Stability and C–H Bond Activation Reactions of Palladium(I) and Platinum(I) Metalloradicals: Carbon-to-Metal H-Atom Transfer and an Organometallic Radical Rebound Mechanism

One-electron oxidation of palladium(0) and platinum(0) bis(phosphine) complexes enables isolation of a homologous series of linear d9 metalloradicals of the form [M(PR3)2]+ (M = Pd, Pt; R = tBu, Ad), which are stable in 1,2-difluorobenzene (DFB) solution for >1 day at room temperature when partnered with the weakly coordinating [BArF4]− (ArF = 3,5-(CF3)2C6H3) counterion. The metalloradicals exhibit reduced stability in THF, decreasing in the order palladium(I) > platinum(I) and PAd3 > PtBu3, especially in the case of [Pt(PtBu3)2]+, which is converted into a 1:1 mixture of the platinum(II) complexes [Pt(PtBu2CMe2CH2)(PtBu3)]+ and [Pt(PtBu3)2H]+ upon dissolution at room temperature. Cyclometalation of [Pt(PtBu3)2]+ can also be induced by reaction with the 2,4,6-tri-tert-butylphenoxyl radical in DFB, and a common radical rebound mechanism involving carbon-to-metal H-atom transfer and formation of an intermediate platinum(III) hydride complex, [Pt(PtBu2CMe2CH2)H(PtBu3)]+, has been substantiated by computational analysis. Radical C–H bond oxidative addition is correlated with the resulting MII–H bond dissociation energy (M = Pt > Pd), and reactions of the metalloradicals with 9,10-dihydroanthracene in DFB at room temperature provide experimental evidence for the proposed C–H bond activation manifold in the case of platinum, although conversion into platinum(II) hydride derivatives is considerably faster for [Pt(PtBu3)2]+ (t1/2 = 1.2 h) than [Pt(PAd3)2]+ (t1/2 ∼ 40 days).


■ INTRODUCTION
With many applications in synthetic organic chemistry, 1 the development of methods for enacting the cleavage of C(sp 3 )−H bonds is an important facet of contemporary organometallic chemistry. 2 Building on pioneering work by Bergman and Graham, 2,3 the activation of these robust and nonpolar σ-bonds by concerted oxidative addition to electron-rich, low-valent platinum group metals is a well-established and exploited mechanism. These reactions proceed via transient three-centertwo-electron M−H−C adducts and involve +2 changes in the formal oxidation state of the metal (ΔOS = +2; Scheme 1). 4 Other distinct manifolds include electrophilic activation (ΔOS = 0), σ-bond metathesis (ΔOS = 0), 1,2-addition across polar metal−ligand multiple bonds (ΔOS = 0), and radical oxidative addition (ΔOS = +1). 2,5 The last is typically associated with the homolysis of appreciably polar σ-bonds by metalloradicals and is an underdeveloped C(sp 3 )−H bond activation strategy. The most long-standing precedent emerged from Wayland's work with rhodium(II) porphyrins in the early 1990s, in which a mechanism involving addition of C(sp 3 )−H bonds across two metalloradicals was established experimentally (Scheme 1). 6 Related bimetallic reactivity has also been invoked in the cyclometalation of a triphenylphosphine-ligated rhodium(II) metalloradical and allylic C(sp 3 )−H bond activation reactions of M II (cyclooctadiene) complexes (M = Rh, Ir). 7 Carbon-to-metal H-atom transfer reactions of a photochemically generated osmium(I) cyclopentadienyl metalloradical with allylic and benzylic substrates is a more recent and notable precedent (Scheme 1). 8 Low-valent paramagnetic derivatives of palladium and platinum are intriguing candidates to participate in radical oxidative addition reactions; however, the chemistry of complexes of this nature is significantly underdeveloped, especially with reference to the advances being made with d 9complexes of nickel. 9−11 Mononuclear palladium(I) and platinum(I) complexes have been invoked as transient intermediates or generated electrochemically in situ, 12,13 but only a handful have been isolated to date. 14−17 Instead adoption of the formal +1 oxidation state for palladium and platinum is almost exclusively limited to bimetallic adducts where formation of a metal−metal bond confers a closed-shell electronic configuration. 18 Building upon work by Stalke, Frenking, Roesky, and coworkers, who examined the electrochemical oxidation of twocoordinate cyclic alkyl(amino) carbene complexes of palladium(0) and platinum (0), 12 15 Pairing the platinum(I) metalloradical with the weakly coordinating [HCB 11 Cl 11 ] − counterion curbs onward reactivity but did not prevent complete conversion of 4 + into a 1:1 mixture of 5 + and 6 + , which occurred within 48  Under certain conditions, we have since observed onward reactivity of 3[PF 6 ] that results in formation of [Pd-(PtBu 2 CMe 2 CH 2 )(PtBu 3 )][PF 6 ] (7), most notably upon exposure of the palladium(I) metalloradical to air. 19 Similar reactivity has recently been noted by Mirica and co-workers while exploring the use of dithiapyridinophane-ligated palladium(I) complexes in Kumada cross-coupling reactions (Scheme 2). 16 Both reactions invoke radical oxidative addition of a C(sp 3 )−H bond but, compared to cyclometalation of 4 + , are less well defined.
We herein present further findings from our work exploring the chemistry of two coordinate palladium(I) and platinum(I) metalloradicals, 20 focused on uncovering the mechanism of C(sp 3 )−H bond activation by late transition metal metalloradicals. Examples with enhanced solution stability are described along with a detailed computational examination of different radical-based mechanisms of C(sp 3 )−H bond cyclometalation and other onward reactivity of 4 + . coordinating anion finds widespread utility for the stabilization of reactive and low-coordinate metal cations, 21 while its use is practically convenient, as the sodium salt is commercially available. Straightforward synthetic procedures that have been optimized for the isolation of solvent-free anhydrous M[BAr F 4 ] (M = Li, Na, K) have also been reported recently. 22 To probe the role of solvent on the stability of 3 + and 4 + , we began by re-evaluating the oxidation of 1 and 2 by cyclic voltammetry using [nBu 4 N][BAr F 4 ] as the electrolyte and DFB and tetrahydrofuran (THF) as solvents (Figure 1). Quasireversible one-electron oxidation was observed in all cases and, noting that considerable variance is to be expected when changing the electrolyte and solvent, 23  were extensively characterized in DFB and unsurprisingly show directly comparable spectroscopic and electrochemical signatures to those previously reported for 3 + and 4 + . 14,15 Most notably, deep blue 3[BAr F 4 ] is characterized by a singlet EPR resonance with axial g-tensor, g ⊥ = 2.343 and g ∥ = 1.978, superimposed with a lower intensity sextet arising from isotropic hyperfine coupling to 105 Pd (I = 5/2, 22% abundance, a = 25.2 mT; no superhyperfine coupling to 31 P was evident; DFB glass @ 200 K), but a 31 P NMR resonance could not be observed between +500 and −500 ppm. In contrast, for green 4[BAr F 4 ] a paramagnetically shifted 31 P NMR resonance could be located at δ −213.9, while an EPR signal could not be detected down to 100 K (DFB glass). The UV−vis spectrum of 4 + has not previously been reported and is most remarkable for a sharp band at 306 nm (ε = 700 M −1 ·cm −1 ) ascribed to a metalcentered transition. 27 Further details, including analysis by timedependent density functional theory (TD-DFT) calculations, are provided in the Supporting Information.
The solid-state structures of the new metalloradicals have also been determined and are isomorphic, crystallizing in the cubic P2 1 3 space group with the P−M−P vector lying along a 3-fold rotation axis (M = Pt, Figure 2 Computational Evaluation of Radical Oxidative Addition Pathways. To further interrogate the mechanism associated with C(sp 3 )−H bond cyclometalation of 4 + , we turned to unrestricted DFT calculations to analyze the viability of possible reaction pathways ( Figure 3). Following benchmarking, geometries were optimized in the gas phase using the PBEh-3c composite method, 30 and single-point energies were calculated at the B2PLYP-D3(BJ)/def-TZVPP 31,32 level of theory with corrections included for London dispersion and solvation effects. 33,34 The prospect for Wayland-like bimetallic radical C−H bond oxidative activation was first examined using an antiferromagnetically spin-polarized four-centered transition state of the form Pt(↓)···C(↑)···H(↓)···Pt(↑) derived from two equivalents of 4 + and producing 5 + and 6 + in one step. 35 The calculations indicate that this intermolecular process is associated with a prohibitively high activation barrier of ΔG ⧧ 298K = 40.0 kcal·mol −1 . The bulky PtBu 3 ancillary ligands appear to encumber the approach of the two metal centers, and the hydridic character of the transition state suggests homolysis of the C−H bond occurs with a significant degree of asymmetry (Pt···C = 3.30 vs 2.04 Å in 5 + ; H···Pt = 1.62 vs 1.50 Å in 6 + ; Pt···Pt = 6.19 Å). With the former in mind, and in an attempt to reconcile the large solvent dependence of the reaction, the possibility for ligand exchange with THF to generate the less bulky metalloradical [Pt(PtBu 3 )-(THF)] + as the H-atom acceptor was also considered. In this scenario, cyclometalation of 4 + occurs with more symmetric homolysis of the C−H bond (Pt···C = 2.61 Å; H···Pt = 1.63 Å) and a reduced activation barrier of ΔG ⧧ 298K = 15.4 kcal·mol −1 , but initial substitution of PtBu 3 to form [Pt(PtBu 3 )(THF)] + renders the overall process energetically inaccessible (ΔG ⧧ 298K = 47.2 kcal·mol −1 ). 36 Likewise, conversion of 4 + into 5 + by reaction with • OMes* via a four-centered transition state of the form Pt(↓)···C(↑)···H(↓)···O(↑) can be ruled out on the basis of a large activation barrier of ΔG ⧧ 298K = 33.5 kcal·mol −1 . More promisingly, intramolecular cyclometalation of 4 + , r e s u l t i n g i n t h e p l a t i n u m ( I I I ) m e t a l l o r a d i c a l [Pt(PtBu 2 CMe 2 CH 2 )H(PtBu 3 )] + (9), is calculated to be a moderately endergonic process (ΔG 298K = +9.4 kcal·mol −1 ). A pathway commencing with concerted C−H bond oxidative addition (ΔG ⧧ 298K = 21.3 kcal·mol −1 ) to the metalloradical can be located between 4 + and 9, but our calculations suggest stepwise insertion into the C−H bond is considerably more favorable with an activation barrier of only ΔG ⧧ 298K = 18.3 kcal· mol −1 (Figure 3). The latter commences with carbon-to-metal H-atom transfer (ΔG ⧧ 298K = 18.0 kcal·mol −1 ) and culminates in the formation of 9 following combination of the resulting pendent C-centered radical with the platinum(II) center in [Pt(PtBu 2 CMe 2 CH 2 • )H(PtBu 3 )] + (8). Radical rebound sequences of this nature were initially considered, but subsequently ruled out, as part of early mechanistic work on C(sp 3 )−H bond activation reactions of diamagnetic iridium(I) cyclopentadienyl complexes by Janowicz and Bergman. 37 There are, however, strong parallels with the established catalytic action of metal-oxo-based enzymes 38 and experimental precedent for the formation of mononuclear Pt(III) complexes. 39 Consistent with the higher relative solution stability of 3 + observed, the activation barrier calculated for intramolecular carbon-to-metal H-atom transfer is considerably larger (ΔG ⧧ 298K = 33.5 kcal·mol −1 ) than for 4 + . This difference is . Anisotropic displacement ellipsoids drawn at 30% probability; hydrogen atoms and anion omitted for clarity. Symmetry equivalent phosphine substituents are generated using the operations: 3 / 2 −z, 1−x, 1 / 2 +y and 1−y, − 1 / 2 +z, 3 / 2 −x. Experimentally observed conversion of 4 + into 5 + by reaction with • OMes* at room temperature can be reconciled by direct H-atom abstraction from 9 ( Figure 4). Noting the challenges associated with accurately predicting the entropic contributions of bimolecular transition states using static DFT calculations, 40 the calculated activation barrier (ΔG ⧧ 298K = 24.4, ΔH ⧧ = 5.6 kcal·mol −1 vs 4 + ) is consistent with the suggested role of • OMes* as an H-atom trap. 41 In the case of the solvent-induced formation of a 1:1 mixture of 5 + and 6 + from 4 + , we propose a reaction sequence commencing with solvent-mediated deprotonation of 9 to give neutral platinum(I) cyclometalated complex [Pt(PtBu 2 CMe 2 CH 2 )(PtBu 3 )] (10) (Figure 4; solvent modeled as a THF dimer). Oxidation of 10 by 4 + would thereafter give 5 + , with the reduced product 2 capturing the proton to afford 6 + . This suggestion is consistent with the product stoichiometry, absence of H/D exchange when conducted in d 8 -THF, known reduction potential of 5 + (E 1/2 = −1.90 V relative to Fc/Fc + ), 14 and synthetic procedures used for preparing 5 + and 6 + from 2 (Scheme 3). Computational analysis suggests that deprotonation of 9 is the rate-determining step, conferring an overall activation barrier of ΔG ⧧ 298K = 32.0 kcal· mol −1 with respect to 4 + . When excess THF is factored in, this barrier is lowered to ΔG ⧧ 298K (20 mM 4 + in THF) = 24.3 kcal· mol −1 .
Isolation and Stability of Tri(1-adamantyl)phosphine-Ligated Metalloradicals. Recognizing the requirement for Catom planarization in carbon-to-metal H-atom transfer reactions, we speculated that phosphine ligands with caged substituents would be less susceptible to cyclometalation and therefore confer enhanced metalloradical stability in solution. Tri(1-adamantyl)phosphine (PAd 3 ) is well suited to test this conjecture and is notable for a similar steric profile to PtBu 3 about the metal (%V bur = 40.5 cf. 40.0%), but appreciably stronger donor characteristics (TEP = 2052.1, cf. 2056.1 cm −1 ). 42 The bis(phosphine) Au(I) complex [Au(PAd 3 ) 2 ] + is a notable diamagnetic derivative, 43 and the nickel metalloradical [Ni(PAd 3 ) 2 ] + has recently been reported. 10 Exploiting   toluene (>80% isolated yields, Scheme 5). These neutral complexes are highly insoluble in common organic solvents (including CH 2 Cl 2 , THF, PhMe, and DFB), presumably resulting from abnormally strong intermolecular dispersion forces, 45  Very broad paramagnetically shifted adamantyl resonances are observed by 1 H NMR spectroscopy for both new metalloradicals in DFB solution. No 31 P NMR resonance was observed for 13 between +500 and −500 ppm, but a broad signal can be identified for 14 at δ −252.4 (fwhm = 120 Hz), upfield of that observed for 4[BAr F 4 ] (δ −213.9). Analysis of 13 by EPR spectroscopy confirms the assignment as a metal-centered radical, with observation of a singlet resonance arising from an axial g-tensor, g ⊥ = 2.333 and g ∥ = 1.979, that is superimposed with a lower intensity sextet arising from isotropic hyperfine coupling to 105 Pd (I = 5/2, 22% abundance, a = 24.6 mT; DFB glass @ 200 K). The magnitude of the hyperfine coupling constant is similar to that recorded for 3[BAr F 4 ], implying only small changes in the character of the singly occupied molecular orbital. Weak shoulders on the 338 mT hyperfine line could be an indication of an unresolved superhyperfine interaction, but this was not modeled. As for 4[BAr F 4 ], no metal-centered EPR spectrum was observed for 14 down to 100 K (DFB glass). The UV−vis spectra of the new metalloradicals (blue, 13; green, 14) are comparable to those of the respective PtBu 3 analogues, with the main bands slightly red-shifted.
The formulations of 13 and 14 have been corroborated in the solid state by X-ray diffraction. The structures were obtained using samples recrystallized from DFB/hexane and are isomorphic (monoclinic C2/c) with no crystallographically imposed cation or anion symmetry (M = Pt, Figure 5). 48 The cations are well-ordered and adopt near-ideal linear geometries (P2−M1−P3 = 178.94(6)°, 13; 179.29(9)°, 14) and eclipsed phosphine conformations, with the dihedral angles < 11°. Isostructural gold(I) and nickel(I) complexes also adopt this geometry in the solid state. 10 The experimental trends are well reproduced computationally using a direct H-atom transfer mechanism between the metalloradicals and 9,10-dihydroanthracene (Table 1). Reac-tions of the palladium complexes are characterized by an activation barrier ca. 12 kcal·mol −1 larger than their platinum counterparts. This difference is inversely correlated with the calculated M II −H bond dissociation energies, with those of palladium (64.4/64.1 kcal·mol −1 ) substantially lower than the bond dissociation energy of the C(sp 3 )−H bonds in 9,10dihydroanthracene. The phosphine ligand substituent has a less pronounced effect, but the overall reaction kinetics and thermodynamics are less favorable for the PAd 3 -ligated metalloradicals. This difference appears to be steric in origin, with the transition states in this case characterized by smaller distortions of the P−M−P angles from linearity. Cyclometalation of 4 + and generation of 5 + can also be achieved by reaction with the 2,4,6-tri-tert-butylphenoxyl radical ( • OMes*) in DFB and in situ through reaction with 2,6-di-tertbutyl-4-methylpyridine and ferrocenium in THF. Computational analysis of the onward reactivity of 4 + rules out mechanisms involving Wayland-like four-center-two-electron and concerted C−H bond oxidative activation, but an energetically feasible pathway involving carbon-to-metal Hatom transfer, combination of the resulting C-centered radical with the metal, and formation of the intermediate platinum(III) hydride complex [Pt(PtBu 2 CMe 2 CH 2 )H(PtBu 3 )] + (9) was identified (Scheme 7). This organometallic radical rebound mechanism reconciles the experimental observations, with conversion of 9 into 5 + proposed to proceed by H-atom abstraction by • OMes* or sequentially by solvent-mediated deprotonation of 9 to the form neutral platinum(I) complex [Pt(PtBu 2 CMe 2 CH 2 )(PtBu 3 )] (10), which can be oxidized to 5 + by 4 + or ferrocenium. The platinum(0) byproduct 2 or an added base thereafter mops up the solvated proton, affording platinum(II) complex 6 + in the former case. Carbon-to-metal Hatom transfer is less accessible for the palladium(I) congener [Pd(PtBu 3 ) 2 ] + (3 + ), due to an inherently weaker M II −H bond, and the PAd 3 -ligated metalloradicals, where the caged phosphine substituents encumber C-atom planarization. Reactions of the metalloradicals with 9,10-dihydroanthracene in DFB at room temperature provide direct experimental evidence for carbon-to-metal H-atom transfer in the case of the platinum, although conversion into platinum(II) hydride derivatives is considerably faster for [Pt(PtBu 3 ) 2 ] + (4 + ) (t 1/2 = 1.2 h) than [Pt(PAd 3 ) 2 ] + (14) (t 1/2 ∼ 40 days). Paralleling their solution stability, the calculated barriers decrease in the order palladium(I) ≫ platinum(I) and PAd 3 > PtBu 3 , reflecting inherent periodic trends in M II −H bond strength and steric constraints, respectively.
These findings demonstrate (a) the synthetic accessibly of low-valent paramagnetic palladium and platinum complexes and (b) the ability of late transition metal-based radicals of this nature to activate C−H bonds in a manner that is mechanistically distinct from their more widely investigated diamagnetic counterparts. Insights of this nature may help inspire the development of new and more effective catalysts for the functionalization of C−H bonds in organic synthesis. ■ EXPERIMENTAL SECTION 1. General Methods. All manipulations were performed under an atmosphere of argon using Schlenk and glovebox techniques unless otherwise stated. Glassware was oven-dried at 150°C overnight and flame-dried under vacuum prior to use. Molecular sieves were activated by heating at 300°C in vacuo overnight. DFB was predried over Al 2 O 3 , distilled from calcium hydride, and dried over two successive batches of 3 Å molecular sieves. 24 THF was vacuum distilled from sodium/ benzophenone and stored over 3 Å molecular sieves. CD 2 Cl 2 was freeze−pump−thaw degassed and dried over 3 Å molecular sieves. d 8 -THF was dried over sodium, vacuum distilled, freeze−pump−thaw degassed, and stored over molecular sieves (3 Å). All other anhydrous solvents were purchased from Acros or Sigma-Aldrich, freeze−pump− thaw degassed, and stored over 3 Å molecular sieves. [ 52 the 2,4,6-tri-tert-butylphenoxyl radical, 53 and PAd 3 42 were prepared using literature procedures. 2,6-Di-tert-butyl-4-methylpyridine was purchased from Sigma-Aldrich and used as received.
Cyclic voltammetry (CV) experiments were carried out in an inert atmosphere glovebox under argon using a PalmSens EmStat3+ Blue potentiostat and a three-electrode setup comprising a glassy carbon (CH Instruments, 3.0 mm diameter) working electrode (WE), coiled platinum wire counter electrode (CE), and silver wire quasi-reference electrode (RE). All potentials are calibrated to the ferrocene/ ferrocenium (Fc/Fc + ) redox couple, which was used as an internal standard. The half-wave potentials, E 1/2 , were determined from E 1/2 = (E P red + E P ox )/2, where E P red and E P ox are the reduction and oxidation peak potential values, respectively. For the irreversible electrochemical process, the half-peak potential, E P/2 , was used as an approximation for E 1/2 . 47 NMR spectra were recorded on Bruker spectrometers under argon at 298K unless otherwise stated. Chemical shifts are quoted in ppm, and coupling constants in Hz. Virtual coupling constants are reported as the separation between the first and third lines. 54 NMR spectra in DFB and THF were recorded using an internal capillary of C 6 D 6 . 24 EPR spectra were acquired on a Bruker EMX spectrometer using a Bruker High Sensitivity cavity (ER 4119 HS). Samples were cooled by nitrogen gas flow through a standard quartz insert from a nitrogen evaporator with a B-VT 2000 temperature control unit. To limit the dielectric loss arising from the solvent, all samples were contained in 2.2 mm i.d. quartz tubes (Wilmad 705-SQ), and the quartz insert was removed for roomtemperature operation. The reported g-factors are referenced to a DPPH standard (g = 2.0036(3)). 55 EPR data were obtained using a 200 mW microwave power at 9.51 GHz, with a 0.5 mT field modulation at 100 kHz. The simulation was performed using the pepper routine in EasySpin 56 with the data fitted directly using a genetic algorithm in the esfit routine. Due to the high spectral width, it was necessary to remove a broad nonlinear cavity baseline prior to fitting by subtraction of a smoothed cubic spline derived from the experimental cavity background recorded under identical conditions. Convolution broadening was applied in the simulations with a combination of Lorentzian and Gaussian line width components necessary to adequately reproduce the observed spectral shape. With this phenomenological line shape model there was no further improvement in the quality of fit from allowing a nonisotropic hyperfine interaction. UV−vis spectra were recorded on an Agilent Cary 3500 UV−vis spectrometer compact Peltier system.
High-resolution (HR) ESI-MS analyses were recorded on a Bruker Maxis Impact instrument. Microanalyses were performed at the London Metropolitan University by Stephen Boyer.   J CB = 3, Ar F ), 124.9 (q, 1 J FC = 273, Ar F ), 117.6 (sept., 3 J FC = 4, Ar F ), 39.0 (vt, J PC = 12, tBu{C}), 31.6 (vt, J PC = 5, tBu{CH 3 }). 31   16. Computational Details. All electronic structure calculations presented in this paper were carried out using the ORCA 5.0.3 program package. 57 To find energetically most favorable conformers, initial conformer searches of selected complexes were conducted using the automated crest approach, 58 which employs the efficient semiempirical extended tight binding method (GFN2-xTB) with a specially adapted implicit solvation model (ALPB). 59 Unconstrained geometry optimizations in C 1 symmetry and analytical (or numerical for the largest systems) frequency calculations of all compounds were carried out at the DFT level, using the PBEh-3c composite method. 30 The orbitals are expanded in a modified valence double-ζ Gaussian basis corresponding to the Ahlrichs-type def2-mSVP set, in conjunction with the def2/J auxiliary basis set for the RI approximation to the Coulomb term. 60 The calculations utilized the def2-ECP for Pd (replacing 28 core electrons) and Pt (replacing 60 core electrons). 61 To account for inter-and intramolecular basis set superposition error (BSSE) and long-range London dispersion effects, the geometrical counterpoise correction (gCP) 62 and atom-pairwise DFT-D3 (Becke−Johnson damping) schemes are utilized. 33 All optimized stationary points were characterized by analysis of their analytical second derivatives, with minima having only positive eigenvalues and transition states having one imaginary eigenvalue. The nature of transition states was confirmed via intrinsic reaction coordinate (IRC) calculations in both forward and reverse direction of the reaction coordinate. 63 Subsequent geometry optimizations of the IRC end points yielded the nearest minima linked by a transition state. The frequency calculations also provided thermal and entropic corrections to the total energy in the gas phase at T = 298.15 K and p = 1 atm within the quasi rigid-rotor/harmonic oscillator (QRRHO) approximation. 64 Open-shell singlet states (M S = 0) corresponding to antiferromagnetically coupled metal centers were modeled with the spin-unrestricted broken-symmetry (BS) formal-ism. 65 The FlipSpin feature of ORCA was used to generate initial guesses for the BS calculations. Geometries with these states were fully optimized, and convergence to the desired BS solution was confirmed by inspection of magnetic orbitals, spin populations, and the expectation value of the ⟨S 2 ⟩ operator. The energies of the BS states were used without spin projection. Single-point energies were computed using the B2PLYP-D3(BJ) double-hybrid functional 31 (including Grimme's D3 atom-pairwise dispersion correction and Becke−Johnson damping) in combination with the def2-TZVPP basis set. 32 Full details of calculations are provided in the Supporting Information. Effects due to the presence of a solvent were treated implicitly with a conductor-like polarizable continuum (CPCM) and Truhlar's SMD model. 34 Solvent parameters corresponded to those of tetrahydrofuran (ε = 7.4, refractive index = 1.000), or, in the absence of defined parameters for DFB solvent, default SMD parameters were selected for fluorobenzene and the dielectric constant adjusted to that of DFB (ε = 13.4, refractive index = 1.443). Geometries were visualized using the ChemCraft software package. 66 ■ ASSOCIATED CONTENT

* sı Supporting Information
The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.3c04167. NMR, EPR, UV−vis, and ESI-MS spectra of complexes and selected reactions; additional electrochemical data and analysis; further computational details, data, and analysis (PDF) Optimized geometries of species examined computationally (XYZ) Animations of selected transition states (MP4)

Accession Codes
CCDC 2257962−2257967 contain the supplementary crystallographic data for this paper. These data can be obtained free of charge via www.ccdc.cam.ac.uk/data_request/cif, or by emailing data_request@ccdc.cam.ac.uk, or by contacting The Cambridge Crystallographic Data Centre, 12 Union Road, Cambridge CB2 1EZ, UK; fax: +44 1223 336033.