Deciphering competing reaction pathways in the α-C–H functionalization of unprotected alicyclic amines via ketone-mediated hydride transfer

Abstract

Alicyclic amines are pervasive in pharmaceuticals and natural products, yet direct α-C–H functionalization of unprotected secondary amines remains a long-standing challenge. Building on the historical oxidation of lithium amides by ketones, we present a combined experimental and computational study that establishes the mechanistic basis for α-functionalization of these substrates. Competition experiments, ReactIR monitoring, kinetic isotope effects, and DFT calculations demonstrate that imine formation proceeds through concerted hydride transfer from solvent-coordinated monomeric lithium amides. Nucleophilic addition of lithium amides to the ketone oxidant is identified as a major yield-limiting side reaction and is shown to depend strongly on amine ring size, oxidant electrophilicity, and steric effects. Regioselectivity in α-substituted amines arises from a combination of steric and electronic factors governing C–H bond activation. These findings clarify the fundamental reactivity of lithium amides and provide a practical platform for the synthesis of polysubstituted alicyclic amines without protecting groups or transition metals.

Mechanistic studies establish that ketone-mediated α-C–H functionalization of unprotected alicyclic amines proceeds via concerted hydride transfer from monomeric lithium amides, in competition with N-addition to the oxidant.

Introduction

Alicyclic amines are ubiquitous core structures in natural products and other bioactive molecules.¹ Synthetic methods to access functionalized alicyclic amines from their parent unsubstituted heterocycles have drawn considerable attention with the most extensively studied approaches involving α-C–H bond functionalization.² This is hardly surprising, given that such methods promise the rapid diversification of simple starting materials. However, a limitation common to most approaches is the need to utilize tertiary or N-protected alicyclic amines as substrates, due to incompatibilities with basic nitrogen atoms or N–H bonds (Scheme 1a).³ N-protecting groups are also often needed to serve as directing groups for a given transformation. In addition, expensive transition metal catalysts are often required. Methods that generate α-substituted secondary alicyclic amines directly from their unprotected parent amines remain limited. Notable exceptions include early transition metal-catalyzed hydroaminoalkylation reactions (Scheme 1b),⁴ and the electrochemical, oxidative α-cyanation of piperidines (Scheme 1c).⁵ More recently, an electrochemical α-oxygenation of unprotected alicyclic amines to form the corresponding lactams has been reported,⁶ in addition to a photochemical method for the addition of carbanions to in situ-generated iminium ions (not shown).⁷ Inspired by studies from the 1960s and ‘70s in which Wittig and coworkers established the ability of Li-amides 1 to undergo oxidation to the corresponding imines 2 upon treatment with a simple ketone oxidant (e.g., benzophenone),⁸ our group recently developed a range of transformations that exploit these historical findings in the α-C–H bond functionalization of unprotected alicyclic amines to provide products such as 3 (Scheme 1d).⁹ While other methods are available to access imines 2, such as N-halogenation followed by base-induced dehydrohalogenation,¹⁰ the Li-amide based approach is unique in that it enables rapid access to imines 2 at low temperatures. This is significant, as these species tend to be unstable or undergo rapid conversion to unreactive imine trimers,^10,11 preventing their subsequent reaction with desirable nucleophiles. A particularly attractive aspect of this chemistry is that amines with an existing α-substituent selectively undergo imine formation at the α′-site, allowing for the synthesis of alicyclic amines with multiple substituents in different ring positions, typically favoring formation of thermodynamically disfavored isomers that are difficult to obtain by other means.¹² Since our first report,^9a this chemistry has been widely adopted in academic and industrial labs alike as the method of choice to rapidly prepare saturated 2-substituted azacycles, facilitating the discovery of novel bioactive substances and the preparation of other valuable materials.¹³ The broad interest in this methodology is further evidenced by the recent development of a flow approach that enables convenient scale-up.¹⁴ In stark contrast to the broad utility of the method, the mechanism of the key imine formation is only understood in its most rudimentary form. Beyond the notion that imine formation most likely proceeds via a hydride transfer step akin to what is depicted in Scheme 1d,¹⁵ key factors dictating chemoselectivity, regioselectivity and divergent behavior of ketone oxidants remain unknown, thus hampering the further development of this chemistry. Here we report the results of detailed experimental and computational studies aimed at providing insights into mechanistic aspects of the key steps of these transformations, including competing reaction pathways.

Scheme 1. Overview of methods for alicyclic amine α-C–H bond functionalization.

Results and discussion

The studies described below share a single objective: to identify the factors that determine whether a lithium amide is oxidized to the desired imine or is diverted into competing pathways. Relative rates, direct spectroscopic observation of the oxidation event, product distributions, site selectivity, and isotope effects are examined in turn. As detailed in the following sections, these experiments converge on a common picture in which rapid, concerted hydride transfer competes with nucleophilic addition of the amide to the ketone, with the balance between the two pathways dictated by the structures of both reaction partners.

Competition experiments probing relative rates

The optimized one-pot synthesis of 2-phenyl-pyrrolidine (4) from pyrrolidine involves deprotonation by nBuLi at −78 °C, then oxidation with benzophenone, followed by addition of phenyl lithium and gradual increase of the reaction temperature to rt (Scheme 2a). Under these conditions, product 4 was obtained in 60% yield.^9a Extending the reaction time of the oxidation step did not lead to any improvement in yield. To gain insights into the relative rate of oxidation in comparison to other potentially competing processes, we explored several competition experiments that involve premixing of the amine and the ketone oxidant. When a solution containing both pyrrolidine and benzophenone was treated with nBuLi at −78 °C, followed by addition of phenyl lithium, 4 was obtained in only slightly reduced yield (54%, Scheme 2b). This indicates that the deprotonation of pyrrolidine by nBuLi is significantly faster than any reaction of nBuLi with benzophenone. Using phenyl lithium as both the base and the nucleophile led to 4 in nearly identical efficiency as under the optimized conditions. Accordingly, the sequence of amine deprotonation/oxidation is vastly faster than the well-known addition of phenyl lithium to benzophenone.¹⁶ A similar experiment was conducted with nBuLi and azacyclotridecane (selected to provide a non-volatile product). While the standard conditions resulted in the formation of product 5 in 64% yield (Scheme 2c), addition of excess nBuLi to a mixture of azacyclotridecane and benzophenone provided 5 in 32% yield (Scheme 2d). Although this yield is significantly reduced, this result is nevertheless remarkable as nBuLi is known to rapidly engage benzophenone, with the main reaction pathway being reduction to benzhydrol rather than addition to form the corresponding tertiary alcohol.¹⁷

Scheme 2. Competition experiments.

ReactIR studies

Considering that both the C O bond stretch of the ketone oxidant and the C N bond stretch of the cyclic imine intermediate should have observable IR absorptions, we performed a series of ReactIR studies to observe the oxidation of cyclic lithium amides by ketones more directly, and to gain quantitative insights into the time required to complete the oxidation step. Initial experiments focused on benzophenone as the oxidant but did not result in compelling IR profiles due to partial peak overlap.¹⁸ However, the oxidation of N-lithio piperidine by trifluoroacetophenone was amenable to analysis. As shown in Fig. 1, the oxidation of N-lithio piperidine is rapid at −78 °C as judged by the disappearance of the C O bond stretch at 1724 cm⁻¹ within less than three minutes. Concurrently, a new signal at 1659 cm⁻¹ was observed, attributable to the formation of the imine intermediate. While this picture is consistent with a concerted hydride transfer mechanism, a rapid sequence involving ET/HAT cannot be ruled out on this basis. After imine formation, the solution was kept at −78 °C for a period of time to monitor the stability of the imine species under these conditions. As no degradation of the imine signal was noted, imine trimerization apparently does not occur in this time span. Phenyl lithium was then added, leading to a relatively fast decline of the imine signal with complete imine consumption being noted within one hour. It should be noted that the apparently rapid initial kinetics are likely due, at least in part, to a dilution effect upon addition of the phenyllithium solution. In this experiment, product 6 was isolated in 65% yield. A similar reaction profile was observed for N-lithio azepane (Fig. 2). In this case, the C O bond stretch at 1724 cm⁻¹ disappeared completely within one minute.

Fig. 1. Trends of functional groups observed over time. Blue line: C O stretch of trifluoroacetophenone at 1724 cm⁻¹; purple line: C N stretch of 1-piperideine at 1659 cm⁻¹. N-Lithio piperidine (1.05 equiv. in 1 mL ether) was added to a solution of trifluoroacetophenone in 1 mL of ether, followed by the addition of PhLi (1.5 equiv.). The reaction was performed at −78 °C.

Fig. 2. Trends of functional groups observed over time. Blue line: C O stretch of trifluoroacetophenone at 1724 cm⁻¹; purple line: C N stretch of azepane imine at 1656 cm⁻¹. N-Lithio azepane (1.05 equiv. in 1 mL ether) was added to a solution of trifluoroacetophenone in 1 mL of ether, followed by the addition of PhLi (1.5 equiv.). The reaction was performed at −78 °C.

Probing competing pathways in reactions of lithium amides with ketones

While the experiments above establish that oxidation is rapid and outcompetes other reactions of the organolithium reagents, they do not account for the moderate yields typically obtained; we therefore turned to the fate of the ketone oxidant itself. Kowalski and Creary et al. found that the reduction of benzophenone by LDA occurs in competition with addition to form an O-lithiated hemiaminal that reverts back to benzophenone upon workup.¹⁹ Others have reported that addition represents the major pathway in reactions of lithium amides with aldehydes, although reduction can become a significant side reaction for particularly electrophilic aldehydes.²⁰ To quantify the extent to which lithium amide addition to the ketone oxidant interferes with the desired oxidation for a given substrate combination, we performed a series of experiments (Scheme 3). When using benzophenone as the limiting reagent in the reaction with N-lithium pyrrolidide, benzhydrol was isolated in 68% yield, while benzophenone was recovered in 32% yield (Scheme 3a).²¹ Based on our results from ReactIR studies, complete consumption of the ketone occurs very rapidly upon addition of the amide. Competing formation of the corresponding O-lithiated hemiaminal (in our hands not easily observable by ReactIR) is the most likely explanation. The possible formation of an aldol product^8a can be ruled out based on our earlier observation that the formation of the requisite metalloenamine (1-azaallyl anion) from a cyclic imine typically requires elevated temperatures and longer reaction times. For instance, the deprotonation of piperideine by LDA at 0 °C requires about one hour to go to completion.^9c Accordingly, the rate of the reduction of benzophenone by N-lithium pyrrolidide is only about twice that of the rate of addition, thus representing a significant side reaction. This finding is qualitatively supported by our observation that the highest yield we've thus far obtained for functionalizing pyrrolidine is 62%,^9a implying a 91% yield for the addition of the nucleophile when assuming a 68% yield for the formation of the imine. The actual yield for the addition step may be even higher, considering that the rather volatile nature of mono-substituted pyrrolidines may lead to some loss of product.

Scheme 3. Probing reduction vs. addition with excess Li amide.

Interestingly, when conducting an identical experiment with benzophenone and N-lithium piperidide, only a 10% yield of benzhydrol was obtained (Scheme 3b). This is consistent with our prior observation that benzophenone is a poor oxidant for piperidine substrates under a variety of conditions. We then tested both pyrrolidine and piperidine with trifluoroacetophenone and t-butyl phenyl ketone, the favorable oxidants for several transformations we have developed (Scheme 3c–f). Due to its substantial solubility in water, quantitative recovery of trifluoroacetophenone proved challenging when conducting our standard aqueous workup. In addition, alcohol 7 is rather volatile. We thus determined the ratios of 7 and recovered trifluoroacetophenone by crude NMR, following quenching with d₄-methanol and using ¹⁹F-NMR alongside an internal standard (trifluorotoluene) to obtain the results. Interestingly, hemiaminals 8 and 9 were observed instead of trifluoroacetophenone. These species also form spontaneously when mixing trifluoroacetophenone with pyrrolidine or piperidine, respectively. Surprisingly, N-lithium pyrrolidide reduced trifluoroacetophenone very inefficiently, with alcohol 7 being obtained in only 11% (Scheme 3c). The major product was hemiaminal 8, indicating that nucleophilic attack is favored by a large margin. Once again, this is consistent with our observation that trifluoroacetophenone is an inferior oxidant for N-lithium pyrrolidide, despite its significantly enhanced electrophilicity over benzophenone. In contrast, reduction outcompetes addition in the case of N-lithium piperidide, making trifluoroacetophenone a viable oxidant for this amide (Scheme 3d). Similar experiments were then conducted with t-butyl phenyl ketone, using crude NMR to determine yields. N-Lithium pyrrolidide reduced t-butyl phenyl ketone to alcohol 10 with excellent selectivity (Scheme 3e). For N-lithium piperidide, the outcome was comparable to the reaction with trifluoroacetophenone (Scheme 3f).

Regioselectivity of the oxidation

Having identified the amide/oxidant pairing as the key determinant of chemoselectivity, we next asked whether the same structural factors govern which α-C–H bond is abstracted in substrates that offer a choice. As mentioned above, we have shown that α-substituted alicyclic amines can be readily converted to α,α′-disubstituted alicyclic amines via regioselective oxidation. To quantify the site selectivity of the oxidation step, we conducted a range of experiments to probe the impact of the ketone oxidant and other reaction parameters (Table 1). Reactions conducted with t-butyl phenyl ketone at −78 °C provided synthetically useful overall yields for imines 11 and 12 with favorable ratios (entries 1–3). While raising the temperature to −40 or 0 °C had only a minimal impact on the reaction outcome (entries 4 and 5), conducting the reaction at room temperature led to a drastic reduction in yield and product ratio (entry 6). While MTBE was found to be an inferior solvent (entry 7), THF and a mixture of THF and ether provided slightly improved product ratios, albeit at the expense of overall yield (entries 8 and 9). While both benzophenone and trifluoroacetophenone provided good overall yields, product ratios were drastically reduced (entries 10–12). A range of other ketones were then evaluated, without providing substantial improvements. An exception is 1-adamantyl phenyl ketone (17), which provided an improved product ratio with only a minor drop in yield (entry 19). This ketone may be a useful oxidant for particularly challenging substrates.

Table 1. Regioselectivity for the oxidation of 2-phenyl pyrrolidine.

Entry	Ketone	Equiv.	Temperature (°C)	Solvent	Time [min]	rr 11 : 12	Yield 11 + 12 a (%)
1	PhCOtBu	1.2	−78	Et2O	10	6 : 1	60
2	PhCOtBu	1.2	−78	Et2O	60	6.9 : 1	63
3	PhCOtBu	1.7	−78	Et2O	10	6.6 : 1	76
4	PhCOtBu	1.2	−40	Et2O	10	5.9 : 1	62
5	PhCOtBu	1.2	0	Et2O	10	5.8 : 1	68
6	PhCOtBu	1.2	rt	Et2O	10	2.1 : 1	33
7	PhCOtBu	1.2	−78	MTBE	10	1.6 : 1	30
8	PhCOtBu	1.2	−78	THF	10	3.3 : 1	30
9	PhCOtBu	1.2	−78	Et2O/THF (1 : 1)	10	7 : 1	50
10	PhCOPh	1.2	−78	Et2O	10	2.1 : 1	78
11	PhCOPh	1.7	−78	Et2O	10	2 : 1	75
12	PhCOCF3	1.2	−78	Et2O	10	2.5 : 1	77
13	13	1.2	−78	Et2O	10	2 : 1	78
14	14	1.2	−78	Et2O	10	1 : 1.2	54
15	15	1.2	−78	Et2O	10	ND	6
16	15	1.2	−78	Et2O	60	1.7 : 1	27
17	16	1.2	−78	Et2O	10	—	NR
18	16	1.2	−78	Et2O	60	ND	4
19	17	1.2	−78	Et2O	10	10 : 1	66
20	18	1.2	−78	Et2O	10	1.3 : 1	37
21	19	1.2	−78	Et2O	10	1.3 : 1	80
22	20	1.2	−78	Et2O	10	1.1 : 1	40

^aNMR yields, regioselectivity ratio (rr) determined by crude NMR using 1,3,5-trimethoxybenzene as internal standard. ND = not determined, NR = no reaction.

A regioselectivity study was also conducted with 2-phenyl piperidine (6), although this turned out to be somewhat challenging due to the relative instability of imine 21. As shown in Scheme 4, oxidation of 6 with trifluoroacetophenone provided a favorable ratio of desired imine 21 over imine 22 (ratio 13.8 : 1). Interestingly, neither benzophenone nor t-butyl phenyl ketone led to any formation of 21 or 22 under otherwise identical reaction conditions (not shown).

Scheme 4. Regioselectivity for the oxidation of 2-phenyl piperidine.

To probe the potential impact of electronic factors on the regioselectivity of the oxidation, we compared reaction outcomes for two additional 2-aryl pyrrolidines (Scheme 5). Whereas the more electron-rich substrate 23 provided a poorer regioselectivity than 2-phenyl pyrrolidine (4), with a 3.3 : 1 ratio of imines 24 and 25 being observed (Scheme 5a), the more electron-poor substrate 26 furnished imines 27 and 28 in a more favorable ratio (7.6 : 1) (Scheme 5b). These examples clearly establish that electronic effects play a major role in the regioselectivity for substrates with essentially identical steric demands. The observed outcome tracks with the strength of the benzylic bond in either substrate, with the electronically more activated (more hydridic) benzylic bond in 23 providing more of the undesired imine regioisomer.

Scheme 5. Regioselectivity for electronically distinct 2-aryl pyrrolidines.

Probing the reactivity of the Li-amide addition product

As illustrated above, addition of the lithium amide to the ketone represents an undesirable side reaction that ultimately limits the yield of the intermediate imine, and thus the overall efficiency of the α-functionalization process. The tetrahedral intermediate formed upon addition of a lithium amide to benzophenone appears to be stable under the reaction conditions, as illustrated by Kowalski, Creary, and coworkers.¹⁹ This notion is consistent with our own results. Nevertheless, we were left wondering if the tetrahedral intermediate could still participate in a productive process that would lead to the imine. We thus decided to perform control experiments to probe this possibility. To cleanly form the requisite tetrahedral intermediate, we opted to add an aryl lithium nucleophile to N-benzoyl pyrrolidine (29). Instead of using phenyl lithium, we selected 4-methylphenyl lithium (30) to be able to distinguish the origin of the aryl group in any product (Scheme 6a). Addition of an excess of 30 to a solution of 29 at −78 °C led to the immediate formation of a precipitate, presumably due to the low solubility of putative intermediate 31. The reaction mixture was allowed to warm to room temperature and stirred for two hours. Following workup, α-aryl pyrrolidine 32 was isolated in 8% yield, in addition to mixed ketone 33 (74% yield) and the corresponding reduction product 34 (13% yield). These results clearly illustrate that 31 can in fact participate in imine formation, if only in a secondary process. Interestingly, in a related experiment with N-trifluoroacetyl piperidine (35) and phenyl lithium, no 2-phenylpiperidine was observed, illustrating the stability of the corresponding tetrahedral intermediate under these conditions (Scheme 6b).

Scheme 6. Conversion of the tetrahedral intermediate to α-functionalized product.

Kinetic isotope effects

To gain insights into the rate-determining step of the reaction, we performed a kinetic isotope effect study. Specifically, we performed an internal competition experiment with azacyclotridecane-2,2-d₂ (36) to form the phenylation product 37 (Scheme 7). Amine 36 avoids issues associated with volatility, simplifying its preparation, and was shown to exhibit a reactivity similar to pyrrolidine.^9a In the standard α-phenylation, a P_H/P_D value of 3 was obtained. This value is consistent with the k_H/k_D value of 3.4 that was determined by Wittig and coworkers for the oxidation of N-lithiated N-benzyl aniline with benzophenone.^8e Further, a similar isotope effect was observed in the related Meerwein–Ponndorf–Verley reaction which has been shown to proceed via hydride transfer.²² These collective observations establish that the rate limiting step of the reaction involves the breakage of the α-C–H bond.

Scheme 7. Kinetic isotope effects.

Potential for electron transfer in the formation of the imine

The isotope effect establishes that α-C–H cleavage is rate-limiting but does not by itself distinguish a concerted hydride transfer from a rapid stepwise sequence, prompting us to search for evidence of radical intermediates. When performing oxidations of lithium amides with benzophenone at −78 °C, we inconsistently observed the formation of a blueish solution immediately upon the addition of the ketone, with the color typically dissipating within a few seconds. This observation suggests the intermediacy of a ketyl radical but is also consistent with the initial formation of a Li-ate complex that undergoes rate-limiting, concerted hydride transfer as proposed by Wittig.^8e In an experiment that was conducted at −100 °C, the blueish color persisted. A solution prepared at this temperature and then frozen in liquid nitrogen provided an EPR signal identical to that obtained when metallic lithium was added to benzophenone in ether at room temperature, followed by freezing in liquid nitrogen. These observations seemed to provide strong evidence for the formation of the lithium benzophenone ketyl radical in the early stages of the reaction. However, closer inspection of the reaction parameters indicated that ketyl radical formation was in fact due to a small excess of n-butyllithium present in some experiments.²³n-Butyllithium is known to reduce benzophenone via SET, and small quantities of the resulting ketyl radical are readily detectable by EPR.¹⁷ When the deprotonation of pyrrolidine was conducted with n-butyllithium as the limiting reagent (pyrrolidine/n-butyllithium ratio 2 : 1), followed by addition of benzophenone, no evidence for the formation of ketyl radicals was obtained. While recent studies have shown the feasibility of electron-transfer from lithium amides to ketones, to our knowledge, this type of reactivity has only been observed in lithium amides lacking α-hydrogens.²⁴

Computational studies

Taken together, the experiments above indicate that hydride transfer and N-addition are closely balanced processes whose relative rates depend on the structures of both the lithium amide and the ketone. We turned to computational chemistry to evaluate the mechanism of this transformation and rationalize several of our experimental observations. The reaction between lithium amides and ketones was modeled using M062X/6-31+G(d)²⁵ level of theory with a PCM²⁶ (diethyl ether) model to account for the effect of solvent on the reaction energetics. The Gibbs free energy barriers reported herein are relative to separated starting materials corresponding to each transition structure. Our initial goal was to evaluate the active lithium species involved in the reaction. Next, we sought to account for the product ratio (hydride transfer versus N-addition) in the reactions between N-lithiated piperidine/pyrrolidine with the two electronically and sterically different ketones – benzophenone and trifluoroacetophenone. Finally, we sought to evaluate the observed regioselectivity in the reaction between 2-phenyl pyrrolidine and t-butyl phenyl ketone.

Identification of most probable lithium amide species

We initiated our theoretical study by modeling the hydride transfer step from N-lithiated pyrrolidine to benzophenone (BP). The transition state corresponding to this step (Fig. 3, BP-Pyr-HTS) has a ΔG^‡ of 8.5 kcal mol⁻¹. We then modeled the possibility that hydride transfer occurs from a dimeric N-lithiated pyrrolidine species (the solvated dimer was suggested by Collum as the active species in solution²⁷); however, this TS (Fig. 3, BP-Pyr-HTS_dimer) has a ΔG^‡ of 12.1 kcal mol⁻¹ suggesting that this is likely not the species involved in hydride transfer in our reaction. Next, we modeled the BP-Pyr-HTS with a molecule of diethyl ether coordinated to Li (Fig. 4, BP-Pyr-HTS_Et2O) and found this structure to have the lowest ΔG^‡ (among the three models) of 6.3 kcal mol⁻¹. While the key distances between Li and the pyrrolidine nitrogen/benzophenone oxygen remain similar in BP-Pyr-HTS (Fig. 3) and BP-Pyr-HTS_Et2O (Fig. 4), the hydride transfer is notably earlier in BP-Pyr-HTS_Et2O suggesting that inclusion of explicit solvent on Li enhances the hydridic nature of N-lithiated pyrrolidine.

Fig. 3. DFT models for the reduction of benzophenone by monomeric and dimeric N-lithiated pyrrolidine computed at M062x/6-31+G* PCM(Et₂O).

Fig. 4. DFT models comparing lowest energy hydride transfer transition states to the lowest energy N-addition transition states for benzophenone with N-lithiated pyrrolidine (blue pathways) and N-lithiated piperidine (maroon pathways) computed at M062X/6-31+G(d) PCM(Et₂O). All models incorporate an explicit solvent (diethyl ether) molecule bound to the lithium atom. Some hydrogen atoms have been removed for clarity, and all key bond-forming distances and stabilizing interactions are indicated in angstroms.

Evaluation of hydride transfer versus N-addition

Having established that BP-Pyr-HTS_Et2O is the likely TS for hydride transfer, we sought to evaluate the relative energies of reduction via hydride transfer versus the competing reaction i.e. N-addition of N-lithiated amine to ketone for benzophenone (BP) and trifluoroacetophenone (TF) (Scheme 3). For benzophenone, the transition state for N-addition of N-lithiated pyrrolidine (Fig. 4, BP-Pyr-NTS_Et2O) has a ΔG^‡ of 8.8 kcal mol⁻¹. The predicted ΔΔG^‡ of 2.5 kcal mol⁻¹ (favoring hydride transfer) is an overestimation of, but qualitatively consistent with, the experimental 68 : 32 ratio of the reduced : N-addition product in this reaction (Scheme 3). Next, we sought to understand the reversal in chemoselectivity in the reaction of BP with N-lithiated piperidine – Scheme 3 shows that N-addition is the preferred product in this reaction. Accordingly, we located BP-Pip-HTS_Et2O (Fig. 4, ΔG^‡ = 12.8 kcal mol⁻¹) and BP-Pip-NTS_Et2O (Fig. 4, ΔG^‡ = 10.2 kcal mol⁻¹) and found the ΔΔG^‡ to be 2.6 kcal mol⁻¹ (favoring N-addition) – which is again qualitatively consistent with the experimental 10 : 90 ratio of the reduced : N-addition product in this reaction. To understand the origin of reversal of chemoselectivity with the two N-lithiated amines, we performed an analysis of the relevant TSs in Fig. 4. The hydride transfer TSs for the two N-lithiated amines are almost identical with the lithium bound to both the amine nitrogen and the carbonyl oxygen (in both cases) as the hydride transfer TS takes place. Therefore, the chemoselectivity observed in these reactions is likely not resulting from differences in the hydride transfer TSs of the two lithiated amines. A key difference between the hydride transfer and N-addition TSs is that the lithium is not coordinated to the carbonyl oxygen atoms in the N-addition TS. Further analysis of the N-addition TSs reveals that BP-Pip-NTS_Et2O is a ‘later’ TS (r_C–N = 2.12 Å) than the BP-Pyr-NTS_Et2O (r_C–N = 2.47 Å). This causes the lithium to have a significantly stronger Li–π interaction with one of the phenyl rings of benzophenone in BP-Pip-NTS_Et2O compared to BP-Pyr-NTS_Et2O (Li–π distances indicated in Fig. 4). This lowers the energy of the BP-Pip-NTS_Et2O to where it is below the energy of the BP-Pip-HTS_Et2O (N-addition is experimentally observed). Since this interaction is weaker in the pyrrolidine, BP-Pyr-HTS_Et2O is lower in energy than BP-Pyr-NTS_Et2O (hydride transfer is observed experimentally).

For TF, the chemoselectivity is reversed compared to BP (Scheme 3) – with pyrrolidine giving N-addition (reduced : N-addition = 11 : 89) and piperidine giving the reduced ketone (reduced : N-addition = 71 : 29) as the favored product. To explain this intriguing reversal, we modeled the relevant TSs for the reaction of TF and N-lithiated piperidine/pyrrolidine (Fig. 5). The first key observation is that the more electrophilic ketone TF (computed electrophilicity of the carbonyl carbon of TF is 2.33 compared to 1.96 for BP) results in lower barriers for each of the four TSs (Fig. 5) compared to the corresponding TSs for BP. Secondly, similar to BP, the hydride transfer TSs for both lithiated amines are characterized by lithium coordinated to the amine nitrogen, carbonyl oxygen, and an ether molecule as the hydride is delivered to the carbonyl carbon. Comparison of the two N-addition TSs (Fig. 5, TF-Pyr-NTS_Et2O and TF-Pip-NTS_Et2O) reveals largely identical features with relatively long C–N bond-forming distances (r_C–N = 2.75 Å and 2.76 Å, Fig. 5). Due to a slight difference in the orientation of the nitrogen lone pairs of pyrrolidine versus piperidine, the lithium that is coordinated to the nitrogen is in a slightly different position in the two TSs. In the pyrrolidine TS, the lithium forms a strong interaction with one of the fluorine atoms of the CF₃ group (2.11 Å), while in the piperidine TS this interaction is significantly weaker (2.43 Å). Since this is an interaction between a highly electronegative and highly electropositive atom these differing distances have a significant impact on the energies of the N-addition TS. The pyrrolidine TS, by virtue of the stronger Li–F interaction becomes lower in energy than the hydride transfer TS (ΔΔG^‡ = 2.2 kcal mol⁻¹, favoring N-addition), while the piperidine TS with its weaker Li–F interaction remains higher in energy than the hydride transfer TS (ΔΔG^‡ = 1.1 kcal mol⁻¹, favoring hydride transfer) – consistent with the experimental chemoselectivity (vide supra). Similar Li–F interactions have previously been noted.²⁸

Fig. 5. DFT models comparing lowest energy hydride transfer transition states to the lowest energy N-addition transition states for trifluoroacetophenone (TF) with N-lithiated pyrrolidine (blue pathways) and N-lithiated piperidine (maroon pathways) computed at M062X/6-31+G(d) PCM(Et₂O). All models incorporate an explicit solvent (diethyl ether) molecule bound to the lithium atom. Some hydrogen atoms have been removed for clarity, and all key bond-forming distances and stabilizing interactions are indicated in angstroms.

Evaluation of the regioselectivity for the reaction of 2-phenyl pyrrolidine with ketones

As a final step in our computational study, we sought to evaluate the origin of the observed regioselectivity in the reactions of N-lithiated 2-phenyl pyrrolidine with BP (2.1 : 1), TF (2.5 : 1), and t-butyl phenyl ketone (tBu) (6 : 1). To this end, we modeled the hydride transfer event from the 2-position and the 5-position of 2-phenyl pyrrolidine to each of these ketones. We used the diethyl ether solvated N-lithiated 2-phenyl pyrrolidine as the active species in each of these reactions and found the difference between hydride transfer from the 5-position and the 2-position to be 0.9 kcal mol⁻¹ for BP and 0.6 kcal mol⁻¹ for TF (see the SI), consistent with the modest regioselectivity for these two ketones. For the most selective ketone tBu, the difference between the two relevant hydride transfer TSs – tBu-2PP-Major_Et2O and tBu-2PP-Minor_Et2O – is 1.3 kcal mol⁻¹ (Fig. 6). These differences are qualitatively consistent with experimental results and suggest that the regioselectivity is primarily driven by steric interactions between the two species.

Fig. 6. DFT models showing relative energies of the regioselectivity-determining transition states (solvated) involving the reduction of t-butyl phenyl ketone (tBu) by the hydride from the 5-position versus 2-position of N-lithiated 2-phenyl pyrrolidine computed at M062x/6-31+G* PCM(Et₂O). All models incorporate an explicit solvent (diethyl ether) molecule bound to the lithium atom. Some hydrogen atoms have been removed for clarity and all key bond-forming distances and stabilizing interactions are indicated in angstroms.

Conclusions

In summary, our combined experimental and computational investigations provide a detailed mechanistic understanding of the ketone-mediated oxidation of alicyclic lithium amides, a key step enabling the direct α-C–H functionalization of unprotected secondary amines. We establish that imine formation proceeds through a low-barrier concerted hydride transfer from solvent-coordinated monomeric lithium amides, with no evidence for primary SET or HAT pathways under the optimized conditions. Kinetic isotope effects confirm α-C–H bond cleavage as the rate-determining step, while ReactIR monitoring reveals that oxidation occurs rapidly at −78 °C without detectable imine degradation. Our studies further identify nucleophilic addition of lithium amides to the ketone oxidant as a major competing pathway. DFT calculations attribute this behavior to differential transition-state stabilization, including Li–π interactions in reactions with benzophenone and Li–F interactions with trifluoroacetophenone, which together rationalize unusual ring-size-dependent chemoselectivity trends. Regioselectivity in α-substituted amines is governed by a combination of steric effects and electronic modulation of C–H bond strengths, and the use of sterically demanding oxidants such as adamantyl phenyl ketone provides improved selectivity in challenging cases. Importantly, while the mechanistic framework established here is expected to be general, the balance between hydride transfer and N-addition is highly sensitive to the structures of both partners and cannot be predicted a priori. When applying this chemistry to a lithium amide that has not previously been examined, the ketone oxidant should therefore be matched to that amide, if necessary by means of the simple competition experiments described above (Scheme 3). More broadly, this work resolves long-standing mechanistic questions surrounding lithium amide–ketone reactivity and establishes a framework for expanding protecting-group-free strategies for the synthesis of polysubstituted alicyclic amines, motifs that are central to modern medicinal chemistry.

Author contributions

D. S. and M. J. V. conceptualized the study. D. A. V. and S. D. performed the bulk of the experiments with assistance by Su. D. and W. C., under D. S. supervision. S. B., M. B., and M. J. V. performed the computational studies. D. S. and M. J. V. wrote the manuscript with input from all authors.

Conflicts of interest

There are no conflicts to declare.

Data availability

Supplementary information (SI): all experimental procedures, spectroscopic data, and computational details. See DOI: https://doi.org/10.1039/d6qo01213b.