Introduction

In the realm of s-block metal amide chemistry, 2,2,6,6-tetramethylpiperidide (TMP) occupies a privileged position on account of its widespread applications in selective deprotonative metalalation reactions.¹ Exemplified by LiTMP and its enhanced Turbo-Hauser base version TMPMgCl.LiCl, these amides combine low nucleophilicity with high Brønsted basicity, ideal for selective cleavage of C−H bonds of aromatic substrates.² The special electronics and architecture of the cyclic TMP group, featuring two methyl substituents on each alpha carbon atoms, not only amplifies its basicity but also significantly increases its steric demand.³

Seminal work by Knochel, Mulvey, and Uchiyama amongst others has demonstrated the exceptional basicity of TMP when in bimetallic environments such as alkali-metal zincates, magnesiates and aluminates.⁴ The introduction of bimetallic cooperation in these systems not only enhances reactivity but also imparts unique selectivity that sets them apart from their monometallic counterparts.⁵ TMP seems to be complicit in this special bimetallic chemistry, since when replaced by other amide groups, diminished reactivities are often observed.⁶

In contrast to these extensive main-group executed applications, TMP is much less visible in transition metal complexes, though a few sporadic studies of Fe and Co studies have hinted at their potential as arene metalators. Thus, Knochel has reported that in situ prepared Fe(TMP)₂.2MgCl₂.4LiCl, made by salt metathesis of FeCl₂.LiCl with 2 equivalents of TMPMgCl.LiCl, can metalate functionalized arenes to generate putative FeAr₂ species that subsequently undergo C−C bond formation processes (Figure 1a).⁷ However the identity of the Fe(II) amide and bis(aryl) intermediates remains blurred as do the roles played by LiCl or MgCl₂ in facilitating the possible formation of ferrate species. Similarly, Mongin has prepared in situ Co(TMP)₂ by treating CoBr₂ with LiTMP in THF which in combination with LiTMP can deprotonate a range of substituted anisoles (Figure 1b), though a large excess of base is needed. Moreover, neither the constitution of Co(TMP)₂ nor the mechanism(s) involved have been studied.⁸

This lack of fundamental knowledge on the synthesis and constitution(s) of M(TMP)₂ (M=Fe and Co) is particularly surprising, especially in light of the well-established reports of Fe and Co bis(amide) complexes, such as M(HMDS)₂; (HMDS=N(SiMe₃)₂; M=Fe, Co), spanning over six decades.⁹ Significantly, while M(HMDS)₂ complexes have demonstrated their ability to act as versatile precursors of many M(II)/M(I) complexes,¹⁰ and to catalyze key transformations (such as hydrosilylation or alkyne trimerization processes), their ability to promote deprotonative metalation reactions has hitherto been limited to highly acidic substrates such as alcohols or secondary phosphines,¹¹ being completely inert towards C−H arene bonds. Opening a new chapter in this literature, our groups recently found that pairing M(HMDS)₂ with sodium amide NaHMDS produced highly reactive sodium transition(metal) ate complexes capable of promoting regioselective ferration and cobaltation of several fluoroarenes (Figure 1c).¹² Mechanistic studies suggested that initially the fluoroarenes are metalated by the sodium amide followed by fast transmetalation to Fe (or Co) which can occur inter- or intra-molecularly.^{12a, 12c} Evidencing the poor metalating power of the TM amides M(HMDS)₂, no reaction was observed when treated with pentafluorobenzene under harsh reaction conditions.

Filling an important knowledge gap, here we report the synthesis and full spectroscopic, electronic, and structural characterization of the transition metal bis(amides) M(TMP)₂ (M=Fe, Co) and explore their reactivity towards fluoroarenes, uncovering their metalating potential to promote direct low polarity metalations without aid of an alkali-metal (Figure 1d). Supported by DFT calculations, we shed light on the mechanisms involved in these reactions and on the role of additives such as LiCl, NBu₄Cl or Lewis donors such as THF, which not only enhance the metalating power of these new earth abundant transition metal amides but also show polybasic behaviours.

Results and Discussion

Approaching the synthesis of M(TMP)₂ via salt metathesis by treating FeBr₂ or CoCl₂ with two molar equivalents of LiTMP in THF at −30 °C (Figure 2aa, i) led to the formation of orange and green solutions respectively. Removing THF in vacuo, extracting with pentane and filtering through celite yielded a crude product that could be purified via dynamic vacuum distillation giving Fe(TMP)₂ (1^Fe) and Co(TMP)₂ (1^Co) as crystalline solids in respective 64 % and 46 % isolated yields. Interestingly, 1^Fe and 1^Co could also be obtained using trans(amidation) by reacting a 2 : 1 ratio NaTMP and M(HMDS)₂ mixture at −30 °C in pentane (Figure 2aa, ii). While 1^Fe and 1^Co are well soluble in pentane, the byproduct of the reaction NaHMDS forms as a precipitate that can be removed by filtration. [(THF)Co(TMP)₂] (1^Co-THF) could also be prepared via salt-metathesis by recrystallisation in THF (Figure 2a, iii, see Supporting Information for details).

Despite their paramagnetic nature, informative ¹H NMR spectra could be obtained in C₆D₆ for 1^Fe and 1^Co displaying three highly broadened and paramagnetically shifted resonances for the beta, gamma, and Me hydrogens of the TMP ligands at 169.97, 138.88, and 81.28 ppm, respectively, for 1^Fe and at 186.95, 172.63 and 8.08 ppm, respectively, for 1^Co. Molecular structures of 1^Fe and 1^Co-THF were established by X-ray crystallographic studies.¹³ Exhibiting a structure reminiscent of those of M(TMP)₂ (M=Zn, Cd),^{14, 15} solvent-free 1^Fe displays a rare monomeric arrangement (Figure 2aa),¹⁶ featuring an Fe(II) center in a gently bent distorted linear geometry [N1-Fe1-N2 angle, 172.09(14)°] which forms two short Fe−N bonds [1.875(3) and 1.871(3) Å]. The opposing TMP ligands are twisted in order to minimize repulsive steric interactions with a twist angle of 73.05(18)° between the C3−N1−C7 and C12−N2−C16 planes. In addition, four medium to long interactions are discernible between Fe and two α-Me groups from each TMP group [ranging from 2.886(5) to 2.953(3) Å] which may explain the slight deviation from linearity of the N1-Fe1-N2 angle. 1^Co-THF features a distorted trigonal Co(II) center which binds terminally to two TMP groups [Co1−N1, 1.893(3); and Co1−N2, 1.884(3) Å] and one THF molecule. This motif is similar to that reported by Power for [(THF)Co(HMDS)₂]¹⁷ though a slightly more obtuse N−Co−N angle is seen in 1^Co-THF [145.48(13)°] most likely due to the greater steric demands imposed by TMP. As far as we can ascertain, 1^Co-THF represents the first example of a structurally defined complex containing a TMP group bonded to Co.¹⁸

The solid-state variable temperature, variable field (VTVF) SQUID magnetization data for 1^Fe (Figure 2b) and 1^Co (Figure 2c) show large room temperature magnetic moments of μ_eff=5.76 B.M. and 5.64 B.M. It must be noted, however, that both complexes significantly exceed the spin-only values for S=2 and S=3/2 spin systems of 4.90 B.M. and 3.87 B.M., respectively, which is attributed to unquenched orbital angular momentum in linear coordination compounds, resulting in magnetic anisotropy.¹⁹ Regardless, both data sets could be fitted with the spin-only Hamiltonian, giving large anisotropic g-values. For 1^Fe, the best fit obtained for the VTVF DC SQUID data yields g-values of g_⊥=2.24 and g_∥=2.50, and a zero-field splitting parameter D=−33 cm⁻¹ (g_av.=2.33). The magnetic characteristics of 1^Fe are commonly observed in linearly coordinated Fe(II) high-spin compounds.²⁰ Additionally, the ⁵⁷Fe Mössbauer spectrum of 1^Fe (Figure 2dd) displays a notably asymmetric, broadened single line at 77 K. This is likely due to unresolved magnetic hyperfine and quadrupole splitting, which can be observed for other linear Fe(II) high-spin complexes and is rationalized as a result of fast relaxation processes.²¹ As found for comparable linear Fe(I)^{19b, 22} or Co(II) complexes,^{19c, 23} the d⁷ configuration of 1^Co leads to even larger axial anisotropy, reflected by g_⊥=2.34, g_∥=3.46, and D=−141 cm⁻¹ as obtained by fitting the VTVF SQUID magnetization data. Finally, 1^Fe and 1^Co show slow magnetic relaxation at low temperatures (see Supporting Information for details).

We started reactivity studies using pentafluorobenzene and 1,2,4,5-tetrafluorobenzene as model substrates. While their aromatic hydrogens are activated towards deprotonative metalation (in terms of pK_a values)²⁴ their high degree of fluorination makes them more prone to undergo C−F bond activation processes, especially in reactions with low valent transition metal complexes.^{25 1}H NMR monitoring of the reactions of equimolar amounts of 1^Fe and 1^Co with these perfluorinated substrates in C₆D₆ led to the almost instantaneous formation of [{M(TMP)Ar^F}₂] (Ar^F=C₆F₅, 2^M; C₆F₄H, 3^M; M=Fe, Co) in quantitative yields with concomitant release of TMP(H), evidencing their metalating ability towards these substrates. Using pentane as solvent enabled isolation of 2^M and 3^M (M= Fe Co) as crystalline solids (31–92 % yield, see Supporting Information for details).²⁶ Emphasizing the enhanced basicity of these transition metal TMP complexes, their reactivity is in sharp contrast with the complete inertness of M(HMDS)₂ amides towards these substrates, even under forcing refluxing conditions.^{12a, 12c} Importantly, while sodium ferrates (and cobaltates) NaM(HMDS)₃ can promote fluoroarene metalations, it should be noted that these reactions occur by initial sodiation of the substrate followed by fast inter- or intramolecular transmetalation to Fe (or Co) (Figure 1c).

Structural studies revealed the dimeric structures of 2^M and 3^M featuring a central planar four-atom {MNMN} core with TMP bridges, whereas the aryl groups bind terminally to the M^II centers (Figures 3 and S66–S69). For brevity, only the main structural parameters of 2^Fe and 2^Co are discussed here.

Confirming successful M−H exchange, 2^Fe and 2^Co contain a pentafluorophenyl anion bound terminally to either a Fe(II) or a Co(II) center via the metalated ipso C atom (average M−C bond length, 2.043 and 1.991 Å for 2^Fe and 2^Co respectively). These aromatic rings lie orthogonal to the plane of the {MNMN} central core, with the divalent metals exhibiting a distorted trigonal planar geometry. One noticeable feature of these structures is their metal-metal distances [2.5128(4) and 2.4221(5) Å for 2^Fe and 2^Co respectively], which are remarkably short when compared with those reported for other dimeric Fe(II) and Co(II) complexes containing amide or alkyl bridges.²⁷ As far as we can ascertain, 2^M and 3^M represent the first examples of direct ferration and cobaltation of fluoroarenes to be isolated and structurally authenticated. While low-valent Fe and Co complexes have been previously used for functionalization of fluoroarenes, via C−H or C−F bond activation processes, these reactions typically occur with a change in the oxidation state of the metal.^{25, 28} Remarkably, the reactivities displayed by 1^Fe and 1^Co align better with those reported for main-group bases.²⁹

Variable temperature SQUID measurements were recorded for dinuclear 2^Fe and 2^Co. Both complexes show temperature-dependent magnetic moments. At 2 K, dimeric 2^Fe and 2^Co possess magnetic moments of μ_eff=0.34 B.M and μ_eff=0.19 B.M., reaching room temperature moments of 3.30 B.M and 1.95 B.M., respectively. This is far below the calculated magnetic moments of 6.93 B.M. (for Fe(II) high-spin) and 5.47 B.M. (Co(II) high-spin) for two uncoupled metal centers and therefore, suggests antiferromagnetic exchange coupling between the two metal ions. The ⁵⁷Fe Mössbauer spectrum of a solid sample of 2^Fe at 77 K (Figure S101) features a single quadrupole doublet with an isomer shift δ=0.63 mm s⁻¹ and a large quadrupole splitting ΔE_Q=4.22 mm s⁻¹. Consistent with antiferromagnetic coupling, the resonances in the ¹H NMR spectrum for 2^M and 3^M in C₆D₆ are sharp and well resolved although they appear at chemical shifts typical for paramagnetic species.^27c In addition, meaningful ¹⁹F NMR spectra could also be acquired (see Figures S9–S15).

We next pondered if the remaining TMP group present in 2^M and 3^M could facilitate the metalation of a second equivalent of substrate. ¹H NMR monitoring of the reactions showed that in C₆D₆ no further reactivity was observed even under forcing reaction conditions (16 h, 80 °C). However, when switching to more polar d₈-THF, in the case of 2^Fe we could see formation of amine TMP(H) producing bis(aryl) [(THF)₂Fe(C₆F₅)₂] (4^Fe-THF) quantitatively (Figure 3). 4^Fe-THF could also be prepared by reacting 1^Fe with 2 equivalents of pentafluorobenzene using THF as solvent and isolated as a solid in a 98 % yield. Remarkably, 1^Co failed to exhibit polybasicity towards an excess of pentafluorobenzene in THF, even under refluxing conditions, forming exclusively 2^Co. The constitution of 4^Fe-THF was established by X-ray crystallographic studies (Figure 3), revealing a discrete centrosymmetric monomeric motif with Fe in a distorted tetrahedral geometry (bond angles range 116.47(9)–109.33(8)°). The Fe−C bond length [2.0863(17) Å] is slightly elongated to that previously noted in 2^Fe. Interestingly, bis(aryl)iron compounds have previously been proposed to be key intermediates in Fe-catalysed cross-coupling reactions.³⁰ Recent mechanistic studies have accessed these compounds by salt metathesis using ArMgX and Fe(II) or Fe(III) salts.³¹ While Knochel et al. used Fe(TMP)₂ ⋅ 2MgCl₂ ⋅ 4LiCl as a base and suggested the in situ formation of such compounds (see Figure 1a),^[7] 4^Fe-THF stands out as a unique example of a FeAr₂ complex obtained by the direct metalation of an arene. ¹H NMR of 4^Fe-THF in C₆D₆ showed two paramagnetically shifted broadened signals at 42.6 and 16.90 ppm for the ligating THF molecules, whereas in the ¹⁹F NMR spectrum only two signals could be seen at 145.7 and 15.5 ppm (see Supporting Information for details). Magnetic studies of the TMEDA solvated analogue of 4^Fe (4^Fe-TMEDA) were carried out and the data compared with the ones recorded for 1^Fe. While the averaged g-value, obtained from fitting the magnetization data of 4^Fe-TMEDA, is similar, the zero-field splitting parameter is expectedly smaller compared to linear 1^Fe (g_av=2.32, |D|=7 cm⁻¹ for 4^Fe-TMEDA vs. g_av=2.31, D=−32 cm⁻¹ for 1^Fe). Notably, while the magnetization data of 4^Fe-TMEDA could be fitted with an averaged g-value, this was not possible for linear 1^Fe. Further, the zero-field ⁵⁷Fe Mössbauer data, with an isomer shift, δ, of 0.71 mm s⁻¹ and a quadrupole splitting, ΔE_Q, of 1.74 mm s⁻¹, are in accordance with a tetrahedrally coordinated iron center with an S=2 spin state (Figure S99).

Intrigued by the contrasting reactivities of 1^Fe and 1^Co toward pentafluorobenzene, we next conducted density functional theory (DFT) calculations at the ωB97XD level (see Supporting Information for details). As depicted in Figure 3, the iron complex demonstrates the ability to undergo the second C−H metalation, forming 4^Fe-THF. In contrast, the cobalt analogue exclusively yields the dimer 2^Co, ruling out any polybasic behavior.

The crucial role of reversible solvent molecule association in the thermodynamics of the process was experimentally suggested and confirmed by the coordination of solvent molecules in the solid state—either to the cobalt reactant [(THF)Co(TMP)₂] (1^Co-THF) or to the iron product [(THF)₂Fe(C₆F₅)₂] (4^Fe-THF). Our initial speciation DFT studies on the M(TMP)₂ complexes aimed to assess solvent coordination for the formation of the tricoordinate [(THF)M(TMP)₂] species, using the implicit solvation model based on density (SMD, see Supporting Information for details). Endergonic association energies were obtained for both metal centers, i.e. +5.5 and +1.2 kcal mol⁻¹ for iron and cobalt, respectively (Figure S84 and S85). These results contrast with the experimentally measured THF binding constants determined from ¹H NMR titrations (i.e. ^FeK_eq=0.44 M⁻¹ ±0.47 % vs ^CoK_eq=113.61 M⁻¹ ±11.58 %, Table S1),³² supporting a markedly favorable THF association for cobalt. Additionally, significant formation of [(THF)Fe(TMP)₂] (1^Fe-THF) was experimentally observed in THF solution. The substantial discrepancy is closely related to the high concentration effect of THF when acting as a solvent. In such cases, THF molecules surround the organometallic complexes, reducing the entropy penalty associated with their coordination and compromising the accuracy of traditional implicit solvation models.³³ To realistically describe this scenario, we adopted a hybrid solvation method. This method integrates an explicitly modeled THF molecule within the computational framework, alongside implicit solvation corrections calculated via the SMD method, as illustrated in Figure S83. Specifically, the THF molecule is positioned in proximity to the metal complex and near the spectator TMP ligand without direct coordination to the metal, as depicted in Figure 4. This approach effectively cancels the translational contribution to the entropy cost required to bring the THF solvent molecule into the vicinity of the metal centre, thus providing a more precise representation.

Applying this hybrid strategy, THF dissociation from the [(THF)M(TMP)₂] (1^M-THF) species was predicted to be endergonic for both metals, establishing these adducts as resting states of the linear M(TMP)₂ (1^M) complexes in THF solution, consistent with experimental data. Specifically, cobalt exhibited a more disfavored THF dissociation compared to iron (i.e. +1.1 vs +6.7 kcal mol⁻¹ for 1^Fe and 1^Co, respectively).

To rationalize this substantial difference, an activation strain analysis on the 1^M-THF adducts was subsequently performed. This involved defining M(TMP)₂ and THF as molecular fragments and decomposing the relative energy of 1^M-THF into distortion (ΔE_dist) and interaction (ΔE_int) energy terms.³⁴ The analysis revealed the cooperative effects of a lower ΔE_dist term in the cobalt fragment (compared to iron) and stronger ΔE_int between cobalt and THF (Table S6) as decisive factors for the higher preference of Co for THF binding (see Figure 2a).

In terms of metalating reactivity, the dissociation of THF from 1^M-THF creates a vacant coordination on the metal center in 1^M, essential for interaction with the C₆F₅H substrate to form I1^M. The computed C−H metalation mechanism, illustrated in Figure 4, depicts similar structures for the resulting I1^M intermediates with both metals. These structures exhibit weak interactions between the metal center and the aryl ring, with the fluoroarene hydrogen directed towards the nitrogen atom of one of the amide ligands.³⁵ This particular arrangement facilitates C−H metalation via TS1^M with computed energies of +21.1 and +22.4 kcal mol⁻¹ for Fe and Co, respectively—well within the range of energies deemed to be feasible at room temperature. The above activation energies were further refined through microkinetic modelling (MKM) based on experimental observations (see Supporting Information for details), with the resulting values labelled with an asterisk in Figure 4. Notably, the estimated TS1^Co energy of 20.4 kcal mol⁻¹ is consistent with the major formation of the cobalt dimer 2^Co observed experimentally within one hour at 298 K (Figure S93 and Table S7).

In addition, distinctive non-covalent interactions (NCI) between the metal center and one of the ortho-F substituents of the fluoroarene were observed for both TS1^Fe and TS1^Co (see Figure 4 for optimized structures with relevant distances and Figure S86 for NCI analyses), suggesting the role of fluorine as ortho-directing group in the C−H metalation process.

An alternative transition state to TS1^M was modelled maintaining the THF coordinated to the metal center, but it was discarded due to its higher energy (Figure S87). This finding strongly suggests the necessity of THF dissociation for the first metalation, a hypothesis that was experimentally confirmed. Specifically, solutions of the complex [(I^tBu)Co(TMP)₂] (I^tBu=1,3-di-tert-butylimidazol-2-ylidene), prepared in situ, were unreactive towards pentafluorobenzene even under harsh conditions (16 h, 65°C; details in SI). This lack of reactivity is likely attributed to the non-lability of the carbene ligand, precluding the formation of the active linear complex 1^Co, which is crucial for the interaction with the fluoroarene (Figure 4).

Geometry relaxations of the TS1^M structures resulted in the metalated intermediates I2^M characterized by a covalent M−C₆F₅ bond and a protonated TMP(H) ligand coordinated to the metal center. Subsequent DFT calculations predict the substitution of the TMP(H) by a THF molecule to be exergonic for iron and slightly endergonic for Co (i.e. −3.6 and +0.6 kcal mol⁻¹, respectively), yielding the tricoordinate species [(THF)M(C₆F₅)(TMP)] (I3^M).

This species, I3^M, then undergoes dimerization to produce the antiferromagnetically coupled dimer [M(C₆F₅)(μ-TMP)]₂ (2^M) with the concomitant dissociation of THF for both iron and cobalt. The experimental magnetic moment of μ_eff=0.34 B.M. (Figure S106) and the greater stabilization of 2^Fe by −33.5 kcal mol⁻¹, compared to the ferromagnetically coupled dimer in a nonet high-spin state, further support the formation of this dimeric species.

To further refine the understanding of the thermodynamics and kinetics involved in the dimerization step, which encompasses double THF dissociation, MKM was employed. These ligand association/dissociation processes are known to be fast and often diffusion-controlled,³⁶ posing challenges in locating the associated transition states. By integrating DFT and experimental data with MKM (see Supporting Information for complete kinetic model), we estimated the energy barrier for the dimerization process to be of +13.0 kcal mol⁻¹. Given the notable discrepancies in computational results typically obtained from different solvation methods in reactions involving multiple solvent molecules, the energies of both 2^Fe and 2^Co (Figure 4) were also estimated using MKM. These values were found to lie within the boundaries defined by the implicit and hybrid solvation approaches (see Supporting Information for details), providing additional validation.

Importantly, the relative stabilities of I3^M and 2^M provide a basis for understanding the divergent experimental behaviors observed for iron and cobalt regarding the C−H metalation of the second equivalent of pentafluorobenzene. The lower stability of I3^Co coupled with the more favorable formation of the dimer 2^Co compared to the iron analogues, restricts the presence of I3^Co in kinetically efficient concentrations, even in THF solution. In contrast, the reversible 2^Fe⇌I3^Fe equilibrium plays a pivotal role in facilitating the interaction with the second equivalent of C₆F₅H (I4^Fe). This interaction eventually leads to the metalated intermediate I5^Fe, characterized by two Fe−C₆F₅ bonds and a protonated TMP(H) group. The formation of I5^Fe is energetically favorable by −11.6 kcal mol⁻¹, requiring a relative energy barrier of 20.5 kcal mol⁻¹ (TS2^Fe) from I3^Fe.

Unlike TS1^Fe, featuring a pseudo-linear geometry with two TMP ligands and a dissociated THF molecule, TS2^Fe displays a tricoordinate iron center with a bound THF molecule ([Fe(C₆F₅)(THF)(TMP)]). The linear alternative for TS2^Fe (without coordinated THF) was also considered, resulting in an energy difference of +6.5 kcal mol⁻¹ higher (Figure S87). This discrepancy in coordination between TS1^Fe and TS2^Fe can be attributed to the loss of strong donation from the spectator TMP ligand present in TS1^Fe (Fe−N=1.832 Å), which electronically compensates for the lack of interaction with the THF molecule.

From I5^Fe, the subsequent dissociation of the protonated TMP(H) and the coordination of the second THF molecule result in the formation of the reaction product [(THF)₂Fe(C₆F₅)₂] (4^Fe-THF). Overall, the formation of this product is thermodynamically favorable by −16.8 kcal mol⁻¹ (Figure 4), in line with its isolation in experiments. These computational findings are further consistent with the monobasic behavior of 1^Fe observed in benzene, leading to the quantitative formation of the dimeric complex 2^Fe.

Notably, in the absence of THF, the deprotonation of a second equivalent of pentafluorobenzene, directly promoted by 2^Fe, was confirmed to be unfeasible (ΔG^≠=+37.6 and +38.7 kcal mol⁻¹ for monomeric and dimeric TSs, respectively; see Figure S88). Hence, the presence of THF provides unique access to the monomeric active species I3^Fe, facilitating the second metalation of pentafluorobenzene. A similar behavior was not observed for cobalt due to the higher stability of the dimer 2^Co, increasing the overall barrier for the putative formation of 4^Co-THF to +35.9 kcal mol⁻¹ (Figure S89). It is also noteworthy that cooperative ligand scrambling involving C₆F₅/TMP exchanges was computationally proposed and experimentally verified when working with a THF solution of 2^Fe in the absence of arene, resulting in mixtures containing 4^Fe-THF and 1^Fe-THF (Figures S33, S34 and S90).

Previous studies assessing the reactivity of Mg and Zn compounds containing TMP groups have revealed that the presence of inorganic salts, in particular LiCl can greatly enhance their kinetic basicity by forming ate complexes.³⁷ This prompted us to consider a similar strategy for the transition metal amides 1^Fe and 1^Co. In probing this, we found by spectroscopic analysis that addition of one equivalent of LiCl to d₈-THF solutions of these complexes is accompanied with a significant shifting of the TMP resonances which are now significantly more shielded. For example, for 1^Fe on addition of LiCl the TMP signals appear at 102.6, 75.70 and −5.40 ppm (vs 150.29, 110.10 and 3.82 ppm for 1^Fe). Moreover, its ⁷Li{¹H} NMR spectrum revealed a very shifted resonance at 69.10 ppm. A similar trend was found for cobalt (see Supporting Information for details). These findings support the formation of lithium metal(ate) complexes (Figure 5). Note that the relevant co-complexes of LiCl with M(HMDS)₂ have previously been prepared and structurally characterized.³⁸ While we could not grow suitable crystals for X-crystallographic analysis using LiCl, we noticed that addition of the THF-soluble ammonium salt (nBu)₄NCl to 1^Fe and 1^Co in THF afforded ate complexes with the TMP resonances appearing at almost identical shifts to those observed when LiCl was added. In the case of Fe, suitable crystals of the ammonium ferrate [{NBu₄}{Fe(TMP)₂Cl}] (5^Fe) could be obtained in a 77 % yield. X-ray crystallographic analysis confirmed formation of a ferrate anion resulting from co-complexation of Fe(TMP)₂ with a Cl⁻ anion (Figure 5). This structure is reminiscent of one recently reported by Werncke for co-complexation of (nBu)₄NBr with Fe(HMDS)₂.³⁹

To test the metalating ability of these ate-activated metal amide complexes, we studied first on the reaction of 1^Co with two equivalents of pentafluorobenzene. Gratifyingly on the addition of one equivalent of LiCl or (nBu)₄NCl the activation of the second TMP group was proven by ¹H NMR (SI for details). While attempts to crystallize a putative [{A}{Co(C₆F₅)₂Cl}] (A=Li(THF)_x⁺ or (nBu)₄N⁺) intermediate failed, addition of NHC I^tBu, once the metalation had occurred, led to the crystallization of [I^tBuCo(C₆F₅)₂] (4^Co-ItBu) in an 82 % yield (Figure 6). 4^Co-ItBu was characterized by X-ray crystallography as well as by ¹H and ¹⁹F{¹H} NMR spectroscopy (see Figure 6 and Supporting Information for details).

Next, we tested these reaction conditions with less activated fluoroarene substrates (in terms of pK_a values) focusing on 1,3-difluorobenzene and ethyl 3-fluorobenzoate. While 1^Co showed no reactivity, with 1^Fe one TMP ligand reacted to give dimeric [{Fe(TMP)Ar^F}₂] (Ar^F=2F-6-(CO₂Et)C₆H₃, 6^Fe; 2,6-F₂C₆H₃, 7^Fe) in 54 and 76 % isolated yields. Remarkably, while the presence of LiCl [or NBu₄Cl] is required to promote ferration of these substrates, these additives are not incorporated in the isolated products (Figure 6). A possible explanation could be that initially [{A}{Fe(TMP)(Ar^F)Cl}] (A=Li(THF)_x⁺ or NBu₄⁺) ate species form, which can be in equilibrium with [Fe(TMP)(Ar^F)] and LiCl [or (nBu)₄NCl]. This equilibrium can be shifted towards the formation of the neutral species, driven by the preferred dimerization of [Fe(TMP)(Ar^F)] to give 6^Fe and 7^Fe respectively, especially when considering that these compounds were recrystallized from non-polar benzene (see Supporting Information for details). The structures of 6^Fe and 7^Fe were established by X-ray crystallographic studies and the compounds were also characterized by ¹H and ¹⁹F NMR spectroscopy. They exhibit similar structural and spectroscopic features to those previously described for 2^Fe and 3^Fe with the exception that for 6^Fe the iron center achieves further coordinative saturation by interacting with the O of the C=O substituent ortho to the C that has experienced ferration [Fe1−O1, 2.296(9) Å].

Extending the applications of 1^Fe and 1^Co beyond fluoroarene metalation, we then studied whether these metal amides could also promote the deprotonation of cyclopentadiene as an alternative route to access the classical organometallic sandwich compounds, ferrocene and cobaltocene. Strikingly, ¹H NMR monitoring of the reactions in deuterated toluene showed that 1^Fe and 1^Co react at room temperature with two equivalents of freshly distilled cyclopentadiene to furnish ferrocene and cobaltocene respectively in quantitative yields with concomitant release of two equivalents of TMP(H). We also tested whether cobaltocene could be synthetized from an in situ made Co(TMP)₂. Pleasingly, the addition of 2 equivalents of cyclopentadiene to a d₈-toluene solution of a mixture of LiTMP (2 equivalents) and CoCl₂ resulted in an immediate color change of the solution, with the instantaneous formation of Cp₂Co (Figure 7) as witnessed by a diagnostic signal at −51.74 ppm.

In addition, the TMP groups on 1^Fe and 1^Co show excellent nucleophilicity, being able to insert two equivalents of CO₂ under mild reaction conditions affording bis(carbamate) complexes [(TMEDA)M(OC(=O)TMP)₂] (8^Fe, 62 % and 8^Co 75 % yield). Monomers 8^Fe and 8^Co feature one divalent metal coordinated to two chelating carbamate ligands via their O atoms and one TMEDA chelate, achieving a distorted octahedral geometry (Figure 7 and SI). Interestingly, the insertion process is significantly faster for Fe than Co, with full conversion to 8^Co over 16 hours compared to 2 hours for formation of 8^Fe. This reactivity is reminiscent to that previously reported by Mulvey for main group metal amides LiTMP or Al(TMP)₂iBu₂.⁴⁰ It should also be noted that transition metal amide complexes can also undergo CO₂ insertion in their M−N bonds⁴¹ although this reactivity has been significantly less explored than those involving M−H⁴² and M−OR (R=alkyl, aryl) bonds.⁴³

Conclusion

Popular in main group chemistry, TMP complexes have been made here with the earth-abundant transition metals Fe and Co and structurally, spectroscopically, and magnetically characterized. Mimicking the reactivity of main-group metal TMPs, these metal (II) amides can promote direct ferration (or cobaltation) of fluoroarenes, without changing their oxidation state or cleaving the C−F bonds of perfluorinated substrates. Their kinetic reactivity can be boosted further by using LiCl or NBu₄Cl as additives facilitating formation of more reactive metal(ate) species. The integration of DFT calculations has enriched our mechanistic understanding, revealing nuanced differences between Fe(TMP)₂ and Co(TMP)₂ in their ability to act as bases using one or two of their basic arms. Additionally, we highlight the pivotal role played by the coordinating solvent THF in promoting the polybasicity of Fe(TMP)₂.

Expanding their synthetic scope, these complexes demonstrate the ability to directly metalate cyclopentadiene, providing an alternative route to the classical organometallic sandwich complexes ferrocene and cobaltocene in quantitative yields. Furthermore, despite the steric bulk of the TMP groups, these metal amides can efficiently insert two equivalents of CO₂ under mild reaction conditions, showcasing their versatility.

Overall, this work transcends the boundaries of main group metal chemistry, unlocking new opportunities for deprotonative metalation. Their redox potential along with the more covalent character of the newly formed M−C bonds opens up a wealth of new opportunities for arene functionalization that are currently under investigation in our groups.

Conflict of interests

The authors declare no conflict of interest.