Dilithium Amides as a Modular Bis-Anionic Ligand Platform for Iron-Catalyzed Cross-Coupling

Abstract

Dilithium amides have been developed as a bespoke and general ligand for iron-catalyzed Kumada–Tamao–Corriu cross-coupling reactions, their design taking inspiration from previous mechanistic and structural studies. They allow for the cross-coupling of alkyl Grignard reagents with sp²-hybridized electrophiles as well as aryl Grignard reagents with sp³-hybridized electrophiles. This represents a rare example of a single iron-catalyzed system effective across diverse coupling reactions without significant modification of the catalytic protocol, as well as remaining operationally simple.

Graphical Abstract

Transition-metal-catalyzed cross-coupling remains one of the most versatile and widely used strategies for the selective formation of carbon–carbon bonds.^1–4 Iron-catalyzed cross-coupling methodologies have received increasing attention due to the high natural abundance and relatively low toxicity of iron, as well as complementary reactivity and generally short reaction times relative to precious metal-catalyzed variants.^5–9 These favorable properties have more recently led to the application of iron-catalyzed cross-coupling reactions in the synthesis of natural products and pharmaceuticals.^10,11 Notably, many of these examples use N-methylpyrrolidone (NMP), which allows for the cross-coupling of β-hydride containing alkyl nucleophiles with a range of aryl, heteroaryl, and alkenyl halides and pseudohalides.^12–16 Although NMP has proved to be an immensely effective additive for such cross-coupling reactions, its reprotoxicity, in combination with typically being used in large excess (200–300 mol %), makes its use in catalysis less desirable.¹⁷ Additionally, the mode of action of NMP was recently demonstrated to be through coordination to the magnesium counterion, resulting in the formation of iron(II) trialkyl ferrate species.^18,19 As NMP is not coordinated to the iron center, modulation of reactivity through traditional ligand modifications is impossible and the development of enantioselective variants more challenging.

Alternatives to NMP which circumvent the issue of toxicity have been sought, with cyclic ureas, alkoxides, and N,N,N′N′-tetramethylethylenediamine (TMEDA) having proved effective alternatives in replicating the reactivity observed with NMP.^20–24 Both the cyclic urea and magnesium alkoxide additives likely operate in a similar fashion to NMP, as well as being used in large excess with respect to iron. As a result, these additives would also possess similar limitations with regards to their modularity and tuning of the reactive iron center for further modification and development. A notable advantage of NMP and these additives, however, is that they do not require a slow and controlled addition of the Grignard reagent. By contrast, the use of TMEDA reported by Fox and co-workers, and indeed most iron-catalyzed Kumada–Tamao–Corriu cross-coupling reactions that use ligands, requires a controlled and often very slow addition of the nucleophile to avoid ligand displacement and associated unwanted side reactions.^24–27 This represents a common practical limitation to these methodologies. Moreover, the precise role of TMEDA and whether it acts as a ligand to the catalytically active iron species in iron-catalyzed Kumada–Tamao–Corriu cross-coupling reactions remains ambiguous. ^25,28–31 This would once again pose a limitation toward modifying reactivity at the iron center. A tunable ligand which allows for similar reactivity to that achieved with these additives, while maintaining the robust catalytic protocol and operational simplicity, would therefore be of significant value.

We hypothesized that a chelating bis-anionic ligand would allow access to a three-coordinate iron(II) ferrate species, analogous to the iron(II) trialkyl ferrate species accessed with NMP. This type of species would potentially display similar reactivity to the trialkyl ferrate, as well as be resistant to ligand displacement by the nucleophile contrasting more typical neutral ligands such as bisphosphines and amines. Altogether this would combine the reactivity and operational simplicity of reactions using NMP and related additives with the ability to modulate reactivity at the iron center afforded by directly coordinating ligands. This could potentially, in turn, lead to a more universal ligand platform for iron-catalyzed cross-coupling reactions.

Toward this goal, dilithium amides were chosen as the bis-anionic ligands, with the ease of synthesis making them an attractive and modular platform (Figure 1). These were also selected owing to the facile deprotonation using organolithium reagents, decreasing the potential for residual unreacted organometallic reagents which could react with the iron center. Additionally, the driving force of forming a lithium halide salt could be advantageous in complexation with the starting iron salt. A series of these dilithium amides were investigated using the prototypical cross-coupling of methyl 4-chlorobenzoate 1a and ethylmagnesium bromide (Table 1). In all cases the ligand and iron salt were prestirred for only 1 min prior to addition of the electrophile and Grignard reagent in quick succession.

Figure 1.

Design concept and rationale behind dilithium amides as ligands for iron-catalyzed cross-coupling reactions.

Table 1.

Optimization of Reaction Conditions for the Coupling of Methyl 4-Chlorobenzoate 1a and EtMgBr^a

Entry	Iron Source	Ligand (mol %)	Solvent	2a (%)b
1	Fe(acac)3	L1 (5 mol %)	THF	68
2	FeCl3	L1 (5 mol %)	THF	83
3	FeCl3	L2 (5 mol %)	THF	75
4	FeCl3	L3 (5 mol %)	THF	74
5	FeCl3	L1 (10 mol %)	THF	75
6	FeCl3	L1 (5 mol %)	Et2O	trace
7	FeCl3	L1 (5 mol %)	CPME	trace
8	FeCl3	none	THF	trace
9	Complex 1	none	THF	84
10c	FeCl3	L1 (5 mol %)	THF	79

^a1a (0.34 mmol; 0.17 M), EtMgBr (0.4 mmol) added over ∼20 s.

^bDetermined by GC analysis using dodecane as internal standard.

^cUsing EtMgCl.

The choice of iron salt in these reactions proved to be significant, with an 83% yield of 2a obtained using FeCl₃ and L1 compared with 68% in the analogous reaction using Fe(acac)₃ (entries 1 and 2). Variations of the ligand did not result in a significant change in yield, although bulkier L1 resulted in the highest yield with 83% of 2a (entries 2–4). Increasing the loading of L1 from 1 to 2 equiv, with respect to iron, resulted in a small decrease in the yield (entries 2 and 5). Finally, changing the solvent from THF to other ethereal solvents proved to have a dramatic effect, with negligible 2a being observed when the reaction was conducted in diethyl ether or cyclopentyl methyl ether (entries 6 and 7). In the absence of any ligand, 2a was not observed in any appreciable amount (entry 8). The L1–iron(III) chloride complex could be isolated as the lithium chloride bridged dimer [L1FeCl]₂[LiCl(THF)₂] from the reaction of L1 with FeCl₃ in THF. Crystals suitable for single crystal X-ray diffraction could be grown from pentane, and although they diffracted weakly, the structure of the complex and formulation of the crystals are unambiguous. This preformed complex gave comparable yield to premixing L1 and FeCl₃ (entry 9), supporting the notion that the diamide does indeed act as a ligand to iron. Finally, no significant halogen effect was observed for the Grignard reagent (entry 10). Notably, these reactions are complete within 15 min and did not require a highly controlled and slow addition of the Grignard reagent. A further advantage is that these reactions were run at 20 °C, rather than the lower temperatures often used for this type of cross-coupling reaction. The scope of this methodology was examined under these conditions using L1 with a range of electrophiles for which NMP and the aforementioned alternatives have been applied (Figure 2).

Figure 2.

Substrate scope for the coupling of various electrophiles with alkyl Grignard reagents. ^a1 (0.34 mmol), FeCl₃ (0.017 mmol), L1 (0.017 mmol), and Grignard reagent (0.4 mmol) added over ∼20 s. ^bDetermined by ¹H NMR analysis using 1,3,5-trimethoxybenzene as internal standard. ^c1 mmol scale of 1a. Values in parentheses denote isolated yields after PTLC.

The Grignard reagent chain length did not affect product yields for the coupling of methyl 4-chlorobenzoate, with all products being obtained in excellent yields (entries 1–3). Notably, however, secondary and tertiary alkyl Grignard reagents gave poor and negligible yields respectively (entries 4–5). A pronounced halogen effect was observed when switching to methyl 4-bromobenzoate or methyl 4-iodobenzoate, resulting in a drastic drop in yield (entry 6–7). In these cases, a mixture of unreacted electrophile and the product of protodehalogenation resulted. Similarly poor reactivity was observed for methyl 3- and 2-chlorobenzoates, resulting in minimal product formation and a mixture of unreacted electrophile and the product of protodehalogenation (see Supporting Information (SI)). Switching from the aryl ester 1a to aryl amides similarly results in high yields (entries 8–9). Facile and selective coupling at an acid chloride occurred readily, even in the presence of an aryl bromide bond (entry 10). The coupling of alkenyl bromides and triflates also occurs with excellent yields (entries 11–14), although this does not occur with the retention of stereochemistry observed for reactions using NMP. Quinoline- and pyrimidine-derived electrophiles also gave moderate to excellent yields (entries 15–18). In the case of 2-haloquinolines, there was no significant halogen effect observed (entries 15–16), contrasting the aryl esters. The position of the halogen substituent proved significant, with high yields maintained for 3-bromoquinoline (entry 17) but diminished for 6-chloroquinoline (entry 18). 2-Chloropyrimidine also proved effective, resulting in good yields (entry 19). Perhaps surprisingly, given the reactivity of haloquinolines, reaction of 2-chloropyridine gave poor yields (entry 20). This represents an example of differential reactivity between this ligand system and reactions using NMP and similar derivatives, suggesting that the diamide is indeed modifying the reactivity of the iron center and not simply allowing formation of the trialkyl ferrate species. Further examples of this differing reactivity are the lack of cross-coupled product formation with 4-chlorobenzotrifluoride, p-tolyl triflate, and 4-chlorobenzonitrile (see SI). Both 4-chlorobenzotrifluoride and p-tolyl triflate remain largely unreacted, despite the reactivity of alkenyl triflates. Instead 4-chlorobenzonitrile resulted in a complex mixture, likely resulting from the lack of reaction of the aryl chloride bond once again and instead reacting at the cyano group. It is also noteworthy that despite the solvent effect observed during optimization, the alkyl Grignard reagent could be used as a solution in either THF or diethyl ether.

While NMP is highly effective for the coupling of alkyl Grignard reagents, it has proved ineffective for the coupling of aryl Grignard reagents with alkyl halides.³² This is potentially the result of the analogous triaryl ferrate species forming the [FeAr₂(μ-Ar)]₂[Mg(NMP)₆] dimer, which was shown to be unreactive toward bromocyclohexane.³³ We hypothesized that these ligands may be able to prevent dimerization of the hypothetical ferrate species formed, resulting in effective cross-coupling of aryl Grignard reagents. We tested this using the coupling of PhMgBr with bromocyclohexane 3a (Table 2). Using the reaction conditions optimized for the coupling of alkyl Grignard reagents, phenylcyclohexane 4a was produced in moderate yield (entry 2). The use of Fe(acac)₃ instead of FeCl₃ had a minimal impact on the yield (entry 1), contrasting the more pronounced effect previously observed with the coupling of alkyl Grignard reagents (vide supra). Once again, the ligand substitution appears to have a minimal effect on the yield (entries 2–4), although this time the least sterically bulky variant L3 gave the slightly higher yield (entry 4). Due to this advantage being negligible, and for consistency, the reaction was further pursued with L1. Once again contrasting the trend observed for the coupling of alkyl Grignard reagents (vide supra), changing the solvent from THF to either diethyl ether or cyclopentyl methyl ether had a dramatically positive effect, with 4a being obtained in excellent yield (entries 5 and 6). Use of the preformed complex [L1FeCl]₂[LiCl(THF)₂] once again gave comparable yields to premixing L1 and FeCl₃ (entry 7).

Table 2.

Optimization of Reaction Conditions for the Coupling of Bromocyclohexane 3a and PhMgBr^a

Entry	Iron Source	Ligand (mol %)	Solvent	4a (%)b
1	Fe(acac)3	L1 (5 mol %)	THF	55
2	FeCl3	L1 (5 mol %)	THF	60
3	FeCl3	L2 (5 mol %)	THF	58
4	FeCl3	L3 (5 mol %)	THF	63
5	FeCl3	L1 (5 mol %)	Et2O	83
6	FeCl3	L1 (5 mol %)	CPME	84
7	Complex 1	none	Et2O	83

^a3a (0.34 mmol; 0.17 M) and PhMgBr (0.4 mmol) added over ∼20 s.

^bDetermined by GC analysis using dodecane as internal standard.

The generality of this reaction was subsequently examined with a range of alkyl electrophiles and substituted aryl Grignard reagents (Figure 3). These were conducted in diethyl ether, as the more readily available and common solvent, using FeCl₃ and L1 for consistency. On the Grignard reagent, electron-donating and moderately electron-withdrawing substituents were well tolerated with little to no effect on the yield observed (entries 2–4). Strongly electron-withdrawing groups, such as with 3,5-bis(trifluoromethyl)phenylmagnesium bromide, resulted in a significant drop in yield (entry 5). Steric encumbrance of the Grignard reagent appeared to shut down reactivity, with 2,6-dimethylphenylmagnesium bromide resulting in no detectable cross-coupled product (entry 6). It should also be noted that despite the pronounced solvent effect, the Grignard reagent could be used as a solution in diethyl ether or THF without any apparent impact on the yield. There is a halogen effect observed for the electrophile, with iodocyclohexane resulting in a moderate drop in yield (entry 7), whereas chlorocyclohexane resulted in a substantial decrease in the yield of cross-coupled product (entry 8). The reaction displays a surprising electrophile dependence, with bromocyclohexane resulting in excellent yields while other ring sizes or even acyclic secondary alkyl bromides give moderate to poor yields of cross-coupled product (entries 9–11). In these cases, the electrophile is consumed to form the products of protodehalogenation and β-hydride elimination. Primary alkyl bromides afforded the cross-coupled product in moderate yield (entry 12) whereas tertiary alkyl bromides resulted in no detectable cross-coupled product (entry 13). Notably, the presence of ether or aryl bromide functionalities on primary alkyl halides did not have a significant impact on the yield, demonstrating the preference for reaction with alkyl bromide bonds (entries 14–15). It should be noted that in the cases where lower yields are observed, predominant side-products observed are those resulting from protodehalogenation and β-hydride elimination.

Figure 3.

Substrate scope for coupling of alkyl halides with aryl Grignard reagents. ^a1 (0.34 mmol), FeCl₃ (0.017 mmol), L1 (0.017 mmol), and Grignard reagent (0.4 mmol) added over ∼20 s. ^bDetermined by ¹H NMR analysis using 1,3,5-trimethoxybenzene as internal standard. ^c1 mmol scale of 3a. Values in parentheses denote isolated yields after PTLC.

NMP has proven to be one of the most effective and applied additives in iron-catalyzed cross-coupling, although its reprotoxicity has resulted in alternatives being sought in recent years. Alternative additives have been able replicate this reactivity and avoid toxicity concerns; however, their scope for further reaction development is limited. Dilithium amides have been developed as a ligand capable of replicating an array of the reactivity observed with these additives, while providing an easily modifiable scaffold for further development and optimization. The first iteration of this promising new class of ligands has already proved capable of coupling alkyl Grignard reagents with a variety of electrophiles in an operationally simple reaction protocol. Furthermore, the same ligand has showed success in the cross-coupling of aryl Grignard reagents with alkyl halides, moving beyond the reactivity of NMP and related additives. This represents a first step toward a bespoke and universal ligand for iron, which will allow access to a wide variety of cross-coupling reactions. Further work will involve the development of more analogs of this ligand type in order to expand the substrate scope, targeting electrophiles that have thus far proved ineffective or where significant side-product formation is observed. Additionally, this work will aim to expand reactivity into other classes of cross-coupling including but not limited to C(sp³)–C(sp³) and C(sp²)–C(sp²) cross-coupling, for which early attempts with this first generation ligand have been unsuccessful, and the use of other nucleophilic coupling partners beyond Grignard reagents. Mechanistic and structural investigations will determine whether these ligands operate as designed as well as direct future developments and modifications.