Divergent Fe-Mediated C–H Activation Paths Driven by Alkali Cations

Abstract

The association of the ferrous complex Fe^IICl₂(dmpe)₂ (1) with alkali bases M(hmds) (M = Li, Na, K) proves to be an efficient platform for the activation of Ar–H bonds. Two mechanisms can be observed, leading to either Ar–Fe^II species by deprotonative ferration or hydrido species Ar–Fe^II–H by oxidative addition of transient Fe⁰(dmpe)₂ generated by reduction of 1. Importantly, the nature of the alkali cation in M(hmds) has a strong influence on the preferred path. Starting from the same iron precursor, diverse catalytic applications can be explored by a simple modulation of the M^I cation. Possible strategies enabling cross-coupling using arenes as pro-nucleophiles, reductive dehydrocoupling, or deuteration of B–H bonds are discussed.

Introduction

Metalation of C–H bonds by transition metals (TMs) is a pivotal transformation in organometallic catalysis, leading to reactive organotransition species. Such compounds are key intermediates in numerous classic transformations such as cross-couplings, carbometalations, electrophilic additions, etc. Moreover, use of hydrocarbons as pro-nucleophiles in metalation processes is a more atom-economic alternative to main-group nucleophiles (e.g., RMgX, RZnX, RLi, etc.) used as transmetalation reagents.¹

Catalytic processes relying on a C–H metalation step with a TM source usually require the C–H bond target to be turned as a metalated intermediate at each catalytic cycle. Among this variety of organometallic species, organoiron compounds featuring well-defined C–Fe bonds have kept drawing particular attention during the last few decades.² Indeed, iron catalysts are appealing candidates for synthetic applications, since they are a sustainable and cheap alternative to noble metals.³ Formation of well-defined organoiron compounds is still particularly challenging due to the intrinsic instability of the C–Fe bond. For the same reason, a fine mechanistic analysis of the reactivity patterns followed by such species is also a delicate question.⁴ Indeed, depending on the iron oxidation state, C–Fe bonds can undergo fast homolytic cleavage,⁵ and two-electron reductive eliminations also easily occur from multiple coordinated R_n[Fe^II] complexes.⁶

Two very distinct strategies can be adopted for the generation of C–Fe bonds from C–H substrates and are particularly applied to the activation of Ar–H bonds. The first one relies on a direct ferration of the C–H bond, which is deprotonated by a suitable base at a Lewis-acidic iron center (Scheme 1a). Classic directed deprotonative ferrations were for example reported by Knochel, using Fe^II(TMP)₂·2MgCl₂·4LiCl (TMP = 2,2,6,6-tetramethylpiperidine).⁷ Nakamura also reported an efficient aryl–aryl cross-coupling involving a directed C_Ar–H deprotonation mediated by PhMgBr in the presence of a Zn^II source, the iron catalyst being associated with 1,10-phenanthroline;^8a the same group also recently reported improved methods of C–H functionalizations relying on a key deprotonative ferration.^8b−8d Similarly, Neidig and Ackermann demonstrated that directed C_Ar–H activation of triazole derivatives at an Fe^II intermediate followed by cross-coupling can be mediated by the Ar₂Zn/dppbz combination (dppbz = 1,2-bis(diphenylphosphino)benzene).⁹ Hevia recently reported an example of nondirected ferration of unactivated arenes (including C₆H₆) using Fe^II(hmds)₂ (hmds^– = N(SiMe₃)₂^–) and NaTMP relying on a cooperative effect between the Fe^II and Na^I cations.¹⁰

Scheme 1. Classic Strategies for the Generation of C–Fe Bonds.

(a) Deprotonative ferration; (b) oxidative insertion; (c) this work: non-redox and oxidative ferrations unlocked by alkali additives.

A second strategy consists in the ferration of C–H bonds by oxidative insertion of a Fe⁰ complex, leading to the formation of heteroleptic hydrido-organoiron(II) species (Scheme 1b). Early reports by Field and Tolman showed that transient Fe⁰(dmpe)₂ (dmpe = 1,2-bis(dimethylphosphino)ethane) generated by photolysis of the ferrous precursor Fe^II(R)(R′)(dmpe)₂ (R,R′ = H, Me or 2-Np, H) easily underwent insertion into a variety of C_sp²–H and C_sp³–H bonds.¹¹ Similar Fe⁰(P)₄ systems have been reported ever since. It was also recently demonstrated that in situ generated Fe⁰(dmpe)₂ was an efficient catalyst for the reductive dehydrocoupling of C_Ar–H and B–H bonds in order to access arylpinacolboranes Ar–B(pin),^12a or for hydrogen isotopic exchange (HIE) between alkenes or arenes and CD₃OD.^12b In the latter case, Fe⁰(dmpe)₂ is generated from 1 in the presence of pinacolborane (pinBH) and CD₃OD under blue light irradiation. Fe⁰(PMe₃)₄ also proved to efficiently promote ortho-directed insertion into C_Ar–H bonds¹³ and was successfully used by several groups such as Kakiuchi’s and Ackermann’s as a catalyst for the addition of Ar–H bonds onto alkenes¹⁴ and allenes,¹⁵ or for the homoallylation of aryl ketones using methylenecyclopropanes.¹⁶

The two activation pathways of a C_Ar–H bond depicted in Scheme 1a,b lead to strongly different aryliron(II) intermediates. In the first case, the H atom coming from the activated C_Ar–H bond acts as a proton H⁺ trapped by a basic partner and is not transferred onto the final product (Scheme 1a). Conversely, in the second case, this H atom affords a Fe-bound hydride H^–, the formal two-electron reduction of the C_Ar–H bond being ensured by the Fe⁰/Fe^II couple. This leads to possible catalytic transformations in which both the Ar group and the H atom are transferred onto a suitable target (Scheme 1b). The two aforementioned strategies (that is nonredox deprotonation, Scheme 1a, or two-electron oxidative insertion, Scheme 1b) usually require drastically different iron precursors and are extremely sensitive to ligand effects. Moreover, generation of efficient precursors of Fe⁰(dmpe)₂ demands the prior multistep synthesis and purification of extremely air-sensitive complexes such as Fe^II(R)₂(dmpe)₂ (R = H, Me) or Fe^II(2-Np)(H)(dmpe)₂, the latter being formed by reduction of FeCl₂(dmpe)₂ (1) by sodium naphthalenide (NaNp).^11a

In this report, we demonstrate that the well-known, easily accessible complex FeCl₂(dmpe)₂ (1) can be associated with alkali amides M(hmds) (M = Li, Na, K) to perform nondirected C_Ar–H activations by either deprotonative or oxidative addition paths (Scheme 1c, paths i–ii). Ar–[Fe^II] compounds formed by deprotonation can be active in cross-coupling with organic iodides, a key demetalation of a dmpe ligand being ensured by Lewis-acidic cations such as Li^I (as well as Mg^II) in the catalytic process. When Ar–[Fe^II]–H compounds are obtained, use of Li^I salts can selectively promote the borylation of Ar–H into Ar–B(pin) by a Ar–H/(pin)B–H dehydrocoupling. On the other hand, use of Na^I salts affords a near equimolar mixture of Ar_D–B(pin) borylation product along with hydrogen isotopic exchange (HIE) between C₆D₆ and (pin)B–H (formation of (pin)B–D).

This finding opens the door to particularly appealing synthetic strategies, since a simple ferrous precursor 1, readily synthesized from commercially available FeCl₂ and dmpe, combined with the suitable alkali salt M(hmds), can be used as a single precursor for generating organoiron species active in either cross-coupling or in dehydrocoupling/HIE catalysis. The role of the alkali cation in the speciation of the iron-containing species formed by action of M(hmds) on the precursor 1 is discussed. Characterization of key C_Ar–H activated products featuring Ar–Fe^II bonds was probed by ³¹P and ¹H NMR spectroscopies, X-ray diffraction, and Mössbauer spectroscopy.

Speciation of the Iron-Containing Species upon Treatment of 1 by M(hmds) (M = Li, Na, K)

We began by investigating the distribution of species, which can be formed under treatment of 1 by M(hmds) (M = Li, Na, K) in C₆D₆ at room temperature, under thermal activation (80 °C) or under UV irradiation (365 nm). Our main objective was to evidence the different C–D activation compounds that can be formed under those conditions. Treatment of 1 by M(hmds) in C₆D₆ can afford a distribution of species, which for some of them have been previously described in the literature (complexes 2–5, Table 1).¹⁷

Table 1. Speciation of Complexes 1–5 Obtained by Treatment of 1 by n equiv of M(hmds) (M = Li, Na, K) under Various Conditions.

entry	M (n)	t (h)	T (°C)	solventa	light	1	2C6D5	3′	3″	4C6D5,D	5–
1	Li (2)	72	20	C6D6	no	97	0	0	0	0	2
2	Na (2)	72	20	C6D6	no	0	0	40	7	5	4
3	K (2)	72	20	C6D6	no	5	0	10	20	4	7
4	Li (2)	72	20	C6D6	yes	5	0	0	0	40b	0
5	Na (2)	72	20	C6D6	yes	0	0	0	0	48b	2
6	K (2)	72	20	C6D6	yes	5	0	0	0	45b	5
7	Li (10)	22	20	C6D6/THF	no	97	0	0	0	0	2
8	Na (10)	22	20	C6D6/THF	no	0	7	16	8	5	3
9	K (10)	22	20	C6D6/THF	no	0	0	8	4	3	10
10	Li (10)	22	80	C6D6/THF	no	0c	0c	0c	0c	0c	18c
						0d	0d	0d	0d	38b,d	15b,d
11	Na (10)	22	80	C6D6/THF	no	0	0	12	13	0	0
12	K (10)	22	80	C6D6/THF	no	0	0	7	5	0	2

^aNeat C₆D₆ or a 9:1 mixture of C₆D₆/THF were used.

^bMixture of cis:trans-4_C6D5,D in a ca. 4:3 ratio.

^cSpectra recorded after cooling down the samples at room temperature during 1 h.

^dSpectra recorded after 10 additional hours at room temperature.

In neat C₆D₆, 1 remained mostly unreacted upon action of 2 equiv of Li(hmds) after 72 h at room temperature (97%, Table 1, entry 1), and less than 5% of 5^– were formed. On the other hand, action of 2 equiv of Na(hmds) under the same conditions led to an almost total conversion of 1. Reduced Fe⁰ complexes 3′ and 3″ were obtained (respectively 40 and 7%, entry 2), as well as the C–D activated complex 4_C6D5,D (5%). Formation of traces of the latter suggests that transient Fe⁰(dmpe)₂ (3) was obtained under those conditions, followed by oxidative insertion in the C–D bond of C₆D₆ (Scheme 1c, path (ii)). Use of K(hmds) (entry 3) afforded a similar distribution of 3′, 3″, and 4_C6D5,D (respectively, 10, 20, and 4%). It is of note that the possibility to use K(hmds) as a reductant of ferrous complexes stabilized by noninnocent coordinating platforms has already been reported in the past.^18a Similarly, Lei demonstrated that Na(hmds) was able to perform one-electron reduction of Cu^II salts into dicoordinated Cu^I(hmds)₂^–.^18b Moreover, the sole detection of NH(SiMe₃)₂ as an organic byproduct by GC-MS also suggests that the alkali amides used in the reduction course of 1 described herein promote formation of the Fe⁰ stage by successive one-electron transfers. The lack of reactivity of Li(hmds), compared to that of Na(hmds) and K(hmds), can be explained by the kinetic inertia of 1 (low-spin d⁶ octahedral complex¹⁹) and the strong electrostatic lithium/amide interaction existing in Li(hmds), which will slow down any reaction of Li(hmds) with the ferrous ion. Given that alkali cations can lead to the formation of π-arene adducts in the presence of an aromatic solvent (C₆D₆ herein),²⁰ the same experiments were carried out in the presence of a more donating solvent (C₆D₆/THF 9:1 mixture). No significant variations were obtained, meaning that reducing properties of Na(hmds) and K(hmds) remained unchanged under those conditions (Figures S15–S17). Treatment of 1 by 2 equiv of M(hmds) was then performed under UV irradiation (λ = 365 nm) during 72 h at room temperature. In all cases (respectively M = Li, Na, or K, entries 4–6), complex 4_C6D5,D was detected in significant quantities (respectively, 40, 48, and 45%) and was the major species detected by NMR along with traces of 5^– and unreacted 1. In other words, this means that a clean reduction of 1 by 2 equiv of M(hmds) into transient 3 can be triggered by UV irradiation, affording 4_C6D5,D by oxidative insertion. Formation of 3 by photolytic evolution of reactive sensitive precursors such as Fe^II(R)(R′)(dmpe)₂ (R,R′ = H, Me or 2-Np, H) has already been used and described in the past.^11,12a However, accessing 3 by reduction of a stable, easily handled precursor such as 1 in the presence of reductants classically used in synthetic chemistry such as M(hmds) is an appealing alternative to the use of the aforementioned sensitive precursors. A previous example of in situ activation of FeCl₂(dmpe)₂ in 3 by tBuOK had been reported by the group of Thomas in 2022.^12b

The action of an excess of M(hmds) on 1 at room temperature (10 equiv during 22 h) was then investigated (entries 7–9). In that case, a C₆D₆/THF 9:1 mixture was used, in order to circumvent solubility issues, which precluded the obtention of interpretable NMR spectra. Almost no conversion was observed with Li(hmds) (entry 7). Reduction by Na(hmds) (respectively K(hmds)) afforded ca. 25% of 3′ and 3″ (respectively 12%, entries 8–9) and 3% of complex 5^– (respectively 10%). Traces of 4_C6D5,D were observed for both Na(hmds) and K(hmds). Overall, the amount of NMR-detected species is smaller with 10 equiv of M(hmds) than with 2 equiv of the latter (M = Na, K; compare entries 8–9 and 2–3).²¹ However, in the case of Na(hmds) (entry 8), ca. 7% of complex (C₆D₅)Fe^II(Cl)(dmpe)₂ (2_C6D5) was also observed, with its diagnostic signal at δ = 67 ppm in ³¹P NMR.^11b Detection of small amounts of 2_C6D5 thus shows that Na(hmds) also enables C–D activation of C₆D₆ by nonredox deprotonation path i (Scheme 1c).

At a higher temperature (80 °C), reduction of 1 by 10 equiv of Na(hmds) (respectively K(hmds)) afforded a mostly NMR-silent distribution (entries 11–12), with reduced 3′, 3″ quantities roughly similar to what was obtained at room temperature. This distribution did not evolve after 10 h at room temperature. This means that the NMR-silent ill-defined reduced iron species obtained in that case are not able to promote the formation of 4_C6D5,D. The situation was strongly different when Li(hmds) was used (10 equiv per mole of 1, at 80 °C during 22 h, entry 10). 5^– (18%) is the sole species detected under these conditions and is almost undetected with Na(hmds) or K(hmds). This highlights the enhanced stability of 5^– in the presence of Li^I countercations, likely by the formation of a tight {5^–;Li⁺} ion pair. However, the speciation progressively evolves at room temperature (Figure S33), and 4_C6D5,D is detected after 10 additional hours at 20 °C (ca. 38%, cis:trans = 4:3, Figure 1a, and Table 1, entry 10, note d) additionally to 5^–, which remains almost unreacted (15%). New species 3* (ca. 10%), which displays ³¹P resonances similar to that of Fe₂⁰(dmpe)₅ (3′), is also detected (Figure 1a,b; a classic splitting of those signals upon use of ⁵⁷Fe-labeled material is observed, confirming the presence of one Fe atom in 3*).²² Additionally, broad resonances in the 50–60 ppm range are also observed and do not display any particular additional splitting upon ⁵⁷Fe-labeling experiments (Figure S33). Given that this area can correspond to diamagnetic Fe⁰-phosphine species,^11,12 formation of ill-defined diamagnetic aggregates stabilized by dmpe ligands is not excluded. Overall, this means that reduced species obtained using Li(hmds) at 80 °C can progressively act as an efficient reservoir of 3, enabling the C–D activation of C₆D₆ affording 4_C6D5,D, which is not the case when Na(hmds) and K(hmds) are used (entries 11 and 12).

Figure 1.

³¹P NMR patterns of (a) 4_C6D5,D; 3* and (b) ⁵⁷Fe-4_C6D5,D; ⁵⁷Fe-3* obtained after thermolysis of respectively 1 and ⁵⁷Fe-1 in C₆D₆ in the presence of 10 equiv of Li(hmds) during 22 h at 80 °C; spectra recorded after 10 additional hours at 20 °C, see Figure S33 for the full spectra; (c) 80 K ⁵⁷Fe Mössbauer spectrum of a frozen solution corresponding to spectrum (b).

Since a significant mass of iron is not detected by NMR speciation, ⁵⁷Fe Mössbauer spectroscopy of the reaction medium was also performed in the same conditions (treatment of ⁵⁷Fe-1 by 10 equiv of Li(hmds) in C₆D₆/THF (9:1) at 80 °C during 22 h, followed by 10 h at 20 °C, Figure 1c). The resulting spectrum could be simulated taking into account the presence of four different iron species. Complex 5^– was detected with Mössbauer parameters analogous to previous reports (δ = 0.57 mm/s, |ΔE_Q|= 0.60 mm/s, green component), accounting for ca. 15% of the overall iron in solution in line with the NMR speciation (Table 1, entry 10).^17b The major species observed by Mössbauer represents ca. 45% of the overall iron quantity (δ = 0.19 mm/s, |ΔE_Q|= 0.57 mm/s, Figure 1c, blue line), which is in fair agreement with the amount of 4_C6D5,D observed by ³¹P NMR speciation (38%). The nuclear parameters of this species are moreover also in agreement with what can be expected from low-spin Fe^II hydride species.²³ The broadness of this signal might be due to the existence of both cis and trans isomers of 4_C6D5,D, which overall lead to the superposition of two doublets with close parameters. Slight differences in Mössbauer parameters may indeed be expected for the signals of a cis/trans isomeric pair, as reported by Herber and Hayter.^23c Moreover, a third, rather broad signal accounting for 35% of the iron speciation displays parameters consistent with low-spin, Fe⁰ species (δ = 0.20 mm/s, |ΔE_Q| = 1.13 mm/s, magenta component).^23a Formation of such a diamagnetic compound is also in line with the ³¹P NMR speciation given in Table 1 and Figure S33, which shows that several ill-defined P-ligated diamagnetic compounds as well as ca. 10% of 3* are formed. Thus, it is likely that several low-spin Fe⁰ species, including 3*, may all contribute to the same Mössbauer signal. Last minor unidentified species (δ = 0.18 mm/s, |ΔE_Q| = 1.92 mm/s, 5% of the total Fe, cyan component) has also been included in order to obtain the best fit to the experimental spectrum. Overall, both Mössbauer and NMR speciations point toward the formation of 5^–, 4_C6D5,D, and a distribution of diamagnetic P-ligated Fe⁰ species upon treatment of 1 by Li(hmds) at 80 °C during 22 h, those species accounting for more than 90% of the iron quantity according to the Mössbauer simulation.

This section shows that C₆D₆ activation by oxidative insertion of in situ generated Fe⁰ species upon treatment of 1 by M(hmds) (M = Li, Na, K) is amenable under photolytic conditions and under thermal activation for M = Li.

Formation of Ar–Fe^II Bonds by Deprotonative Ferration

In addition to the detection of Fe⁰ complexes (which can yield 4_C6D5,D) that has been described in the previous section, formation of a small amount of 2_C6D5 (Table 1, entry 8) obtained by combination of 1, C₆D₆, and Na(hmds) also suggests that deprotonative ferration can be reached with the 1/M(hmds) system. Given that conditions discussed in Table 1 mostly led to easy reduction of 1 with Na(hmds) and K(hmds), the reactivity of Li(hmds) was more closely investigated. 1 was then subjected to treatment by 2 equiv of Li(hmds) in the presence of crown-ether 15-C-5.²⁴ It was anticipated that quenching of the Li^I cation by 15-C-5 would lead to an enhancement of the hmds anion basicity, thus favoring the C–D bond deprotonation.

When 1 is treated by 2 equiv of Li(hmds) associated with 2 equiv of 15-C-5 at 20 °C during 72 h, reduction into 3′ and 3″ is observed (25% overall, Table 2, entry 1). Deprotonation of the C₆D₅–D bond also occurs as a side pathway, and 10% 2_C6D5 are detected. Under photolytic conditions (entry 2), 40% of 4_C6D5,D are detected, and the quantity of 2_C6D5 remains roughly unchanged (14%). In line with the results discussed in Table 1 (comparison between entries 2–3 and 5–6), 3′ and 3″ are not observed upon irradiation: it is not excluded that they are formed under photolytic conditions and that their evolution toward 3 and, ultimately, to 4_C6D5,D is triggered by light. However, under thermal activation (80 °C, entry 3), 2_C6D5 was detected (34%) along with unreacted 1 (60%), and small amounts of ill-defined P-ligated diamagnetic species were detected (ca. 5% Fe_x⁰(P,P)_y, 50–60 ppm area in ³¹P NMR, see Figure S36). This means that deprotonation of the C₆D₅–D bond by Li(hmds) can be achieved while limiting the reduction in the presence of 15-C-5 under thermal conditions. In the absence of crown-ether at 80 °C (Table 2, entry 4), 90% conversion of 1 is observed. 2_C6D5 is not detected, and reduced ill-defined diamagnetic Fe_x⁰(P,P)_y species are formed (ca. 30% according to ³¹P NMR speciation, see Figure S37).²⁵ In other words, it means that quenching the acidity of the Li^I cation does not preclude formation of 2_C6D5, but it seems to enhance its stability at 80 °C.

Table 2. Speciation of Complexes 1–5 Obtained by Treatment of 1 by 2 equiv of Li(hmds) in the Presence or the Absence of 2 equiv of 15-C-5 under Various Conditions.

entry	M (n), additives	t (h)	T (°C)	solvent	light	1	2C6D5	3′	3″	Fex0(P,P)y	4C6D5,D	5–
1	Li (2), 2 equiv 15-C-5	72	20	C6D6	no	50	10	25% overalla		0	0	5
2		72	20	C6D6	yes	40	14	0	0	0	40b	0
3		22	80	C6D6	no	60	34	0	0	5	0	traces
4	Li (2), no 15-C-5	22	80	C6D6	no	10	0	0	0	30	0	traces

^a8% 3* were observed.

^bMixture of cis:trans-4_C6D5,D in a ca. 4:3 ratio.

In order to further analyze the reactivity of 2_C6D5, its brominated analogue PhFe^II(Br)(dmpe)₂ (2_Ph′) was synthesized by anion metathesis between 1 and PhMgBr and isolated.²⁶³¹P NMR analysis shows that signatures of 2_Ph′ and 2_C6D5 are close (δ = 67 ppm for both, see Figures 2a and S3–S4 for ³¹P and ¹H NMR spectra of 2_Ph′ and Figure S36 for 2_C6D5). Use of labeled ⁵⁷Fe-1 showed unambiguously that ⁵⁷Fe-2_Ph′ features a ⁵⁷Fe-ligated phenyl anion C₆H₅^–, characteristic ³¹P–⁵⁷Fe and ¹H_Ar–⁵⁷Fe couplings being observed (J(³¹P–⁵⁷Fe) = 42 Hz; J(¹H_meta–⁵⁷Fe) = 4 Hz; J(¹H_para–⁵⁷Fe) = 3 Hz; see Figure 2b).

Figure 2.

¹H (left) and ³¹P (right) NMR spectra of complexes 2_Ph′ (a) and ⁵⁷Fe-2_Ph′ (b) generated by addition of 1 equiv of PhMgBr (1 M in THF) onto 1 and ⁵⁷Fe-1; see Figures S3–S8 for full spectra; (c) crystal structure of 2_Ph′ (hydrogen atoms omitted for clarity, ellipsoids drawn at the 30% probability level); (d) 80 K ⁵⁷Fe Mössbauer spectrum of a frozen C₆D₆/THF (9:1) solution of ⁵⁷Fe-2_Ph′ formed by addition of PhMgBr onto ⁵⁷Fe-1, with δ = 0.24 mm/s and |ΔE_Q| = 0.46 mm/s.

Single crystals of 2_Ph′ suitable for X-ray diffraction analysis were obtained by cooling down a C₆H₆/THF (9:1) solution layered with pentane at −35 °C during 16 h (Figure 2c). Characterization of this structure by X-ray diffraction completes the NMR description of 2_Ph reported by Tolman and Ittel in 1979.^11b2_Ph′ is one of the rare examples of well-defined structurally characterized low-spin aryliron species in an octahedral environment, with a C_ipso–Fe distance of 2.08 Å. Camadanli described the formation of low-spin hydrido aryliron(II) species by insertion of Fe⁰(PMe₃)₄ into the ortho C–H bond of an aromatic ketimine, where the C=N bond acts as a directing group.¹³ In those species, the C_ipso–Fe bond is slightly shorter than in 2_Ph′, probably due to the presence of the intramolecular N–Fe ligation. By comparison, the cyclometalated complex [(benzamide)(dppe)(Ph)Fe^II]⁻ described as a dimer by Neidig and Ackermann and involving a nontethered phenyl anion features Ph–Fe bonds of 2.036/2.039 Å, closer to those of 2_Ph′.⁹ The ⁵⁷Fe Mössbauer spectrum of ⁵⁷Fe-2_Ph′ was also recorded in a frozen C₆D₆/THF (9:1) solution (Figure 2d). The observed Mössbauer parameters (δ = 0.24 mm/s and |ΔE_Q| = 0.46 mm/s) are consistent with a low-spin Fe^II center (diamagnetic, S = 0). Similar parameters were reported by Neidig and Gutierrez for the formation of an octahedral aryliron(II) complex featuring depe ligands (depe = 1,2-bis-diethylphosphinoethane), although no X-ray structure was reported for this species.²⁷

We then investigated the fate of 2_Ph′ in the presence of Li^I salts such as LiBr or LiOTf. When 2_Ph′ was treated by 5 equiv of those salts in neat C₆D₆, degradation of respectively 70 and 50% of 2_Ph′ was observed after 22 h at 80 °C, along with an insoluble dark precipitate. Concomitant formation of Ph–Ph was confirmed by GC-MS, attesting to a reduction of the Fe^II ion from 2_Ph′. On the other hand, 2_Ph′ remained unreacted in the absence of Li^I salts under the same conditions (Scheme 2a). Those results demonstrate that 2_Ph′ is thermally stable in the absence of a Lewis-acidic cation, explaining the persistence at 80 °C of its analogue 2_C6D5 when Li^I cation acidity is quenched in the presence of 15-C-5 (Table 2, entry 3).

Scheme 2. Reactivity of Isolated 2_Ph′ (a) or In Situ Generated 2_C6D5 (b) in the Presence of Li^I Salts; X = Br, OTf.

Decomposition of 2_Ph′ in the presence of Li^I salts into unidentified reduced iron species points toward an interaction of the lithium cation with the iron coordination sphere prior to the reduction. When 1 is treated by Li(hmds) at 80 °C during 22 h (conditions of Table 1, entry 10), a broad distribution in the −49/–50 ppm area is also observed in ³¹P NMR (Figure 3a). This is diagnostic of the release of free dmpe in the reaction medium, along with various adducts formed between the latter and Li^I salts, e.g., [dmpe·Li(Y)] (Y = Cl, hmds). σ-Ligation of Li^I cations by phosphines indeed leads to a minor shift of the latter in ³¹P NMR, as reported by Lappert and Layh,^28a or more recently by Reid for the Li(dmpe)₃⁺ cation (δ(³¹P) = −45.9 ppm in toluene-d₈).^28b

Figure 3.

³¹P{¹H} NMR spectrum (C₆D₆/THF (9:1), 20 °C, restricted to the free dmpe area) of a solution of 1 treated by 10 equiv of M(hmds) after 22 h at 80 °C; M = (a) Li; (b) Na; * = free dmpe, ¤ = [dmpe·Li^I] adducts; see Figures S27–S32 for the full spectra.

The instability of 2_Ph′ in the presence of Li^I salts (Scheme 2a) and the observation of dmpe-lithium adducts upon treatment of 1 by Li(hmds) (Figure 3a) attest to the demetallating power of the Li^I cation, able to decoordinate the dmpe from the iron. As summarized in Scheme 2b, consumption of 2_C6D5 in the absence of 15-C-5 suggests that the former readily undergoes a demetalation by Li^I cations, affording [dmpe·Li^I] adducts along with NMR-silent species. Conversely, trapping the Li^I cation by 15-C-5 prevents demetalation of dmpe from iron, enhancing stability of 2_C6D5 even at 80 °C. No [dmpe·M^I] adducts are observed using the Na/K hmds salts, in agreement with their weaker coordinating properties (Figure 3b). This observation is also in line with the strength of the Na^I–P ligation described by Reid in the case of the 1,2-bis-dimethylphosphinobenzene (diphos) ligand, much weaker than that of the Li^I–P analogues (discussion based on the comparison of the metric parameters of the X-ray structure of the [M(diphos)₃]⁺ adducts; M = Li, Na).^28b The interaction observed between the Li^I cation and the dmpe ligand in the system discussed herein is also in line with other reports from the recent literature. This again demonstrates the crucial role of main-group cations in the evolution of a distribution of organotransition species.²⁹ Bedford also reported a similar reactivity of Zn^II cations in Negishi couplings catalyzed by various (P,P)Fe^II salts, where Zn-(P,P) ligation is a key step of the catalytic cycle.³⁰

Association of 1 with alkali bases M(hmds) (M = Li, Na, or K) thus affords a simple platform, which unlocks the access to both deprotonative (nonredox) and oxidative ferrations (Scheme 1c, paths (i) and (ii)). The nature of the preferred path is impacted by the alkali countercation. Under thermal activation, K(hmds) promotes formation of reduced species, affording 4_C6D5,D by oxidative insertion of 3 into C₆D₆, and Na(hmds) allows detection of both 2_C6D5 and 4_C6D5,D intermediates. While treatment of 1 by Na(hmds) and K(hmds) mostly affords silent species (probably reduced aggregates), use of Li(hmds) can lead to either deprotonative ferration (formation of 2_C6D5) under thermal conditions or 4_C6D5,D under harsher conditions. Reduction of 1 followed by C–D activation yielding 4_C6D5,D is however observed for M = Li, Na and K under photolytic activation.

Possible catalytic applications of this divergent ferration system are discussed in the next section of this article, with a focus on the cross-coupling, reductive dehydrocoupling, and hydrogen isotopic exchange (HIE) chemistry.

Reactivity in Cross- and Dehydrogenative Coupling as well as in HIE Chemistry

Cross-Coupling of Arenes with Organic Electrophiles

The possible generation of Ar–Fe^II compounds such as 2_C6D5 from 1 by deprotonation of an Ar–H pro-nucleophile with M(hmds) (M = Li, Na) invited us to investigate the possible catalytic applications of this system in cross-coupling transformations. As outlined in the introduction of this report, organoiron(II) species are indeed key intermediates in cross-coupling processes; they are moreover known for efficiently activating aliphatic electrophiles by single electron transfers (SETs).³¹ The cross-coupling of C₆H₆ with n-C₁₂H₂₅I in the presence of 0.1 equiv of 1 and M(hmds) (1 equiv vs n-C₁₂H₂₅I) has been chosen as a benchmark reaction (Table 3); the main byproduct obtained under those conditions was n-C₁₂H₂₆, formed by reduction of the starting C–I bond.

Table 3. Cross-Coupling between C₆H₆ and R-I Mediated by 1 in the Presence of M(hmds) (1 equiv); Yields Based on GC-MS Analysis.

entry	M	R	T (°C)	t (h)	additives	Ph-R (%)	R-H (%)
1	Li	n-C12H25	40	72		2	3
2	Li	n-C12H25	80	72		25a	10
3	Na	n-C12H25	80	72		7	17
4	K	n-C12H25	80	72		9	9
5	Li	n-C12H25	80	72	15-C-5 20 equiv	1	2
6b	Li	n-C12H25	80	72		30a	4
7b	Li	cyclohexyl	80	72		35a	n.d.
8c	Li	n-C12H25	80	72		4	traces
9	Li	n-C12H25	80	72	dmpe 1 equiv	3	5
10	Li	n-C12H25	80	72	dmpe 3 equiv	2	3
11d	Li	n-C12H25	80	72		0	0
12d	Li	n-C12H25	80	72	dmpe 1 equiv	0	0

^aIsolated yield.

^b0.1 equiv of Fe(hmds)₂ + 1 equiv of dmpe vs Fe; other iron sources.

^c0.1 equiv of FeCl₂

^dNo iron source was used.

A first coupling attempt using Li(hmds) as a base at 40 °C during 72 h only led to traces of the coupling product (Table 3, entry 1). Increasing reaction temperature up to 80 °C however afforded a modest 25% coupling yield (entry 2). Owing to the competitive reduction of 1 into 3, 3′ and 3″, Na(hmds) was a much less efficient alkali base and afforded Ph-(n-C₁₂H₂₅) with a poor 7% yield (entry 3). In line with the formation of more reduced iron species in the reaction medium, reduction product n-C₁₂H₂₆ is obtained with a higher 17% yield. In a similar way, use of K(hmds) only afforded 9% of coupling product (entry 4). Addition of 15-C-5 to Li(hmds) (entry 5) led to a strong decrease of the coupling yield (1%). Since 15-C-5 proved to enhance the stabilization of 2_C6D5 in the presence of Li^I salts (Scheme 2b), this shows that the deprotonation of the Ph–H bond is not the more energetically demanding step of the coupling sequence. Conversely, this demonstrates that the electrophilicity of the Li^I cation is requested in the coupling process. Moreover, the best coupling yield (30%, entry 6) is obtained when Fe^II(hmds)₂ is associated with only 1 equiv of dmpe, evidencing the best performances of unsaturated, single-dmpe coordinated intermediates. Under the same conditions, a similar 35% yield was obtained when iodocyclohexane was used (entry 7). Traces of coupling product are also detected when FeCl₂ is used in the absence of dmpe (entry 8). Use of an excess of dmpe is highly detrimental to the reaction outcome, since the coupling yield dropped to 3% (respectively 2%) when 1 equiv of dmpe (respectively 3 equiv) is added to 1 in the reaction medium (entries 9–10, compare with entry 2). Control experiments performed in the absence of iron confirm that the latter is mandatory in the coupling process, since no product is detected when Li(hmds) is used alone or in the presence of dmpe (entries 11–12). At this stage, two possible mechanisms can explain the formation of the C–H coupling product. Either the electrophile is activated by Fe^II species such as an arylated complex derived from 2_Ph, or the formation of a reduced Fe⁰ complex (e.g., 3, 3′ or 3″) is required to do so. A stoichiometric experiment was performed, in which reduced iron species are formed by addition of 2 equiv of Li(hmds) on 1 at 80 °C (conditions of Table 2, entry 4), prior to reaction with n-C₁₂H₂₅I in the presence of 1 additional equivalent of Li(hmds) (Scheme 3).

Scheme 3. Stoichiometric Experiment for the Cross-Coupling between C₆H₆ and n-C₁₂H₂₅I.

In those conditions, 20% of cross-coupling product Ph-(n-C₁₂H₂₅) are formed, along with 25% n-C₁₂H₂₆, 7% of 1-dodecene, and 10% n-C₂₄H₅₀. Those three latter products are formed by reduction of the starting halide n-C₁₂H₂₅I. In particular, the n-C₁₂H₂₆/Ph-(n-C₁₂H₂₅) ratio is much higher in conditions of Scheme 3 (5:4) than in cross-coupling conditions (2:5, Table 3, entry 2). Moreover, 1-dodecene and n-C₂₄H₅₀ are solely detected as traces in cross-coupling conditions (Table 3, entry 2), whereas they represent 27% of the conversion of n-C₁₂H₂₅I in conditions of Scheme 3. This suggests that a sacrificial reduction of n-C₁₂H₂₅I occurs when reduced iron species are formed upon treatment of 1 with Li(hmds) and that those species do not contribute to the formation of the coupling product Ph-(n-C₁₂H₂₅). This thus points toward the possible implication of nonreduced Fe^II intermediates such as 2_Ph or 5^– in the coupling process. Given that the arylated complex 2_Ph has been identified as an intermediate on the reduction route of 1 to the Fe⁰ stage, its reactivity as a coupling intermediate has been further investigated.

The brominated analogue 2_Ph′ was used as an analytically pure compound, and its treatment by an excess of n-C₁₂H₂₅I at 80 °C during 16 h afforded the expected coupling product Ph-(n-C₁₂H₂₅) in a 24% yield (Scheme 4a), demonstrating that 2_Ph′ is able to promote the coupling step. A similar experiment was performed using 2_Ph′ generated in situ by transmetalation of PhMgBr with 1 (in conditions of Figure 2, affording a near-quantitative formation of 2_Ph′ as evidenced by Mössbauer spectroscopy, Figure 2d). This led, after treatment by 10 equiv of n-C₁₂H₂₅I in C₆H₆, to 43% of coupling product, whereas the coupling yield dropped to 27% in the presence of 15-C-5 (1 equiv vs PhMgBr, Scheme 4b). First, this result importantly demonstrates that the ability of 2_Ph′ to promote a cross-coupling-type reactivity with n-C₁₂H₂₅I is enhanced in the presence of Lewis-acidic cations such as Mg^II. It is important to note that the Ph-(n-C₁₂H₂₅) yield remains unchanged when thermolysis of 2_Ph′ is performed in C₆D₆ instead of C₆H₆ and that only traces of the deuterated compound C₆D₅-(n-C₁₂H₂₅) are obtained (Scheme 4b). This confirms that the aromatic core contained in Ph-(n-C₁₂H₂₅) comes from the Ph^– anion bound to the iron. Given that the coupling reactivity of 2_Ph′ with n-C₁₂H₂₅I in the presence of Mg^II salts associated with crown-ether 15-C-5 is very close to that of the salt-free, pure complex 2_Ph′ (respectively 27%, Scheme 4b, and 24%, Scheme 4a), this means that Mg^II cations play a key role in the implication of 2_Ph′ in the catalytic cycle, which is quenched by the presence of the crown-ether. ³¹P NMR monitoring of reaction of a ⁵⁷Fe-2_Ph′ sample (Scheme 4c, top, δ = 67 ppm, J(³¹P–⁵⁷Fe) = 42 Hz), generated in conditions of Figure 2, with 3 equiv of n-C₁₂H₂₅I at 80 °C during 4 h also shows that ⁵⁷Fe-1 and its bromide analogue ⁵⁷Fe-1_Br are generated in addition to the coupling product Ph-(n-C₁₂H₂₅) (Scheme 4c, bottom, respectively δ = 59.0 ppm, J(³¹P–⁵⁷Fe) = 42 Hz, and δ = 58.5 ppm, J(³¹P–⁵⁷Fe) = 41 Hz). The stoichiometric reactivity of 2_Ph′ in cross-coupling (Scheme 4a,b) shows without doubt that the presence of a Lewis-acidic cation (Mg^II in Scheme 4a,b) plays a beneficial role in the reaction outcome. In line with the observations of adducts such as [dmpe·Li(Y)] by ³¹P NMR (Figure 3a), it can be suggested that Mg^II salts enable the decoordination of one dmpe ligand from 2_Ph′, unlocking the access to unsaturated ferrous species able to promote the cross-coupling. Despite the apparent HSAB mismatch between the hardness of Mg^II ions and softness of P-ligands, weak, labile interaction between PMe₃ and Mg^II ions in Grignard reagents has already been probed.^32a Moreover, Arnold also reported an example of structurally characterized Mg^II adduct with alkyl 1,2-diphosphines, demonstrating the possible interaction between alkyl phosphines and Mg^II salts.^32b

Scheme 4. Stoichiometric Reactivity of 2_Ph′.

Used either (a) pure or (b) generated by addition of 1 equiv of PhMgBr on 1 followed by 15 h at 20 °C with R′I (R′ = n-C₁₂H₂₅); (c) ³¹P NMR monitoring of reaction of ⁵⁷Fe-2_Ph′ (top) with 3 equiv of n-C₁₂H₂₅I (bottom) during 4 h at 80 °C; (d) plausible catalytic cycle for the cross-coupling of C₆H₆ with R′I involving 2_Ph as a catalyst reservoir (X = Cl, I); (e) cross-coupling attempt performed on PhF (pK_a of the C–H bonds italicized).

Since 2_C6D5 is readily obtained by direct ferration of C₆D₆ mediated by Li(hmds) (Table 2, entry 3), this also means that 2_C6D5 can act as an intermediate in the cross-coupling between C₆D₆ and n-C₁₂H₂₅I. Li^I and Mg^II cations can thus act as a dmpe reservoir in a catalytic process: the main role of the dmpe therefore mostly seems to provide a coordination sphere able to stabilize transiently unsaturated Ph–Fe^II intermediates toward decomposition pathways, even if decoordination of the dmpe ligand from the iron center also seems to be mandatory in another step of the coupling cycle.

Those results unambiguously show that the coordination strength of the alkali countercation is a key parameter of the coupling discussed in Table 3. A likely explanation is that 2_Ph, owing to its octahedral low-spin d⁶ configuration, also displays a certain inertia toward reaction with n-C₁₂H₂₅I. The Li^I cation is then prompt to promote the decoordination of one dmpe ligand of 2_Ph, affording a nonsaturated 14-VE complex PhFe^II(dmpe)Cl (7_Ph), which will react with the organic electrophile in a SET step to afford Fe^III intermediate 7_Ph⁺ (Scheme 4d). Similar trends were reported by Gutierrez and Neidig, who showed that saturated octahedral 18-VE (P,P)₂Fe^II(Ar)X proved to be inert in coupling processes, whereas unsaturated tetrahedral 14-VE analogues (P,P)Fe^II(Ar)X displayed excellent performances.²⁷ The requirement of a decoordination of one dmpe ligand during the process is moreover confirmed herein by the detrimental effect of the excess of dmpe on the reaction outcome (Table 3, entries 9–10, to be compared with entry 2). An outer-sphere radical recombination between 7_Ph⁺ and n-C₁₂H₂₅^• then affords the coupling product Ph-(n-C₁₂H₂₅) with 7_X, as observed in classic aryl-alkyl Fe-catalyzed couplings (Scheme 4d).³¹7_X (in equilibrium with 1, as evidenced in Scheme 4c) finally re-enters the catalytic cycle after deprotonation of another equivalent of benzene. The proficiency of a SET/radical recombination sequence as key steps has been probed by the coupling of C₆H₆ with 1-bromohexene as a radical clock, which selectively afforded cyclization product PhCH₂C₅H₉ (see Section S4b). It is also not excluded that a dmpe-free intermediate such as Fe^II(hmds)₃^– (5^–, evidenced by ¹H NMR and Mössbauer speciations, see Table 1, entry 10 and Figure 1c) also initiates a catalytic cycle with deprotonation of C₆H₆, leading to a dmpe-free Ph-Fe^II compound. The latter would react in a SET step with the electrophile, since traces of cross-coupling product are also observed when FeCl₂ is used as an iron precursor in the absence of dmpe (Table 3, entry 8). This would be in line with the reducing activity of similar ferrate species such as [Me₃Fe^II]⁻, which was reported by Uchiyama, those ate complexes proving to readily undergo single-electron oxidation by a variety of electrophiles.³³ In our case, the poor amount of cross-coupling product obtained in the absence of dmpe however suggests that the latter is required to stabilize on-cycle, undercoordinated intermediates.

As detailed above, formation of 2_Ph′ under coupling relevant conditions and ability of the latter to provide the expected cross-coupling product Ph-(n-C₁₂H₂₅) by reaction with n-C₁₂H₂₅I (Scheme 4a–c) suggests that metalation of the Ar–H bond occurs prior to formation of the radical n-C₁₂H₂₅^•. An alternative mechanism would be the formation of n-C₁₂H₂₅^• followed by addition of the latter onto C₆H₆ in a radical nucleophilic addition mechanism. In this alternative mechanism, a C–H bond then undergoes homolytic cleavage to restore the aromaticity. In order to decipher whether the coupling step may also proceed following this alternative radical-based mechanism, a cross-coupling attempt has been performed using fluorobenzene (PhF) as a C–H bond source in the conditions of Table 3, entry 2 (n-C₁₂H₂₅I treated by 0.1 equiv of 1 and 1 equiv of Li(hmds) in neat PhF at 80 °C during 3 days). PhF has been chosen as a mechanistic probe since it can be activated by both the deprotonative pathway and radical addition mechanism;³⁴ it moreover displays three C–H bonds with very distinct acidities (pK_a(ortho H) = 36.8, pK_a(meta H) = 42.3, pK_a(para H) = 43.7, Scheme 4e).^34a Cross-coupling of PhF with n-C₁₂H₂₅I in the presence of Li(hmds) (1 equiv per mole of electrophile) led to a poor 15% yield in coupling products, significant quantities of Ph-N(SiMe₃)₂ being formed by defluorination of the starting material. A 11:5:1 distribution of the ortho/meta/para isomers of the cross-coupled product C₆H₄F(n-C₁₂H₂₅) is observed. This distribution can be indicative of the proficiency of a radical-based mechanism, since it is very similar to the 11.2:7.6:1 ratio reported in such conditions under Pd-catalyzed addition of alkyl radicals onto PhF.^34b In a pure deprotonative mechanism, an almost exclusive metalation at the ortho position would be expected, in line with the ability of PhF to undergo this process.^34c Therefore, taken together, those results demonstrate that some part of the coupling activity can also be attributed to a radical-based mechanism, in addition to the deprotonative way discussed earlier and involving the intermediate 2_Ph. This would explain the detection of traces of C₆D₅-(n-C₁₂H₂₅) as a coupling product when C₆D₆ is used as a solvent in the thermolysis of 2_Ph′ (Scheme 4b). 2_Ph′ would thus undergo oxidation by SET with n-C₁₂H₂₅I, but C₆D₅-(n-C₁₂H₂₅) would be formed by addition of the n-C₁₂H₂₅^• radical onto C₆D₆, bypassing its recombination with the iron-bound C₆H₅ group. In this radical mechanism, it is not excluded that nonarylated species such as Fe^II(hmds)₃^– (5^–) also act as one-electron reductants. However, it is also possible that the nature of the preferred mechanism (deprotonative way or radical addition) also depends on the Ar–H target.

C–H/B–H Bond Metathesis: Reductive Dehydrocoupling

Given that formation of elusive species 3 proved to occur upon reduction of 1 with Li, Na, and K hmds salts under photochemical activation, leading to 3′, 3″, or 4_C6D5,D (Table 1, entries 4−6), we also aimed at exploring potential catalytic applications of this system. As outlined in the introduction of this report, formation of hydrido aryliron species 4_Ar,H (Scheme 1b) is a key step in various transformations such as reductive dehydrocoupling. It was for example reported in 2015 that borylation of C_Ar–H bonds could be performed by photolytic generation of 3 from Fe^II(R)₂(dmpe)₂ (R = H, Me) in good to moderate yields.^12a Performances of 1 associated with M(hmds) (M = Li, Na, K) in the reductive dehydrocoupling of C₆D₆ and pinacolborane pinBH, leading to formation of perdeuterophenyl pinacolborane C₆D₅-B(pin), were then investigated (Table 4). The transformation was monitored at room temperature by multinuclear NMR analysis (³¹P, ¹¹B, ¹H). No borylation product C₆D₅-B(pin) was detected after 3 days when 1 was treated by 2 equiv of Na(hmds) in C₆D₆ in the presence of 10 equiv pinBH per mole of Fe (Table 4, entry 1). Under those conditions, no reduction of 1 and no further evolution to 4_C6D5,D were observed. To circumvent this matter, 1 was treated with 2 equiv of Na(hmds) during 72 h prior to addition of pinBH, affording the expected distribution of 3, 3′, and 4_C6D5,D species as detailed in the previous section (see the SI). After addition of pinBH, no evolution was observed after 7 days at 20 °C when the medium was kept in darkness (entry 2). However, upon irradiation at 365 nm, ¹¹B NMR monitoring showed progressive formation of C₆D₅-B(pin) (δ = 31.3 ppm, see the SI), see entry 3. A 15% borylation yield (respectively 30%) was observed after 72 h (respectively 6 days). After 10 days, 35% of C₆D₅-B(pin) were formed, with an overall incomplete 80% pinBH conversion, suggesting a deactivation of the catalyst. Interestingly, substitution of Na(hmds) by Li(hmds) led to a much more efficient conversion, since 75% of C₆D₅–B(pin) are formed after 3 days (entry 4). Progressive consumption of pinBH leads to a 80% borylation yield after 4 days; no starting material is detected after 6 days, with a 87% C₆D₅–B(pin) yield (isolated yield: 60%). The isolated yield obtained using this method is moreover higher than that was reported using the more sensitive precursor Fe^II(Me)₂(dmpe)₂ (34% PhB(pin) after 3 days^12a). K(hmds) led to a poor 15% borylation yield after 10 days and an incomplete 40% conversion, also suggesting a catalyst deactivation (entry 5).

Table 4. Borylation of C₆D₆ by pinBH and Deuteration of pinBH Mediated by 1 in the Presence of M(hmds) (M = Li, Na, K) under Photochemical Activation (in C₆D₆/THF 9:1)^a.

entry	M	t (days)	conditions	C6D5–B(pin)	pinBD	pinBH
1	Na	3	no light	0	0	100
2	Na	7	no lightb	0	0	100
3	Na	3	hνb	15	17	66
		6		30	30	29
		10		35	35	21
4	Li	3	hνb	75	8	6
		4		80	4	3
		6		87 (60)c	0	0
5	K	4	hνb	5	<5	88
		10		15	15	60

^a11B NMR yields are given.

^bPremix of M(hmds) with 1 (72 h) prior to addition of pinBH; see Figures S56–S58 for ¹¹B NMR monitoring.

^cIsolated yield.

The slow but complete evolution of the pinBH conversion using Li(hmds) as a reductant of 1 under irradiation clearly demonstrates that in situ generated active species 3 does not decompose over time, whereas catalyst deactivation is observed with Na(hmds) and especially with K(hmds). A possible explanation is that Na(hmds) and K(hmds) have a too strong reducing power, leading to higher concentrations of Fe⁰ species (such as 3, 3′, and 3″) and hence to decomposition of the latter over time by formation of unreactive metallic aggregates, leading overall to poorer catalytic activities. In line with this remark, significant quantities of unreacted 1 are observed by ³¹P NMR monitoring (Figures S48, S51, S54) of the reaction medium during catalysis when Li(hmds) is used as a precursor; a smaller quantity of 1 is detected when Na(hmds) is used, and only traces of 1 are detected for K(hmds), in line with the increasing reducing ability of the M(hmds) salts in the Li/Na/K series. Owing to the less strong reducing power of Li(hmds), the dynamic concentration of transient Fe⁰ species in the reduction course of 1 with Li(hmds) would then remain low enough to prevent their aggregation. ³¹P NMR monitoring shows that complexes 4_C6D5,D, Fe^II(H)₂(dmpe)₂ along with Fe^II(H)(D)(dmpe)₂ (δ = 76.7 (triplet) and 67.1 (triplet) ppm,^12a) are formed during the photolyses with Li, Na, and K hmds salts, which attests to the feasibility of the C–D and B–H bond activations by oxidative additions of 3 in this catalytic system. Formation of H₂ and of its isotopologue H–D is also confirmed by ¹H NMR (Figure S52), also attesting to the proficiency of the C–D and B–H bond activations in the dehydrocoupling process (Scheme 5, cycle (i)). Those species are not detected in the absence of irradiation. Moreover, pinBD is also detected by ¹¹B NMR during the photolysis with a diagnostic signal at δ = 28.4 ppm (triplet, ¹J(¹¹B–²D) = 21 Hz), regardless of the nature of the M(hmds) cation (Table 4). This is also in line with the known reactivity of 3 in C_Ar–H and B–H bond activations.^12a

Scheme 5. Plausible Borylation (i) and C–D/B–H σ-Bond Metathesis (ii) Cycles Mediated by In Situ Generated 3 in the Presence of M(hmds) (M = Li, Na); (P,P) = dmpe, Ar = C₆D₅.

In the recent literature, pinacolborane has been clasically associated with several nucleophilic partners such as tBuONa to efficiently reduce divalent transition-metal complexes L[M^II] of the third row (L = P- or N-based pincers and M = Fe, Co).³⁵ In those cases, tBuONa acts as a “masked reducing agent”, enabling formation of [pinB(OtBu)(H)]⁻;Na⁺, which itself allows the reduction of the [M^II] complex. It is thus not excluded that this mechanism also contributes to the generation of the active Fe⁰ species in the case discussed herein, additionally to the direct reduction of 1 under irradiation as evidenced by speciation in Table 1. However, since no reduction of 1 by Na(hmds) was observed in the presence of pinBH (Table 4, entry 1), a reduction mechanism relying on boron hydrides seems unlikely in the case of the Na(hmds). More importantly, the outcome of the transformation is also strongly impacted by the nature of the M(hmds) cation. As discussed earlier, Li(hmds) indeed allows a full conversion of pinBH, yielding 87% of C₆D₅–B(pin) (Table 4, entry 4), whereas Na(hmds) affords a modest 35% yield (entry 3). In the former case, a fast C–D/B–H σ-bond metathesis is observed (Scheme 5, cycle (ii)). However, both pinBD and pinBH are consumed by the borylation path (entry 4, and Scheme 5, cycle (i)), leading to C₆D₅–B(pin). On the other hand, when Na(hmds) is used, a progressive increase of the pinBD yield is observed (up to 35%, entry 3). This suggests that the C_Ar–B(pin) bond formation path is less operative when Na(hmds) is used instead of Li(hmds), but that the C–D/B–H σ-bond metathesis path still operates efficiently with Na(hmds).

The nature of the alkali cation can thus govern the reactivity of 4_C6D5,D, formed after oxidative insertion of 3 into the C_Ar–D bond (Scheme 5). It has been suggested previously that formation of unsaturated species by loss of at least one arm of a dmpe ligand was involved in the C_Ar–B bond formation step.^12a When Li(hmds) is used, this step can be triggered by the demetallating power of Li^I, leading to unsaturated complex 8_C6D5,D (Scheme 5, cycle (i)). Demetalation of one dmpe is also probably triggered by the photoirradiation of the system, which usually leads to a ligand labilization in saturated, inert intermediates, such as low-spin octahedral ferrous complexes. 8_C6D5,D affords the borylation product C₆D₅–B(pin), along with 8_H,D. Reintroduction of dmpe in the iron coordination sphere followed by reductive elimination regenerates 3 and affords H–D, detected by ¹H NMR (Scheme 5, cycle (i)). However, in the presence of Na(hmds), owing to the weak demetallating power of the Na^I cation, cycle (i) in Scheme 5 is difficultly followed due to the slow formation of 8_C6D5,D. Instead, a C–D/B–H σ-bond metathesis is observed, possibly by reaction between C–D and B–H activated intermediates 4_C6D5,D and 4_B(pin),H (Scheme 5, cycle (ii)). This ligand scrambling step between two saturated octahedral species might involve a bimetallic transition state (TS₁), as suggested in Scheme 5. The use of Na(hmds), in this case, slows down the demetalation of a dmpe ligand from one of the Fe^II ions, enabling the σ-bond metathesis to occur between those two saturated complexes, overall leading to the formation of pinBD and C₆D₅H.

These preliminary results show that 1 can be used as an efficient precursor to induce the formation of Fe⁰ species such as 3 in a reducing medium leading to activation of C_Ar–D and B–H bonds under photochemical conditions. More importantly, the judicious choice of the alkali countercation of the reductant, herein M(hmds), can drive the formation of C_Ar–B bonds by reductive dehydrocoupling (M = Li, Na), or the deuteration of B–H bonds affording B–D species by σ-bond metathesis can be observed (M = Na). This opens the way to promising synthesis of deuteroboranes R₂BD from boranes R₂BH, using C₆D₆ as a mild deuterium source and 1 as an easily handled iron catalyst, thus making a sustainable alternative to the classic formation of B–D bonds by addition of borodeuterides such as NaBD₄ or LiAlD₄ on electrophilic halogenoboranes R₂BX. Those methods indeed lead to stoichiometric amounts of salts as byproducts and rely on expensive deuteride sources (molar price of NaBD₄ (respectively LiAlD₄) being roughly 6 (respectively 30) times higher than that of C₆D₆).³⁶

Conclusions

In this report, we demonstrated that alkali salts M(hmds) can be used as bases or reductants to promote C–Fe bond formations from C–H substrates, respectively, by deprotonative ferration at a Fe^II center or by oxidative insertion of transient Fe⁰ species. These striking differences clearly show that the reactivity of M(hmds) toward C₆H₆ and 1 is impacted by the nature of the alkali countercation M⁺. Depending on the latter, hmds^– can act as a base and follow a nonredox ferration path affording Ar–Fe^II species able to react with organic halides in a cross-coupling process (M = Li). On the other hand, hmds^– can also act as a two-electron reductant of 1, affording elusive species Fe⁰(dmpe)₂ and promoting C–H insertion reactions leading to heteroleptic Ar–Fe^II–H intermediates (M = Li, Na, K). Under photolytic activation, this path can be efficiently used to promote either the catalytic borylation of arenes (the best yield and selectivity being afforded by M = Li) or the deuteration of boranes R₂BH using C₆D₆ as a deuterium source by σ-bond metathesis (being observed for M = Na, along with borylation). In all cases, the coordinating properties of the alkali cation M⁺ can govern the evolution of the iron complex distribution, playing a crucial role in the control of the oxidation state of the latter species as well as of their coordination sphere. Those countercations are thus far from being mere spectators in the catalytic transformations discussed herein and strongly contribute to the generation of the key on-cycle species from their inert resting states. This report thus also provides another example of the crucial importance of alkali cations in organometallic catalysis, which have long proved to strongly contribute to the selectivity of a large variety of transformations.³⁷

Methods

Materials and Reagents

All the samples were prepared in a Jacomex Campus glovebox under an argon atmosphere or vacuum Schlenk lines. Glassware equipped with a J. Young valve was dried overnight at 120 °C before use. Deuterated solvents purchased from Eurisotop were thoroughly degassed by freeze–thaw cycles and dried overnight on 4 Å molecular sieves. Nondeuterated solvents were dried over a Na/benzophenone mixture and distilled before use. NMR tubes equipped with a J. Young valve were used for all NMR experiments. Chemicals were obtained from commercial sources and used as purchased, unless specified otherwise or after drying when necessary. A Thorlabs’ M365LP1 LED (nominal wavelength = 365 nm, fwhm = 9 nm) mounted to the end of a 57 mm heat sink enabling heat dissipation has been used for the irradiation setup. The details of the synthetic procedures to prepare both iron complexes, Fe(Cl)₂(dmpe)₂ (1) and FeBr(C₆H₅)(dmpe)₂ (2_Ph′), are described in the Supporting Information.

General Procedure for Cross-Coupling Reactions

The reaction with Fe(Cl)₂(dmpe)₂ (1) and Li(hmds) at 80 °C is given as an example (Table 3, entry 2). In an argon-filled glovebox, a 12 mL vial equipped with a magnetic stir bar was loaded with Fe(Cl)₂(dmpe)₂ (3.5 mg, 8.4 μmol, 0.1 equiv), C₆H₆ (1.5 mL), and decane as an internal standard for GC-MS analysis. The mixture was placed under stirring, and n-C₁₂H₂₅I (20.2 μL, 82 μmol, 1 equiv) was added dropwise followed by Li(hmds) (15.0 mg, 82 μmol, 1 equiv). The color of the reaction medium gradually changed from emerald green to yellow/orange. The reaction mixture was stirred for 5 min and transferred into an NMR tube fitted with a J. Young valve. The tube was placed in an oil bath at 80 °C and left for 72 h. The color turned pale yellow. The tube was cooled to room temperature and quenched with 2 mL of H₂O. The solution was extracted with 2 mL of MTBE and analyzed by GC-MS.

General Procedure for Dehydrocoupling Reactions

The reaction with Fe(Cl)₂(dmpe)₂ (1) and Li(hmds) is given as an example (Table 4, entry 4). In an argon-filled glovebox, a 12 mL vial equipped with a magnetic stir bar was loaded with Fe(Cl)₂(dmpe)₂ (3.7 mg, 8.7 μmol, 0.1 equiv), C₆D₆ (0.6 mL), and THF in a 9:1 ratio. The mixture was placed under stirring, and Li(hmds) (2.9 mg, 17.3 μmol, 0.2 equiv) was added. The color of the reaction medium gradually changed from emerald green to yellow/orange. The reaction mixture was stirred for 3 days at room temperature in order to afford the expected distribution of 3, 3′, and 4_C6D5,D species, monitored by ³¹P NMR. Pinacolborane (12.6 μL, 87 μmol, 1 equiv) was added, and the mixture was transferred into an NMR tube fitted with a J. Young valve. The tube was irradiated, and the reaction was monitored by ¹¹B, ³¹P{¹H} and ¹H NMR.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacsau.3c00649.

• vw479_0m (CIF)

• General procedures, ¹H, ³¹P, ¹¹B, ¹⁹F NMR spectra (PDF)

Author Contributions

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

G.L. thanks the ERC (Project DoReMI StG, 852640) and the CNRS (Project IrMaCAR) for their financial support. The NMR shared facilities of Chimie ParisTech (Dr. M.-N. Rager) are thanked for technical support.

The authors declare no competing financial interest.