Bond-Weakening Catalysis: Conjugate Aminations Enabled by the Soft Homolysis of Strong N–H Bonds

Abstract

The ability of redox-active metal centers to weaken the bonds in associated ligands is well precedented, but has rarely been utilized as a mechanism of substrate activation in catalysis. Here we describe a catalytic bond-weakening protocol for conjugate amination wherein the strong N–H bonds in N-aryl amides (N–H bond dissociation free energies ~100 kcal/mol) are destabilized by ~33 kcal/mol upon by coordination to a reducing titanocene complex, enabling their abstraction by the weak H-atom acceptor TEMPO through a proton-coupled electron transfer process. Significantly, this soft homolysis mechanism provides a method to generate closed-shell, metalated nucleophiles under neutral conditions in the absence of a Brønsted base.

Accessing metalated intermediates under mild conditions is an enabling feature of many contemporary catalytic methods. For example, in electrophilic activation schemes, the Brønsted acidity of bonds in a ligated substrate are significantly increased through coordination to an electron-deficient metal,¹ facilitating irreversible deprotonation by a comparatively weak Brønsted base. Such mechanisms are central to the success of the numerous metal-mediated C–H activation technologies,² as well as the soft enolization protocols commonly used for the activation of carbonyl compounds.³ Analogously, coordination to redox-active metal centers can significantly decrease ligand homolytic bond strengths (Figure 1A).⁴ In these mechanisms, formal hydrogen atom transfer (HAT) from a datively coordinated ligand is kinetically coupled to one-electron (1e⁻) oxidation of the metal center and formation of a new M–X bond. When favorable, these changes can provide the driving force necessary to enable the homolysis of normally strong bonds with weak H-atom acceptors. While this manner of bond weakening has been observed for many different metals and ligand classes,⁴ and the degree of destabilization can be substantial, these mechanisms have not been systematically exploited in catalytic applications.⁵

Figure 1.

Complexation-induced bond weakening and design of catalytic amination protocol based on the soft homolysis of strong N–H bonds.

We recently questioned whether complexation-induced bond weakening could provide new opportunities to develop catalytic “soft” HAT processes. In such reactions, the bond dissociation free energies (BDFEs) of normally strong σ bonds in a substrate would be destabilized upon coordination to a reducing metal center, enabling formal HAT to TEMPO or a similarly weak H-atom acceptor via proton-coupled electron transfer (PCET).⁶ Rather than generating a free radical, this elementary step would directly furnish a closed-shell, organometallic intermediate that could be utilized in subsequent bond-forming processes (Figure 1B). From a synthetic perspective, such reactions would have value in their ability to generate metalated nucleophiles under neutral conditions in the absence of an identifiable Brønsted base. Toward this goal, we report here the development of a conjugate amination reaction between N-aryl amide derivatives and electron-deficient olefins jointly catalyzed by TEMPO and Cp*₂Ti^IIICl (Figure 1C). The details of our study are outlined below, including kinetic, computational, and spectroscopic results that support bond weakening as the operative mechanism of substrate activation.

We first focused on identifying combinations of reducing metal centers and oxidizing H-atom acceptors that could function cooperatively to activate ligated substrates, but that would not react with each other directly. TEMPO was selected as the model H-atom acceptor given its ease of handling, commercial availability, and the strongly negative potentials required for its 1e⁻ reduction (E_1/2 = −1.95 V vs Fc/Fc⁺ in MeCN).⁷ In considering reducing metal complexes, we were initially drawn to pioneering work from Cuerva and co-workers on the bond-weakening ability of Cp₂Ti^IIICl.⁸ In a particularly dramatic example, those authors estimated that the O–H bond dissociation energies in the aqua complex of this Ti^III species are reduced by nearly 60 kcal/mol relative to the BDFE of free water (Figure 1A). In turn, this degree of weakening enables these aqua complexes to function as efficient stoichiometric H-atom donors for alkyl radicals.

While the capacity of these complexes to effect bond weakening was encouraging, we were mindful of reports from Waymouth and co-workers that Cp₂Ti^IIICl readily reacts with TEMPO to form an isolable Ti^IV complex,⁹ albeit one that exhibits an unusually weak Ti–O bond. However, those authors further demonstrated that the strength of the Ti–TEMPO association can be modulated by modifying the ancillary cyclopentadienyl ligands,¹⁰ with the Ti–O bond in bulkier complexes being significantly destabilized. Along these lines, we questioned whether we could identify a sufficiently encumbered titanocene complex where complex formation with a large ligand such as TEMPO would be rendered thermodynamically unfavorable, but coordination to a less sterically hindered substrate would remain viable, leading to cooperative bond activation and formation of a metalated Ti^IV nucleophile.

Our initial efforts focused on developing methods for amide activation. Amide N–H bonds are homolytically strong (N–H BDFE ~110 and 100 kcal/mol for N-alkyl and N-aryl amides, respectively)¹¹ and are energetically inaccessible as substrates for direct HAT to TEMPO (O–H BDFE in TEMPO-H 67 kcal/mol).⁷ However, upon complexation to Ti^III, these N–H bonds might be weakened to the point that TEMPO can function as a formal H-atom acceptor, resulting in the formation of a metalated Ti^IV aza-enolate intermediate that could serve as a competent nucleophile in a range of C–N bond-forming reactions. To test these predictions, we elected to study the conjugate amination of carbamate 1 jointly catalyzed by TEMPO and a series of reduced titanocene complexes. We envisioned a plausible catalytic cycle (Scheme 1) wherein 1 would first bind to the titanocene to furnish complex 2. This coordination event would weaken the carbamate N–H bond sufficiently to render proton-coupled electron transfer by TEMPO thermodynamically favorable and afford a closed-shell, Ti^IV aza-enolate intermediate 3.^6,12 Conjugate addition of this metalated nucleophile to the pendant olefin acceptor followed by metal migration would result in the formation of titanium enolate 4 that could undergo proton exchange with reduced TEMPO-H to furnish a neutral closed-shell product.^13,14 Electron transfer from the strongly reducing TEMPO anion (E_1/2 = −1.95 V vs Fc/Fc⁺ in MeCN)⁷ to the Ti^IV cation (E_p = −1.44 V vs Fc/Fc⁺ in MeCN)^15a is exergonic by ~500 mV and should readily regenerate the Ti^III catalyst and TEMPO radical, closing the catalytic cycle.

Scheme 1.

Proposed Catalytic Cycle

To evaluate this design, we evaluated a series of sterically differentiated titanocene catalysts in conjunction with TEMPO in the conjugate amination of carbamate 1 (Table 1). Unsubstituted and monosubstituted titanocenes capable of forming strong complexes with TEMPO were not catalytically active and returned the starting material unchanged (entries 1, 2). The preformed Cp₂Ti^IV(Cl)–TEMPO complex was similarly ineffective (entry 3). However, the more reducing and sterically hindered Cp*₂Ti^IIICl and mixed-ring Cp(Cp*)Ti^IIICl both proved to be effective catalysts and afforded the product in nearly quantitative yield after 1 h (entries 4, 5). As suggested above, this difference in reactivity likely results from the increased steric bulk imparted by the pentamethylcyclopentadienyl ligands, which precludes strong association to TEMPO (vide infra). This model reaction proceeds well in MeCN and DMF, but functions poorly in less coordinating aromatic, ethereal, and chlorinated solvents (entries 5–9). Other stable nitroxyls, such as AZADO (2-azaadamantane N-oxyl), were also found to be catalytically competent in these reactions (entry 10). Control experiments confirmed that the reaction does not proceed in the absence of either Ti^III or nitroxyl catalyst or with related Ti^IV complexes (entries 11–13). Finally, the cyclization of 1 was shown to proceed with similar efficiency with only 1 mol % of both catalysts (entry 14).

Table 1.

Optimization Studies^a

entry	Ti catalyst	nitroxyl catalyst	solvent	% yield
1	Cp2TiCl	TEMPO	MeCN	0
2	(t-BuCp)2TiCl	TEMPO	MeCN	0
3	Cp2TiIVCl(TEMPO)	TEMPO	MeCN	0
4	Cp(Cp*)TiCl	TEMPO	MeCN	95
5	Cp*2TiCl	TEMPO	MeCN	98
6	Cp*2TiCl	TEMPO	DMF	96
7	Cp*2TiCl	TEMPO	C6H6	15
8	Cp*2TiCl	TEMPO	THF	25
9	Cp*2TiCl	TEMPO	DCM	20
10	Cp*2TiCl	AZADO	MeCN	96
11	-	TEMPO	MeCN	0
12	Cp*2TiCl	-	MeCN	0
13	Cp*2TiIVCl2	TEMPO	MeCN	0
14a	Cp*2TiCl	TEMPO	MeCN	95

^aOptimization reactions performed at 0.05 mmol scale. Yields determined by ¹H NMR analysis of the crude reaction mixture relative to an internal standard.

^bReaction performed using 1 mol % of Cp*₂TiCl and 1 mol % TEMPO.

Having identified optimal conditions for the Ti^III-catalyzed cyclization of 1, we next conducted a preliminary survey of the scope of this process. On preparative scale, model carbamate 1 rapidly cyclized to give oxazolidinone 5 in 94% isolated yield (Table 2). N-Aryl amides were also viable substrates, providing N-phenylpyrrolidone 6 and N-acetylindoline 18 in 87% and 98% yield, respectively. Other heteroatoms were successfully incorporated to afford thiazolindinone 7 and imidazolidinones 8 and 14 from the corresponding thiocarbamate and urea starting materials. Gratifyingly, substrates containing α-Me or γ-Me substituents cyclized with high diastereoselectivity, giving products 9 and 11 in high yields. A substrate bearing a β-Me substituent readily cyclized to furnish adduct 10, which features a fully substituted tertiary carbinamine center. An ortho-substituted phenylcarbamate successfully cyclized to form tetrahydroquinoline 17 in 86% yield. Products 7, 8, and 12 were generated in high yield through conjugate addition to enones, and no ketyl-derived side products were observed.¹⁶ Electronically distinct substitution of the aryl group was possible, providing heterocycles 13, 14, and 16 in nearly quantitative yield, as well as ortho-substituted benzodioxole 15 with similar efficiency. To demonstrate that these reactions proceed under effectively neutral conditions, a base-labile fluorenylmethyl ester was prepared and successfully cyclized to afford 19 in 94% yield with no detectable loss of the ester protecting group. Notably, cyclization to form 19 using standard Brønsted bases, such as KOtBu in THF, led to quantitative ester deprotection. These conditions were not effective for intermolecular alkylation reactions or for the alkylations of N-alkyl amide substrates (N–H BDFEs ~110 kcal/mol). Efforts to address these limitations are currently ongoing.

Table 2.

Conjugate Amination Products^a

^aReactions performed at 0.5 mmol scale unless otherwise noted. Yields are for purified material. Diastereomeric ratios determined by analysis of the ¹H NMR spectra of the crude reaction mixtures.

^bReaction run at 0.05 mmol scale.

Experimental and theoretical observations are consistent with the proposed soft homolysis mechanism of substrate activation. First, the degree of bond weakening resulting from coordination of the N-phenyl propionamide to Cp*₂TiCl was assessed through density functional calculations (UB3LYP/6-31G(d) CPCM(MeCN). These studies indicated that the N–H bond strength decreased from 99 to 66 kcal/mol upon coordination to Cp*₂TiCl (Scheme 2), a weakening effect of 33 kcal/mol.^17,18 Notably, this value is sufficiently low that abstraction from the N–H bond to TEMPO is excepted to be nearly thermoneutral (O–H BDFE for TEMPO-H 67 kcal/mol). The analogous carbamate substrates were even further destabilized, with N–H bonds weakened by 39 kcal/mol.¹⁹

Scheme 2.

DFT Calculations (UB3LYP 6-31G(d) CPCM(MeCN))

Next, we sought to demonstrate the compatibility of the titanium reductant and nitroxyl oxidant in solution. Isodesmic calculations indicated that the bonding between Cp*₂TiCl and TEMPO is 23 kcal/mol less favorable than in the TEMPO complex of Cp₂TiCl, which Waymouth reported to have a Ti–O BDFE ~25 kcal/mol.^10,15b This implies that the Ti–O bond for the TEMPO complex of Cp*₂TiCl is exceptionally weak, with a BDFE only ~2 kcal/mol.¹⁸ Additional experimental support for the ability of TEMPO and Cp*₂TiCl to persist together in solution was obtained through EPR spectroscopy. Solutions containing equal amounts of TEMPO and Cp*₂TiCl in MeCN at room temperature produced spectra where signals from both species were simultaneously observable (Figure 2). This is consistent with the notion that the equilibrium for bonding between these species in solution is not highly favorable, and that meaningful concentrations of both oxidant and reductant can exist in solution on the time scale of the amination reaction. Inclusion of 1 equiv of carbamate 1 in these solutions also resulted in no observable spectral changes, suggesting a solvent-bound resting state for the titanocene catalyst.^19–21

Figure 2.

EPR spectra and simulations demonstrating compatibility of Cp*₂TiCl and TEMPO in MeCN solution.

Lastly, preliminary kinetic studies of the model amination reaction revealed first-order rate dependencies on the concentrations of carbamate 1, Cp*₂Ti^IIICl, and TEMPO. In addition, a significant primary kinetic isotope effect of 7 ± 1 was observed by comparing the rates of individual reactions run with N–H and N–D isotopologues of 1. These computational, spectroscopic, and kinetic data are consistent with a soft homolysis mechanism of substrate activation involving rate-limiting N–H abstraction from the Cp*₂Ti^IIICl complex of 1 by TEMPO.

In conclusion, we have developed a conjugate amination protocol jointly catalyzed by Cp*₂TiCl and TEMPO wherein a variety of evidence suggests that substrate activation occurs through complexation-induced bond weakening. More generally, the compatibility of the Cp*₂TiCl–TEMPO system in this work suggests an intriguing 1e⁻ parallel to frustrated Lewis pair chemistry, where sterically encumbered partners are not able to react with one another, but still function cooperatively to effect substrate bond activation. We anticipate that the elements of reaction design presented herein are general and will provide a basis for the development of other transformations that operate by a similar principle.