⁸Co (8c) (all crystallized from DMA/ether), under the conditions noted above (Table 1), with the exception of the amount of PhCl used (500 mg). Yields were determined after quenching the reaction at various time intervals. Surprisingly, the more electron rich and sterically congested L⁴Co (4) exhibits faster product generation than L⁸Co (8c) during the first hour, whereas L⁷Co (7) is kinetically comparable to 8c. At the 1.0 h mark, the reaction is complete by 88% for L⁴Co (4), 65% for L⁸Co (8c), and 76% for L⁷Co (7). Interestingly, the MeCN-crystallized version of L⁸Co, [K(NCMe)(L⁸)Co^II–NCMe] (8b), has been previously shown⁴⁴ to be significantly slower; only 14% of the reaction is complete after 1.0 h in d⁵-PhCl, indicating potential interference by the strongly coordinating acetonitrile. However, 8b is still much faster than the corresponding [K(NCMe)(L⁸)M^II–NCMe] (M = Mn, Fe) reagents, presumably reflecting the superior electrophilicity of Co^III–^•NR as discerned from E_p,a values associated with the M^II/M^III couple of the divalent catalysts (Fe: 0.228 V, Mn: 0.518 V, Co: 0.837 V).⁴⁴

Figure 8.

Yield of aziridine (%) as a function of time (min) in the reaction of styrene (2.0 mmol) by PhI = NTs (0.25 mmol) mediated by L⁴Co (4), L⁸Co (8c), and L⁷Co (7) (0.0125 mmol with respect to Co) in chlorobenzene (0.500 g) at 30 °C.

Hammett Analysis.

Several para-substituted styrenes (para substituent: Me, ^tBu, F, Cl, CF₃, NO₂; 1.0 mmol each) were subjected to competitive aziridination (PhINTs, 0.25 mmol) versus styrene (1.0 mmol), mediated by L⁴Co (5 mol %) in chlorobenzene (0.200 g), in the presence of 5 Å molecular sieves (25 mg). Hammett plots of log(k_X/k_H) (determined by ¹H NMR from the ratio of the corresponding aziridines) as a function of the substituent polar parameter σ_P or even the resonance-responsive parameter σ⁺ did not provide any reliable linear free-energy correlations (Figure S4, Table S5). In contrast, Jiang’s dual-parameter correlation that incorporates both polar (σ_mb) and spin-responsive (σ_JJ*) parameters (log(k_X/k_H) = ρ_mbσ_mb + ρ_JJ*σ_JJ* + C)⁵⁷ provides a reasonable linear correlation (R² = 0.98; Figure 9). The negative ρ_mb value is consistent with a small positive charge developing at the benzylic carbon, whereas the always positive ρ_JJ* value denotes an incipient radical character for the same site. The ratio |ρ_mb/ρ_JJ*| = 1.12 is similar to the one previously observed for L⁸Co catalyzed aziridinations (|ρ_mb/ρ_JJ*| = 1.0),⁴⁴ and indicates competitive contributions of polar and spin-delocalization effects. Polar effects are dominant in many Rh,^24d Cu,^24d,58 and Fe^18g,59,60 catalyzed aziridinations, for which Hammett plots can be fit with the assistance of polar parameters alone (σ_p, σ⁺), but the need for incorporating spin-delocalization parameters (σ*, σ_JJ*)^57,61 with a wide range of |ρ_mb/ρ_JJ*| values (0.04–2.02) is also evident in many other metal-catalyzed aziridinations.^{18e,22d,g,23f,g} More recently, larger |ρ⁺/ρ_JJ*| values have been reported for the aziridination of styrene by PhINNs mediated by [Co^III(TAML^red)]⁻ (5.71) and [Co^III(TAML^sq)] (8.64), in accordance with a novel mechanism that involves a partial single-electron transfer from the styrene to the metal-nitrene as a component of the turnover-determining step.⁴¹ A similar mechanistic scenario has also been advanced for Fe-mediated aziridinations, but in this case Hammett correlations can be successfully accommodated with polar parameters alone (σ⁺).⁴²

Figure 9.

Linear free energy correlation of log(k_X/k_H) vs σ_mb, σ_JJ* for the aziridination of para-substituted styrenes (X = Me, ^tBu, F, Cl, CF₃, NO₂) mediated by L⁴Co (4).

Kinetic Isotope Effect and Stereochemical Integrity.

Evaluation of the secondary kinetic isotope effect was accomplished by ¹H and ²H NMR with the assistance of deuterated styrenes (α-d-styrene, cis- and trans-β-d-styrene; 1.0 mmol each) in competitive aziridinations (PhINTs, 0.25 mmol) with styrene (1.0 mmol) catalyzed by L⁴Co (4) or L⁸Co (8c) (5 mol %) in chlorobenzene (Table 3). KIE values close to 1.0 were obtained with α-d-styrene for both catalysts, indicating that the α-styrenyl is unlikely to be involved in the initial nitrene attack to styrene. In contrast, the β-styrenes are associated with inverse KIE values for L⁴Co (cis: 0.90 (±0.02), trans: 0.92 (±0.02)) that can be attributed to a limited sp² → sp³ rehybridization of styrene’s C_β site upon development of the initial N–C_β bond (Scheme 2). More modest inverse KIE values are also noted in cis- and trans-β-d-styrene aziridinations mediated by L⁸Co (8b⁴⁴ or 8c) (cis: 0.96 (±0.02), trans: 0.98 (±0.02)), suggesting only minimal N–C_β bond formation in the transition state.

Table 3.

Secondary KIE Values in Aziridination of Deuterated Styrenes vs Styrene

catalyst
[K(L4)CoII]n (4)	1.01(±0.01)	0.90(±0.02)	0.92(±0.02)
[K(DMA)3(L8)CoII]2 (8c)	1.00(±0.01)	0.96(±0.02)	0.98(±0.02)

Scheme 2.

The kinetics of the aziridine ring closure (formation of the second N–C_α bond) was further evaluated by ²H NMR in the aziridinations of cis- and trans-β-d-styrene (Table 4), by examining the degree of stereochemical scrambling in the resulting aziridines (cis/trans partitioning due to C_α–C_β bond rotation; Scheme 2) in competition with N–C_α bond formation. The ratio of cis/trans aziridine (94:6) and trans/cis aziridine (92:8) resulting from the L⁴Co-mediated aziridination of cis-β-d-styrene and trans-β-d-styrene, respectively, signifies the interference of very small energy barriers in aziridine-ring closure. A slightly larger barrier is indicated for the aziridination of the more sensitive cis-β-d-styrene by L⁸Co (cis/trans aziridine: 89/11). On the other hand, the stereochemical scrambling observed in the aziridination of cis-β-methyl-styrene is more pronounced with L⁴Co than L⁸Co.

Table 4.

Exploration of Stereochemical Integrity in the Aziridination of cis- and trans-β-d-Styrene

catalyst
[K(L4)CoII]n (4)	94:6 (cis:trans aziridine)	92:8 (trans:cis aziridine)
[K(DMA)3(L8)CoII]2(8c)	89:11 (cis:trans aziridine)	92:8 (trans:cis aziridine)

Computational Studies.

The structure and electronic description of the presumptive [L⁴Co]NTs intermediate were explored by DFT calculations at the B3LYP/6–31+G(d) level of theory. Free energy calculations suggest that the intermediate-spin quartet ground state (S = 3/2) lies only 0.1 kcal mol⁻¹ lower than the high-spin sextet state (S = 5/2), and 2.8 kcal mol⁻¹ below the doublet state (S = 1/2). As mentioned above, the calculated free energies for [L⁸Co]NTs indicate that the high-spin sextet is the ground state, in agreement with the weaker ligand field provided by the L⁸ versus L⁴ ligand.

Calculated structures for the three spin-states of [L⁴Co]NTs along with key metrical parameters are presented in Figure 10. The most conspicuous feature of these structures is the dissociation of one arm from the equatorial coordination sphere of the metal (Co–N = 3.93 (quartet), 3.67 (sextet), 4.08 (doublet) Å). The axial Co–N_amine bond is also elongated (Co–N = 2.43 (quartet), 2.63 (sextet), 2.99 (doublet) Å), if not dissociated, by comparison to that of L⁴Co (2.151(11) Å). These features have been previously noted in the DFT structure of [L⁸Co]NTs, although the latter exhibits an additional Co–F equatorial contact (Co–F = 2.37 Å).⁴⁴

Figure 10.

DFT structures (minimal metal coordination) for [L⁴Co]NTs active species in different spin states (from left to right: quartet, sextet, doublet) optimized at the B3LYP/6–31+G(d) level of theory. Hydrogen atoms were omitted from the figure for clarity.

Most importantly, the calculated spin densities for all three spin states of [L⁴Co]NTs place a full oxidizing equivalent over the dissociated arm, hence generating a widely delocalized N-aryl amidyl radical (Figure 11). For the ground-state quartet, the computed spin density consists of ~3.2 unpaired e⁻ on Co, 0.79 e⁻ on the nitrene N atom and −1.0 unpaired e⁻ on the noncoordinating arm. Similar spin densities are calculated for the sextet (Co: 2.9 e⁻, N: 0.9 e⁻, ligand arm: 1.0 e⁻) and the doublet state (Co: 2.8 e⁻, N: −0.66 e⁻, ligand arm: −1.0 e⁻). This spin density distribution is accommodated by an electronic structure such as [(^•L⁴)Co(II)–^•NTs]⁻, featuring a high-spin Co(II) center (S = 3/2) and two oxidizing equivalents on the noncoordinating ligand arm and the N atom of the nitrene residue, respectively. In sharp contrast, the spin density of the ground-state sextet of [L⁸Co]NTs (Figure 11) is largely distributed on Co (3.3 e⁻) and the N atom of the nitrene (1.1 e⁻). In this case, the noncoordinating arm is redox innocent, and the residual spin density is centered over ligating N atoms in a typical spin polarization fashion. Hence, the sextet state of [L⁸Co]NTs is better accommodated with an [(L⁸)Co(III)–^•NTs]⁻ electronic description (S_Co = 2).

Figure 11.

Spin density on the calculated lowest-energy spin state of the putative cobalt nitrenoid intermediates: quartet [L⁴Co]NTs (left) and sextet [L⁸Co]NTs (right).

A global electrophilicy index (GEI) was also computed for [L⁴Co]NTs and [L⁸Co]NTs by employing Stephan’s improved methodology.⁶² For the quartet spin state, GEI is calculated to be 5.0 eV for [L⁴Co]NTs and 5.7 eV for [L⁸Co]NTs. The corresponding values for the sextet spin state are 4.7 and 6.1 eV for [L⁴Co]NTs and [L⁸Co]NTs, respectively. Thus, [L⁸Co]NTs is more electrophilic than [L⁴Co]NTs on the basis of the GEI criterion.

Unfortunately, all efforts to map the aziridination reaction coordinate starting from [L⁴Co]NTs and styrene have not been successful in locating an acceptable transition state for the initial N–C_b bond formation. All three spin states (quartet, sextet, doublet) of [L⁴Co]NTs generated large activation barriers for this initial step (~50 kcal/mol). However, when dispersion-corrected DFT was applied,⁶³ as appropriate for polarizable bulky groups such as the tert-Bu, the corresponding barriers were reduced by approximately 20 kcal/mol. These barriers are still significant by comparison to those we have previously identified for the reaction of [L⁸Co]NTs (sextet) and styrene (23.4 kcal/mol for the turnover-limiting N–C_b bond formation).⁴⁴ Further experimentation and attendant DFT calculations will be required to unravel reliable trends and contributing factors with the assistance of catalysts that feature substituents spanning the CF₃ to CMe₃ range.

FURTHER DISCUSSION AND CONCLUSIONS

In a rigorous recent account, Latour and co-workers⁴² highlight the importance of the electron affinity (EA) of iron-nitrene/imido species as a guiding principle for predicting their reactivity in a wide range of iron-mediated aziridinations. In these catalytic reactions, the iconic substrate styrene undergoes aziridinations by various iron-nitrene compounds (Fe = NR), under a general mechanistic scheme that designates the formation of the initial N–C_b bond as the rate-determining step, usually encountered in aziridinations with a two-step mechanism ([M]NR radical addition to styrene, ring-closure radical rebound). More importantly, this first step is front-loaded by significant charge transfer from styrene to the iron-nitrene and, thus, is crucially influenced by the EA of the active oxidant. The applicability of the EA as a general predictor of reactivity seems to be wide, but at the present time is largely confined within the realm of catalysts that provide Hammett correlations for the aziridination of para-substituted styrenes that can be accommodated with polar parameters alone (σ_P, σ⁺), or by a combination of polar and spin-delocalization parameters (σ*, σ_JJ*) with dominant polar contribution.

In an almost concurrent publication, de Bruin and co-workers⁴¹ advance a similar argument with the assistance of electrophilic Co(III)-nitrene radical aziridination reagents, generated from the reaction of anionic [Co^III(TAML^red)]⁻ or the one-electron oxidized and neutral [Co^III(TAML^sq)] with PhINNs. Hammett plots for para-substituted styrene aziridinations are fitted with both polar (σ⁺) and spin-delocalization parameters (σ_JJ*) with large |ρ⁺/ρ_JJ*| values (5.71 for [Co^III(TAML^red)]⁻ and 8.64 for [Co^III(TAML^sq)]), hence these systems can also be considered as good candidates for testing the EA criterion. Indeed, the larger |ρ⁺/ρ_JJ*| value for [Co^III(TAML^sq)] and DFT calculations indicate that the energy barrier for the initial, rate-limiting reaction of the incipient [Co^III]NNs and styrene to generate the N–C_b bond is lower for [Co^III(TAML^sq)] versus [Co^III(TAML^red)]⁻, in agreement with an anticipated higher EA value for the nitrene species generated from [Co^III(TAML^sq)]. Surprisingly, the experimental rate for the aziridination of styrene by these two catalysts favors [Co^III(TAML^red)]⁻ versus [Co^III(TAML^sq)], but this has been attributed largely to the instability of the latter reagent. A distinctive feature of the Co(III) systems is associated with the redox noninnocent character of the tetraanionic ligand (TAML^red)⁴⁻, which can be successively oxidized in one-electron steps to (TAML^sq)³⁻ and (TAML^q)²⁻. The authors argue convincingly that the emerging radical nitrene species [Co^III(TAML^sq)(^•NNs)]⁻/Co^III(TAML^q)(^•NNs)₂]⁻ and [Co^III(TAML^q)(^•NNs)], resulting from [Co^III(TAML^red)]⁻ and [Co^III(TAML^sq)], respectively, interact with styrene in the rate-limiting step by means of an asynchronous transition state, encompassing significant single-electron transfer from styrene to the oxidized TAML ligand and a nucleophilic attack by the nitrene lone pair (in lieu of the ^•NNs p radical) at the incipient styrenyl radical cation. The attendant single-electron relocation (TAML → Co(III) → ^•NNs) reestablishes the N lone pair and retains the Co(III) oxidation state. Similar participation of charge transfer in the rate-limiting transition state between high-valent metal nitrenes/imidos and sulfides has been recently showcased for many other nitrene-transfer catalysts⁶⁴ and is now established as a common mechanistic feature.

In a previous comprehensive study from our lab,⁴⁴ we have shown that a library of Mn(II) catalysts, supported by the vast majority of the ligands used in the present work (L^1–15H₃), mediates alkene aziridinations with reactivity that increases in parallel with increasing electrophilicity of the putative [Mn^III]–^•NTs active oxidant. Moreover, the electrophilicity criterion holds across the base metals, inasmuch as the reactivity of the best performing Mn(II) reagent L⁸ Mn is inferior to that of the more acidic L⁸Co. Although the electron affinity of the metal-nitrene reigns supreme for all these reagents, the molecular interaction between [M^III]–^•NR and styrene in the rate-limiting formation of the initial N–C_b bond, is quite distinct with respect to the reagents explored by Latour and de Bruin. Indeed, Hammett correlations for the aziridination of styrenes mediated by our L⁸M reagents (M = Mn, Fe, Co) reveal rather modest positive charge buildup on the α-styrenyl carbon (increasing with metal acidity in the expected order: Fe < Mn < Co), and require the inclusion of competitive spin-delocalization contributions (|ρ_mb/ρ_JJ*| = 0.75 (Mn), 1.17 (Fe), 1.00 (Co); the correlation for Fe was rather weak). The fact that these reagents demonstrate more modest charge-transfer characteristics is consistent with the operation of presumptive metal nitrenes ([M^III]–^•NR) resting at a lower oxidizing level than the high-valent iron and cobalt nitrenes of Latour and de Bruin, respectively. Overall, these Mn(II) reagents and congeners can also be accommodated under the general EA criterion advanced by Latour (after all, they are catalysts engaged in typical electrophilic radical reactions), although they are not characterized by a dominant charge-transfer component. Incidentally, a strongly enhanced radical contribution, as in the case of Betley’s iron dipyrrinato complexes (|ρ_mb/ρ_JJ*| = 0.04 for NAd),^18e has been interpreted⁴² as the result of a competitive energy barrier for the second, ring-closing step (radical rebound).

The library of the Co(II) reagents (S = 3/2) reported in this work showcases some surprising deviations from the EA criterion. Although the importance of the electrophilicity of the metal-nitrene can still be detected in the relative enhanced yields provided by the Co(II) compounds possessing acyl-versus aryl- or alkyl-substituted ligands, the trend is certainly not as smooth and predictable as that previously encountered with the Mn(II) reagents.⁴⁴ Indeed, a closer inspection of the acyl-substituted subset of the Co(II) library of reagents reveals a wide range of yields in the aziridination of styrene that cannot be correlated with the anticipated electrophilicities of the corresponding cobalt-nitrene moieties. To further pinpoint the provenance of these disparities, we selected the high-yielding L⁸Co (−COCF₃ arm) and L⁴Co (−COCMe₃ arm) for further investigation. The L⁸Co was previously studied⁴⁴ in tandem with the L⁸Mn and L⁸Fe congeners and found to be more reactive and selective than the other two base metal analogues. Mechanistic and computational studies showed that all three L⁸ M reagents follow a two-step styrene aziridination path (turnover-limiting addition of [L⁸ M^III]–^•NTs to the β-styrenyl carbon followed by product-determining ring-closure via radical rebound), with activation barriers in the order Fe > Mn > Co for both steps. The trend is consistent with the anticipated metal-nitrene electrophilicities (first step) and ease of reduction from M(III) to M(II) (second step), hence highlighting the dominant role of EA in both steps of styrene aziridination (aliphatic olefins do not follow the same trend for the second step).

The representative case of L⁴Co, however, presents a conundrum, inasmuch as its reactivity in terms of styrene aziridination yields is comparable to that provided by L⁸Co. More importantly, the rate of product buildup in the first hour of the reaction mediated by L⁴Co is superior to that of L⁸Co (Figure 8). These results are difficult to reconcile for a reagent such as L⁴Co, whose Co(II/III) couple is cathodically shifted by 500 mV versus that of L⁸Co, and its nitrene derivative [L⁴Co]NTs is computed to have a lower global electrophilicity index (GEI) than that of [L⁸Co]NTs, in agreement with the electronic nature of the CMe₃ and CF₃ substituents. In addition, the enhanced reactivity of L⁴Co deviates from that of the corresponding L⁴Mn and L⁴Fe reagents, which exhibit significantly lower yields (and rates) in styrene aziridinations by comparison to the L⁸Mn and L⁸Fe analogues, in line with their electrophilic characteristics.

Mechanistic analysis of the operation of L⁴Co in styrene aziridination indicate that both polar (σ_mb) and spin-delocalization (σ_JJ*) parameters are needed to fit Hammett plots, suggesting that both modest positive charge buildup and radical stabilization participate in the turnover-limiting step. The unexpected preponderance of the polar effect for L⁴Co by comparison to L⁸Co can be traced both in the slightly higher values of absolute ρ_mb (−0.58) and relative |ρ_mb/ρ_JJ*| (1.17) than those observed for L⁸Co (ρ_mb = −0.56, |ρ_mb/ρ_JJ*| = 1.0). The secondary KIE values obtained from the competition between styrene and selectively deuterated styrene in aziridinations confirm that both catalysts operate via an initial, turnover-limiting N–C_b bond formation step, but also indicate that the L⁴Co mediated pathway incorporates more significant rehybridization of the β-styrenyl carbon in the transition state, hence placing this TS energetically closer to the resulting radical intermediate [L⁴Co]N(Ts)–CH₂–^•CH₂Ph. Moreover, the ring-closing step (radical rebound) seems to operate via a miniscule energy barrier for both L⁴Co and L⁸Co styrene aziridinations, but the one for L⁴Co is even more suppressed than that for L⁸Co, as judged by the superior retention of stereochemistry in the aziridination of the sensitive substrate cis-β-styrene. This runs counter to what is usually the main driving force for the aziridine-ring closure of styrene, namely the ease of reduction from Co(III) to Co(II).^22g,44,65 All these mechanistic observations would have been perfectly in line, had the supporting L⁴ ligand been more electron-withdrawing than L⁸.

DFT calculations on the electronic and geometric disposition of the presumptive [L⁴Co]NTs vis-à-vis the previously explored [L⁸Co]NTs highlight how a small ligand modification can result in a major electronic rearrangement. First, the ground state of [L⁴Co]NTs is computed to be the quartet (S = 3/2), positioned slightly underneath the sextet (S = 5/2). The sextet is the clear ground state of [L⁸Co]NTs, presumably due to the weaker ligand field provided by the L⁸ ligand. Geometrically, both cobalt nitrenes are quite similar, their most outstanding feature being the elongation of one of the equatorial N residues to a noncoordinating position. However, spin-density calculations reveal that the noncoordinating arm of [L⁴Co]NTs is one-electron oxidized, whereas the corresponding arm of [L⁸Co]NTs is redox innocent. The single-electron distribution of the resulting N-aryl amidyl radical in [L⁴Co]NTs is spread throughout the noncoordinating arm, with almost half of the spin density being localized on the N atom. Apparently, the electron withdrawing CF₃ residue protects the noncoordinating arm of [L⁸Co]NTs from a similar one-electron oxidation. The overall electronic picture for the ground state of the two cobalt nitrenes is schematically summarized in Figure 12. As noted above, the [L⁴Co]NTs sextet (α-spin on the N-aryl amidyl radical) is calculated to be only 0.1 kcal/mol higher in free energy relative to the quartet.

Figure 12.

Schematic distribution of spins in the ground state of [L⁴Co]NTs and [L⁸Co]NTs.

Whereas a definitive justification for the higher reactivity of L⁴Co, in spite of lower electrophilicity, versus L⁸Co cannot be provided at the present time, the following observations should be taken into account:

1. Although it cannot be excluded, it is deemed rather unlikely that the ease of formation of the cobalt-nitrene itself (presumably favoring [L⁴Co]NTs) will be a contributing factor, since our previous calculations for the reaction of [L⁸M^II] (M = Mn, Fe, Co) and PhINTs indicate almost instantaneous generation of [L⁸M]NTs. Rate-limiting metal-nitrene formation is more common with organic azides (RN₃).⁶⁶

2. The fact that L⁴M (M = Mn, Fe), as well as a wide range of other Mn(II) reagents, exhibit reactivities consistent with the EA criterion, whereas L⁴Co and other Co(II) reagents demonstrate deviations, suggest that ligand-centered contributions to the overall oxidizing ability of the reagent may enable more favorable reactivity channels. Indeed, [LCo^III]NTs is more likely to store oxidizing equivalents on ligand residues, as inferred by the cyclic voltammograms of the LCo^II reagents, and anticipated due to the superior oxidizing power of Co(III) versus Mn(III) or Fe(III). Among other possibilities, N-aryl amidyl radicals are known to add to alkenes, at least intramolecularly,⁶⁷ and more electrophilic N-aryl sulfonamidyl radicals can even add intermolecularly.⁶⁸ Although they are not expected to outcompete the metal-bound nitrene radical, they might offer stabilizing interactions not yet realized. On the other hand, the similarity of the Hammett parameters for L⁴Co and L⁸Co suggests that the electronic differences in the ground states of [L⁴Co]NTs and [L⁸Co]NTs may have only a small effect on their reactivities, but this point requires further elaboration once more information is available for the corresponding transition states.

3. Multiple spin-state reactivity channels,^65c,69 such as those offered by the almost isoenergetic quartet and sextet states of [L⁴Co]NTs, may afford enhanced reactivity profiles in aziridinations⁷⁰ by comparison to a potentially single spin-state operation by the [L⁸Co]NTs sextet.

4. London dispersion (LD) interactions applying intramolecularly between highly polarizable alkyl substituents (also known as σ–σ interactions) are now well established stabilizing forces of sterically congested molecules in solution, by means of favorable enthalpic contributions.⁷¹ The tert-Bu group and other conformationally rigid alkyl groups (flexible alkyl groups have an unfavorable entropic impact)⁷² have been credited as “dispersion energy donors”,⁷³ and deemed responsible for stabilizing many highly congested organic and inorganic compounds.^71–74 More importantly, LD forces have started receiving recognition as contributors to observed chemical reactivity and catalytic outcomes.⁷⁵ In the more tight reaction cavity of cobalt reagents, the stabilization offered by tert-Bu groups via LD interactions can play a significant role, as already noted in our initial dispersion-corrected DFT calculations. Interestingly, the L⁷Co reagent, which carries the less polarizable i-Pr substituent, demonstrates lower aziridination rates, not unlike those of L⁸Co.

Future experimental and computational research will seek to disentangle and quantify the factors contributing to the enhancement of catalytic reactivity above and beyond the underlying electrophilic character of the active oxidant, and further explore whether reagents with other rigid alkyl substituents can be superior mediators of nitrene-transfer chemistry.