Lithium Diisopropylamide-Mediated Reactions of Imines, Unsaturated Esters, Epoxides, and Aryl Carbamates: Influence of Hexamethylphosphoramide and Ethereal Cosolvents on Reaction Mechanisms

Abstract

Several reactions mediated by lithium diisopropylamide (LDA) with added hex-amethylphosphoramide (HMPA) are described. The N-isopropylimine of cyclohex-anone lithiates via an ensemble of monomer-based pathways. Conjugate addition of LDA/HMPA to an unsaturated ester proceeds via di- and tetra-HMPA-solvated dimers. Deprotonation of norbornene epoxide by LDA/HMPA proceeds via an intermediate metalated epoxide as a mixed dimer with LDA. Ortholithiation of an aryl carbamate proceeds via a mono-HMPA-solvated monomer-based pathway. Dependencies on THF and other ethereal cosolvents suggest that secondary-shell solvation effects are important in some instances. The origins of the inordinate mechanistic complexity are discussed.

Introduction

Hexamethylphosphoramide (HMPA) is one of the most prominent additives used to influence the yields, rates, and selectivities of organolithium reactions.¹ A preponderance of what is known about solvation of lithium ions by HMPA derives from the studies of Reich and coworkers.² Their spectroscopic analyses of lithium salts in the limit of slow exchange of free and lithium-ion-coordinated HMPA offer intimate details of HMPA-mediated deaggregation and ionization. Despite the sound understanding of how HMPA influences the structures of lithium salts, the oft-cited influence of HMPA on reactivity is not well understood and leaves many questions unanswered.³ Does the marked tendency of HMPA to serially solvate organolithiums foreshadow a similar structural diversity in the rate-limiting transition structures? Do dramatic accelerations result from the capacity of HMPA to promote high solvation numbers and low aggregation numbers in the rate-limiting transition structures? Why does HMPA sometimes fail to elicit high reactivities?²

We have begun addressing some of these questions in the context of lithium diisopropylamide (LDA),^4-9 a prevalent base in organic synthesis.¹⁰ We describe herein investigations of four reactions mediated by LDA/HMPA--lithiation of imine 1 (eq 1),^7f,8b,11,12 1,4-addition to unsaturated ester 3 (eq 2),^13,14 opening of epoxide 5 (eq 3),^10,15-17 and ortholithiation-Fries rearrangement of carbamate 7 (eq 4).¹⁸ All have found niches in organic synthesis, and all but the 1,4-addition have proved useful in studies of LDA structure-reactivity relationships.¹⁹ The results support an emerging picture of unusual mechanistic diversity imparted by HMPA;^4,5 they are summarized at the start of the discussion for the benefit of the non-specialist.

(1)

(2)

(3)

(4)

Background

Any examination of structure-reactivity relationships must be prefaced by a clear understanding of structure.^19,20 LDA offers an optimal template in that it forms exclusively disolvated dimer 9 over all THF concentrations and disolvated dimer 10 at ≥ 1.0 equiv of HMPA.^6a,21 (The absence of deaggregation is notable given that HMPA usually deaggregates organolithiums.²)

A brief survey of previous rate studies of LDA/HMPA-mediated reactions is instructive. LDA/HMPA-mediated reactions afford considerable mechanistic variability.^4,5 For example, enolizations of a hindered ester by LDA/THF and LDA/HMPA/THF proceed at similar rates by distinctly different mechanisms.^4b The sole detectable pathway in neat THF involves disolvated monomers (11), whereas HMPA causes enolization to proceed via monosolvated monomers (12) and putative triple ions (13). (Three dimensional depictions of transition structures throughout this paper derive from computational studies,^7,8 analogies with observable structural forms,^4,6,7f and conjecture.)

Dehydrobrominations of alkyl bromides proceed by mono-, di- and trisolvated monomers (14-19) as well as triple ions (20).⁵ The highly variable solvation numbers are unusual when compared with LDA-mediated reactions in standard ethereal solvents.¹⁹

A seemingly minor point has baffled us. LDA/HMPA-mediated dehydrobrominations are insensitive to the proportions of THF in THF/hexane cosolvent. Conversely, enolization proceeding via putative triple ion 13 is inhibited by THF.^4b A model based on the solvation of free (uncoordinated) HMPA by THF²² with an affiliated net stabilization of the ground state was postulated,^4b but this explanation was subsequently dismissed as incorrect or at least inadequate.⁵ The dual role of ethers as both ligands and media continues to be vexing.^23,24

Results

General Methods

Reactions carried out under standard conditions using 1.0-4.0 equiv of [⁶Li,¹⁵N]LDA^6a confirm that a number of mixed aggregates are formed, as described in the context of each case study.^{25 6}Li and ¹⁵N NMR spectroscopic data are summarized in Table 1. To avoid autoinhibition often caused by mixed aggregation^4c,26 during the rate studies, pseudo-first-order conditions were established by using substrates at low concentrations (0.004 M). LDA, HMPA, and THF were maintained at high, yet adjustable, concentrations using inert cosolvents.²⁷ The loss of 5 was monitored using gas chromatography relative to an internal dodecane standard.⁹ The loss of 1, 3, and 7 were monitored using in situ IR spectroscopy.^4,28 All follow first-order decays, affording pseudo-first-order rate constants (k_obsd) that are independent of the initial concentrations of substrate (±10%).²⁹ Isotope effects (k_H/k_D) determined using deuterated analogues 1-d₄ and 7-d₅ (Table 2) are consistent with rate-limiting proton transfers for 1 and 7. The isotope effect obtained with epoxide 5-d₂ is unusually small, emblematic of a proton transfer that is not rate limiting.

Table 1.

⁶Li and ¹⁵NMR spectroscopic data.^a

compd	δ 6Li(mult, JLiN)	δ15N (mult, JLiN)
21	0.72 (s)	---
27a	0.88 (d, 5.2)	74.8 (q, 5.4)
27b	1.03 (d, 5.2)	76.7 (q, 5.2)
38	0.88 (d, 5.0)	76.0 (q, 5.0)
39	0.91 (d, 5.1)	75.0 (q, 5.0)
43	0.78 (d, 5.2)	76.4 (q, 5.2)

^aSpectra were recorded in THF solutions of 0.10 M [⁶Li,¹⁵N]LDA total lithium titer, 0.40 M total HMPA (bound and unbound), and 0.025 M substrate. Coupling constants are reported in hertz. Multiplicities are denoted as follows: s = singlet, d = doublet, t = triplet, q = quintet. The chemical shifts are reported relative to 0.30 M ⁶LiCl/MeOH at -90 °C (0.0 ppm) and neat Me₂NEt at -90 °C (25.7 ppm). Spectra also contained [⁶Li,¹⁵N]LDA dimer 10: ⁶Li NMR δ 1.64 (t, 4.9); ¹⁵N NMR δ 73.3 (q, 4.8).

Table 2.

Summary of rate studies for the LDA-mediated reactions (eq 1-4)

Substrate	T (°C)	solvent	THF order	HMPA order	LDA order	kH/kD
1	20	THF	0	---	0.54±0.02a	10±1b
1	-55	THF/HMPA	2.4±0.2	0	0.60±0.03c	----
			0	2.0±0.3d	0.54±0.07e	7.8±0.1e
3	-78	THF/HMPA	2.0±0.3	0	1.0±0.1f	---
			--g	1.87±0.06h	1.10±0.06i	---
5	20	THF	0	---	0.73±0.04j	3.6±0.4k
5	0	THF/HMPA	0	0	1.0±0.1a	1.3±0.1f
				-0.6±0.2
7	-40	THF	0	---	0.49±0.01	16k
7	-40	THF/HMPA	0	0	0.51+0.04	14.2±0.2l

^a[LDA] = 0.10 M in 10.0 M THF/hexane.

^bSee ref 8b.

^c[HMPA] = 0.10 M in 8.0 M THF/hexane.

^d[LDA] = 0.10 M; [THF] = 8.0 M in hexane cosolvent.

^e[HMPA] = 0.40 M in 8.0 M THF/hexane.

^f[HMPA] = 0.40 M in 10.0 M THF/hexane.

^gThe order in THF could not be measured at high HMPA concentration due to insolubility.

^h[LDA] = 0.10 M; [THF] = 10.0 M in hexane cosolvent.

ⁱ[HMPA] = 0.10 M in 10.0 M THF/hexane.

^j[LDA] = 0.10 M; [THF] = 10.0 M in hexane cosolvent.

^kSee ref 43.

^l[LDA] = 0.10 M; [HMPA] = 0.50 M in 10.0 M THF/hexane.

Selected rate data are depicted in Figures 1-6, and the reaction orders constituting the rate laws are summarized in Table 2. Rate data for the LDA/THF-mediated lithiations of 1 and 7 in the absence of HMPA reported previously are also included in Table 2 for comparison.^8b,26 Rate studies of LDA/THF-based lithiation of 5^9a in THF with no added HMPA have not been reported previously and are described below.

Figure 1.

Plot of k_obsd vs [HMPA] for the lithiation of imine 1 (0.005 M) by 0.10 M LDA in THF/hexane at -55 °C: (A) 2.0 M THF; (B) 8.0 M THF. Curve A derives from a linear least squares fit. Curve B depicts the result of an unweighted least-squares fit to k_obsd = k[HMPA]ⁿ + k′ (k = (1.3 ± 0.1) × 10^-3; k′ = (1.0 ± 0.2) × 10^-4; n = 2.0 ± 0.3).

Figure 6.

Plot of k_obsd vs [HMPA] for the deprotonation of epoxide 5 (0.004 M) by 0.10 M LDA in THF (10.0 M)/hexane at 0 °C. The curve depicts the result of an unweighted least-squares fit to k_obsd = k [HMPA]ⁿ + k′ (k = (3.3 ± 1.9) × 10^-4; k′ = (1.3 ± 2) × 10^-2; n = -0.6 ± 0.2).

Imine Lithiations

Previous ⁶Li and ¹⁵N NMR spectroscopic studies indicate that lithiated imine 2 is a monomer in THF.³⁰ Analogous studies in which imine 1 is metalated by a modest excess of [⁶Li,¹⁵N]LDA (2.0-4.0 equiv) in the presence of 0.40 M HMPA reveal LDA dimer 10 along with a species displaying a singlet consistent with (but not rigorously assigned to) monomer 21. Mixed aggregates are not formed.

LDA/THF-mediated metalations of imine 1 proceed via mono-solvated monomer (22).^8b,11 Conversely, the LDA/HMPA/THF-mediated metalation of imine 1 displays a stifling complexity. A plot of k_obsd versus HMPA concentration shows a first-order HMPA dependence in 2.0 M THF and a second-order HMPA dependence in 8.0 M THF (curves A and B, respectively, in Figure 1). The HMPA concentration-independent pathway--the non-zero intercepts in curves A and B--is also independent of the THF concentration. The THF concentration dependence can be illustrated from a different perspective in which the HMPA concentration is held constant and the THF concentration is varied (Figure 2). At low HMPA concentration the rates are independent of the THF concentration (curve A), whereas at high HMPA concentration an exponential (2.4 ± 0.4 order) dependence on the THF concentration is evident (curve B).

Figure 2.

Plot of k_obsd vs [THF] for the lithiation of imine 1 (0.005 M) by 0.10 M LDA in hexane with HMPA at -55 °C: (A) 0.10 M free HMPA in cyclopentane; (B) 0.50 M free HMPA in cyclopentane; (C) 0.50 M free HMPA in 2,5-Me₂THF. Curves A and C derive from linear least squares fits. Curve B depicts the result of an unweighted least-squares fit to k_obsd = k [THF]ⁿ + k′ (k = (7.6 ± 0.8) × 10^-7; k′ = (3.0 ± 0.1) × 10^-4; n = 2.4 ± 0.4).

The solvent dependencies are consistent with three independent terms in the idealized³¹ partial rate law described by eq 5. We wondered whether the THF dependence derives from a sterically sensitive, primary-shell solvation by THF or a sterically insensitive secondary-shell effect.^4b,23,24 Compelling evidence of secondary-shell solvation was uncovered in a plot of k_obsd versus THF concentration using the weakly coordinating 2,5-dimethyltetrahydrofuran (2,5-Me₂THF) as cosolvent (Figure 2, curve C). When the polarity of the medium is held constant using a poorly coordinating cosolvent,³² the apparent second-order THF dependence disappears.³³ Thus, although the rate law is formally described by eq 5, we choose to cull out the apparent secondary-shell solvation effects and provide the simplified version described by eq 6.

$$k_{\mathrm{obsd}}=k_3+k_4\left[\mathrm{HMPA}\right]+k_5{\left[\mathrm{HMPA}\right]}^2{\left[\mathrm{THF}\right]}^2$$ (5)

$$k_{\mathrm{obsd}}=k_3+k_4\left[\mathrm{HMPA}\right]+k_5{\left[\mathrm{HMPA}\right]}^2$$ (6)

Plots of k_obsd versus LDA concentration reveal half-order LDA dependencies at both low and high HMPA concentrations consistent with monomer-based metalations (Figure 3). The fractional LDA orders in conjunction with HMPA and ethereal cosolvent dependencies are consistent with the idealized rate law³¹ in eq 7. A literal interpretation of eq 7 implicates transition structures [(i-Pr₂NLi)(HMPA)(1)]^‡, [(i-Pr₂NLi)(HMPA)₂(1)]^‡, and [(i-Pr₂NLi)(HMPA)₃(1)]^‡ for which we offer depictions 23, 24, and 25, respectively; however, we advise caution against over interpreting such a three-term rate law.

$$\text{-}d\left[\mathbf{1}\right]/d\mathrm{t}=\left\{k^{\prime }{\left[\mathrm{LDA}\right]}^{1/2}+k^{″}\left[\mathrm{HMPA}\right]{\left[\mathrm{LDA}\right]}^{1/2}+k^{‴}{\left[\mathrm{HMPA}\right]}^2{\left[\mathrm{LDA}\right]}^{1/2}\right\}\left[\mathbf{1}\right]$$ (7)

Figure 3.

Plot of k_obsd vs [LDA] for the lithiation of imine 1 (0.005 M) in 0.40 M free HMPA/8.0 M THF/cyclopentane at -55 °C. The curves depicts the result of an unweighted least-squares fit to k_obsd = k[LDA]ⁿ (k = (1.4 ± 0.2) × 10^-3; n = 0.54 ± 0.07).

Conjugate Additions

LDA in THF solution undergoes conjugate addition to 3 in lieu of either α- or γ-deprotonation.^13a However, the results are highly unusual and part of an emerging mechanistic story to be discussed in another context.³⁴ LDA/HMPA-mediated conjugate additions proceed smoothly and provide tractable rate data as follows.

Conjugate addition of LDA to unsaturated ester 3 in HMPA/THF (eq 2) affords β-amino ester 26 in 76% yield. Analogous addition with 2.0-4.0 equiv of [⁶Li,¹⁵N]LDA forms two mixed dimers that we believe are the two geometric isomers (27a,b). Enolization of amino ester 26 with 2.0-4.0 equiv of [⁶Li,¹⁵N]LDA affords only one isomer. The stereochemical assignments of the E and Z isomers has not been made. Enolization with 1.25 equiv of LDA affords enolate 4 in an unknown aggregation state.³⁵

Rate studies carried out under pseudo-first-order conditions afforded evidence of several mechanisms. A plot of k_obsd versus HMPA concentration reveals a second-order HMPA dependence in 10.0 M THF (Figure 4). An HMPA concentration-independent pathway is evidenced by a significant non-zero y-intercept. Plots of k_obsd versus LDA concentration at both low and high HMPA concentrations reveal first-order dependencies, indicating that both the HMPA-independent and HMPA-dependent pathways are dimer based. In principle, transition structures [(R₂NLi)₂(HMPA)₄]^‡ (30) and [(R₂NLi)₂(HMPA)₂]^‡ (possibly 28 or 29) are consistent with the rate data. However, second-order dependencies on the THF concentration with distinct non-zero intercepts at both low and high HMPA concentrations (Figure 5, curve A) suggest two additional transition structures corresponding to [(R₂NLi)₂(HMPA)₂(THF)₂]^‡ and [(R₂NLi)₂(HMPA)₄(THF)₂]^‡.³⁶ One could imagine describing the former transition structure as mixed-solvated triple ion 31. The high solvation number of [(R₂NLi)₂(HMPA)₄(THF)₂]^‡--formally a hexasolvated dimer--stretches our imaginations. A plot of k_obsd versus THF concentration at low HMPA concentration (0.10 M) using 2,5-Me₂THF as the cosolvent to hold the polarity of the medium relatively constant (Figure 5, curve B) simply adds further confusion by revealing a first-order THF dependence rather than the second-order dependence noted in hexane.

Figure 4.

Plot of k_obsd vs [HMPA] for the Michael addition to ester 3 (0.004 M) by 0.10 M LDA in THF (10.0 M)/hexane at -78 °C. The curve depicts the result of an unweighted least-squares fit to k_obsd = k[THF]ⁿ + k′ (k = (7.1 ± 0.3) × 10^-3; k′ = (8.2 ± 0.2) × 10^-2; n = 1.87 ± 0.06).

Figure 5.

Plot of k_obsd vs [THF] for the Michael addition to ester 3 (0.004 M) by 0.10 M LDA with 0.10 M free HMPA at -78 °C in (A) 2,5-Me₂THF as cosolvent; (B) hexane as cosolvent. Curve A derives from a linear least squares fit. Curve B depicts an unweighted least-squares fit to k_obsd= k[THF]ⁿ + k′ (k = (3.8 ± 0.5) × 10^-6; k′ = (3.8 ± 0.3) × 10^-4; n = 2.0 ± 0.4).

Epoxide Metalation

We examined the LDA/HMPA-mediated metalation of several epoxides and confirmed16a,b that epoxides 32 and 33 are unreactive even at ambient temperatures, above which the integrity of LDA/HMPA/THF becomes suspect.^37,38 Metalation of epoxide 5 in HMPA/THF (eq 3) affords alcohol 34 deriving from a transannular C-H insertion by a carbenoid intermediate in 80% isolated yield.

LDA/THF-mediated metalations of epoxide 5 display an odd (0.73 ± 0.04) fractional order in LDA and a clean zeroth-order dependence on the THF concentration. The idealized rate law (eq 8)³¹ is consistent with reaction via mono-solvated monomers (eq 9) and disolvated dimers (eq 10). (Previous studies of the LDA-mediated metalation of epoxide 5 using chelating ligands showed no evidence of a monomer-based pathway.) Transition structures 35-37 seem reasonable,^7,8,39 although the deviation from an optimal 180° C-H-N angle appears significant in 35.^7,9a,40

$$­d\left[\mathbf{5}\right]/d\mathrm{t}=k_1{\left[\mathrm{THF}\right]}^0{\left[\mathrm{LDA}\right]}^{1/2}\left[\mathbf{5}\right]+k_2{\left[\mathrm{THF}\right]}^0{\left[\mathrm{LDA}\right]}^1\left[\mathbf{5}\right]$$ (8)

$$1/2{\left(i­{\mathrm{Pr}}_2\mathrm{NLi}\right)}_2{\left(\mathrm{THF}\right)}_2+\mathbf{5}\to {\left[{\left(i­{\mathrm{Pr}}_2\mathrm{NLi}\right)}_2\left(\mathrm{THF}\right)\left(\mathbf{5}\right)\right]}^{‡}$$ (9)

$${\left(i­{\mathrm{Pr}}_2\mathrm{NLi}\right)}_2{\left(\mathrm{THF}\right)}_2+\mathbf{5}\to {\left[{\left(i­{\mathrm{Pr}}_2\mathrm{NLi}\right)}_2{\left(\mathrm{THF}\right)}_2\left(\mathbf{5}\right)\right]}^{‡}$$ (10)

Metalation of epoxide 5 (0.25 M) using [⁶Li,¹⁵N]LDA (0.10 M) with HMPA (0.40 M) reveals a mixed dimer that is replaced by another mixed dimer. The latter was confirmed to be 38 by mixing [⁶Li,¹⁵N]LDA with alcohol 34. The first-formed mixed dimer is believed to be 39 derived from the observable lithiation before carbenoid formation/C-H insertion.⁴¹ Quenching with D₂O failed to afford significant levels of deuterated 34, which is not particularly surprising in light of Seebach’s studies of D₂O quenches when i-Pr₂NH is present. ⁴² In the absence of HMPA, little or no metalated epoxide is observed (although some minor ⁶Li resonances are noted.)

The intermediacy of 39 introduces an inordinate complexity that was unappreciated when the solution kinetics were investigated. Because we monitored the metalation using GC analysis of quenched samples, mixed dimer 39 registered as starting material rather than product. Consequently, the measured reaction orders listed in Table 2 are not easily interpreted. The small isotope effect (k_H/k_D = 1.3 ± 0.1) makes sense if the metalation is reversible. The observed reaction displays a net inhibition by HMPA (Figure 6), but the metalated intermediate causes us to resist further interpretation of the rate data at this point. These suspect data are archived in supporting information.

Carbamate Ortholithiation

Previous studies of the LDA/THF-mediated ortholithiation of carbamate 7 show a rate-limiting metalation via monosolvated monomer (40) followed by a rapid (post-rate-limiting) anionic Fries rearrangement. Formation of mixed dimer 42 causes a marked autoinhibition when only 1.0 equiv of LDA is used. Analogous behavior is observed in the presence of HMPA: mixed dimer 43 is observed to the exclusion of any intermediate aryllithium derivatives. Under pseudo-first-order conditions a large k_H/k_D confirms a rate-limiting proton transfer. (The particularly large isotope effect is characteristic of ortholithiations.⁴³) A first-order HMPA concentration dependence and a half-order LDA concentration dependence implicate disolvated monomer-based metalation (41). A zeroth-order dependence on the THF concentration shows that medium effects are unimportant.

Discussion

We studied four LDA-mediated reactions--lithiation of imine 1 (eq 1), conjugate addition to unsaturated ester 3 (eq 2), α-deprotonation of epoxide 5 (eq 3), and ortholithiation of carbamate 7 (eq 4). All are important in organic synthesis, and HMPA has played a role in controlling reactivity in each case. The results are summarized below and discussed in more detail subsequently. We must reiterate an important point: The rate studies provide only the stoichiometries of the transition structures at the rate-limiting steps;²⁰ the three-dimensional renditions of the transition structures are based on computational studies,^7,8 analogies with observable structural forms,^6,21,32a and conjecture. We routinely offer this caveat in the context of rate data, but it seems especially germane in the context of LDA/HMPA-mediated reactions.

Summary

The rate studies were prefaced by structural studies showing aggregate changes throughout the reaction coordinate when moderate excesses of LDA/HMPA are used. LDA/HMPA-mediated metalation of imine 1 affords an LDA-free lithiated imine believed to be monomer 21 based on prior studies.³⁰ Analogous reaction of epoxide 3 forms low concentrations of an intermediate lithiated epoxide as mixed dimer 39. Subsequent carbenoid-derived insertion leads to LDA-alkoxide mixed aggregate 38. Michael addition of LDA/HMPA to unsaturated ester 5 provides mixed dimer 27 as a putative E-Z mixture. Ortholithiation of aryl carbamate 7 affords an undetectable aryllithium that undergoes facile (post-rate-limiting) Fries rearrangement to give LDA-aryloxide mixed dimer 43 as the only observable product. Previous studies of mixed aggregation in LDA/HMPA mixtures are consistent with this picture of a highly salt-dependent penchant toward mixed aggregation.^4c,6b,7e

Rate studies were carried out under pseudo-first-order conditions to preclude mixed aggregation effects. The intermediacy of metalated epoxide 39 proved disruptive to detailed rate studies. For the most part, however, the results are tractable and reveal considerable mechanistic diversity. The influence of THF in all LDA/HMPA-mediated reactions is discussed in a subsequent section.

Metalation of imine 1 by LDA/HMPA is mechanistically complex, displaying concentration dependencies implicating several monomer-based pathways (23-25). This complexity contrasts with analogous metalations in the absence of HMPA in which a single pathway involving a monosolvated monomer (22) is detected. LDA/HMPA-mediated conjugate addition to ester 3 proceeds via disolvated and tetrasolvated dimer-based pathways for which we depict 28 or 29 for the former and triple ion 30 for the latter. LDA/HMPA-mediated metalation of epoxide 5 was only marginally informative because of the formation of an observable lithiated epoxide (39). Even the metalation in the absence of HMPA, however, proved to be quite complex, implicating a combination of monosolvated monomer (35) and disolvated dimer (36 or 37). Last, the ortholithiation of carbamate 7 proceeds via a di-HMPA-solvated monomer (41).

Are mixed and secondary-shell solvation important?

Secondary-shell solvation--the influence of solvent as simply a medium--has received scant attention in organolithium chemistry.^4,23,24 Extensive investigations of LDA-mediated metalations have generally revealed that primary-shell solvation is critical, but secondary-shell solvation is unimportant. In the case of LDA/HMPA-mediated reactions, one might presume that THF, normally a strongly coordinating solvent, would be relegated to the role of inert cosolvent. Indeed, LDA/HMPA-mediated dehydrohalogenations discussed as background (vide supra)⁵ and metalations of epoxide 5 and aryl carbamate 7 described herein show no influence by THF whatsoever. In short, the reaction rates in HMPA/THF/hydrocarbon mixtures are independent of the THF concentrations.

On occasion, however, supposedly inert cosolvents can influence organolithium structure and reactivity.^23,24 For example, monomer-based ester enolization (see 12) is not measurably influenced by THF, whereas the putative triple-ion-based pathway (see 13) is inhibited by THF. There is no evidence that primary-shell solvation by THF is the culprit. Similarly, LDA/HMPA-mediated metalation of imine 1 via putative monomer 25 is markedly accelerated by THF whereas lower-solvated analogs 23 and 24 are not influenced by the THF concentration. By using THF/2,5-Me₂THF cosolvent mixtures (rather than THF/hexane) we observe a loss of the THF concentration dependence, suggesting that the influence of the cosolvent is purely through secondary-shell (medium) effects. It is tempting, plausible, and convenient to conclude that the influence of the ethereal cosolvent is due exclusively to secondary-shell solvation. It may not be that simple, however.

The LDA/HMPA-mediated conjugate addition to ester 3 proceeds via a disolvated dimer-based pathway (28 or 29) and a tetrasolvated dimer-based pathway (30). Both have an affiliated second-order THF dependence in THF/hexane mixtures. The role of THF, however, is very strange. By holding the polarity of the medium fixed and varying the THF with 2,5-Me₂THF as the cosolvent, the second-order THF dependence gives way to a first-order dependence. Given the pronounced steric demands of both LDA and HMPA and consequent buttressing, one cannot rule out primary-shell solvation by HMPA and THF. Are we suggesting that there is both primary- and secondary-shell solvation by THF? Possibly, but not with any conviction. We must confess that, despite continued efforts to distinguish primary and secondary-shell solvation effects, the latter remains largely inscrutable.

Is such mechanistic complexity unusual?

Surveying mechanisms underlying LDA/HMPA-mediated reactions one cannot help but notice the large variation of both monomer- and dimer-based pathways. An inventory of LDA/HMPA-mediated reactions reported to date includes reactions based on [(i-Pr₂NLi)(HMPA)_1-3]^‡ and [(i-Pr₂NLi)₂(HMPA)_2-4]^‡--six total (not including putative mixed solvated forms). If each case study is viewed in isolation, the results seem reasonable. We have no reason to doubt the veracity of the data and, consequently, do not doubt the diversity of the behavior. Nevertheless, the case studies taken together afford a complexity that is daunting, vexing, and unique to LDA/HMPA-mediated reactions.

Is HMPA special?

The short answer is yes. HMPA presents a profound conflict between a marked Lewis basicity affiliated with the P=O dipole^2a,44-46 and the exceptional steric demands of the splaying dimethylamino groups.⁷ Semiempirical computational studies of lithium amide solvation that compare HMPA with (H₂N)₃P=O attest to the high affinity of phosphoramides for lithium ion.⁷ However, the exothermicity of serial solvation tapers off gradually for (H₂N)₃P=O drops off markedly for HMPA. Reich and coworkers have observed this experimentally in the serial solvation of Li cations by HMPA in ethereal solvents: Three HMPA substitutions occur quantitatively; a fourth HMPA substitutes reluctantly.² We also cannot ignore the idea expressed previously that solvation of uncoordinated HMPA could contribute significantly to its penchant for binding to lithium cations. Solvent-solvent interactions in solutions of HMPA have been discussed.²³

Conclusion

LDA/HMPA elicits marked increases in mechanistic complexity compared with LDA in the absence of HMPA. This increased complexity may derive from the pronounced steric demands of both the LDA and HMPA. We have documented a plethora of mechanisms for LDA/HMPA-mediated reactions and have asked as many questions as we have answered. It is curious that the influence of HMPA on reaction rates can, as exemplified by the relative rate constants in Table 1, border on insignificant. The often marginal influence on reaction rates seems incongruent with the significant influence HMPA imparts on mechanism and its importance in organic synthesis.

Experimental Section

Reagents and Solvents

Ethereal solvents, hydrocarbons, and HMPA were vacuum transferred from calcium hydride. The hydrocarbon stills contained 1% tetraglyme to dissolve the ketyl. Imines 1 and 1-d₄¹¹ and epoxides 5 and 5-d₂⁹ were recrystallized. Unsaturated ester 3 was prepared as described in Supporting Information. LDA was prepared as a solid from commercial n-BuLi and purified using a standard literature procedure.^6a Air- and moisture-sensitive materials were manipulated under argon or nitrogen following standard glovebox, vacuum line, and syringe techniques.

Kinetics

The rate studies were carried out using methods based on in situ IR spectroscopy^4,28 or gas chromatography⁹ as described in detail previously.

Table 3.

Relative rate constants for reactions of LDA.^a

substrate	Temp (°C)	kHMPA:kTHF
1	-55	3:1
3	-78	4:125
5	0	1:6
7	-40	1.5:1

^aLDA is 0.10 M in either 10.0 M THF/hexane or 0.40 M HMPA and 10.0 M THF/hexane.