Carbon-Nitrogen Bond Formation Using Sodium Hexamethyldisilazide: Solvent-Dependent Reactivities and Mechanisms

Abstract

The solvent-dependent reactivity of sodium hexamethyldisilazide (NaHMDS) toward carbon-centered electrophiles reveals reactions that are poorly represented or unrepresented in the literature including direct aminolysis of aromatic methyl esters to give carboxamides, nitriles, or amidines, depending on the choice of solvent. S_NAr substitutions of aryl halides and opening of terminal epoxides are also examined. A combination of ¹H and ²⁹Si NMR spectroscopic studies using [¹⁵N]NaHMDS, kinetic studies, and computational studies reveal the complex mechanistic basis of the preferences for simple aryl carboxamides in toluene and dimethylethylamine and arylnitriles or amidines in THF. A prevalence of dimer- and mixed dimer-based chemistry even starting from the observable NaHMDS monomer in THF solution is notable.

Graphical Abstract:

Introduction

After decades of communal disinterest in organosodium chemistry, we began trying to shine a light on its potential by expanding the toolbox for generating and using organosodium species.^1–3 This involved some minor adjustments to protocols for generating and handling sodium diisopropylamide (NaDA) ^2,3 as well as developing a new reagent, sodium isopropyl(trimethyl)silyl amide (NaPTA), that manifests desirable solubilities and reactivities as a strong base.⁴ Our primary approach, however, is decidedly structural and mechanistic with the faith that understanding how solvation and aggregation influence reactivity and selectivity will propel applications through a combination of serendipity in our lab and need-driven progress by others. Even the most prominent organosodium reagent, sodium hexamethyldisilazide (NaHMDS),⁵ while garnering the attention of crystallographers,⁶ had evaded spectroscopic, mechanistic, and computational scrutiny until recent studies of its solvent-dependent structure and reactivity toward enolization.⁷ This brings us to the current work. Aminolyses illustrated with generic examples in Scheme 1 are legion. S_NAr substitutions and peptide bond formations are of unquestioned importance in pharmaceutical chemistry.^8,9 With that said, any reaction in which the electrophile is merely heated in ammonia or simple alkylamine will be difficult to improve upon. However, should these simple protocols fail, whether owing to low reactivity or poor selectivity, the experimentalist is left with few options. In these instances, metal amides could offer solutions through control of their coordination spheres.

Scheme 1.

Generic aminolyses

We describe herein a survey of the reactivity of NaHMDS toward carbon-centered electrophiles of potential interest in synthesis accompanied by detailed structural and mechanistic studies. NaHMDS plays the role of a highly reactive analog of ammonia and a preface to expanding investigations of sodium alkylsilazides.⁴ Some of the transformations surveyed as opportunities for mechanistic studies piqued our interest as synthetically promising and have surprisingly little or no presence in the NaHMDS literature. The more mechanistically inclined will find the prevalence of dimer- and mixed aggregate-based reactivity surprising.¹⁰ We also inadvertently stumbled into the complex world of organosilicon chemistry.

Results and Discussion

Aminolyses.

Substitutions of representative organic substrates by NaHMDS are summarized in Table 1. The yields are of isolated, purified products that have been desilylated during workup. ¹H and ²⁹Si NMR spectroscopic monitoring (vide infra) shows the reactions are often pristine and reveals unisolated silylated and sodiated intermediates. The times and temperatures listed in Table 1 are those required to achieve high conversion as confirmed by in situ monitoring using NMR or IR spectroscopies.

Table 1.

Reactions of NaHMDS with electrophiles in various solvents.

entry	substrate	conditions	product	yield
1	1	2.0 equiv NaHMDS 25 °C, 5 h toluene	2	72%
2	3	3.0 equiv NaHMDS 25 °C, 0.3 h DMEA	4	95%
3	5	3.0 equiv NaHMDS 25 °C, 0.3 h toluene	4	90%
4	3	1.0 equiv NaHMDS 50 °C, 1.0 h THF	6	86%
5	3	3.0 equiv NaHMDS 50 °C, 0.3 h THF	7	92%
6	8	3.0 equiv NaHMDS 70 °C, 0.3 h toluenea	9	85%
7	8	3.0 equiv NaHMDS 70 °C, 1.0 h THF	10	96%
8	11	2.0 equiv NaHMDS 70 °C, 1.0 h toluene	12	76%
9	11	3.0 equiv NaHMDS 70 °C, 1.0 h THF	13	95%
10	14	3 equiv NaHMDS 50 °C, 1 h DMEA	15	78%
11	14	3.0 equiv NaHMDS/THF 70 °C, 2.0 h 25 °C, 24 h	16	92%
12	17	2.0 equiv NaHMDS 25 °C, 0.05 h DMEA	16	95%
13	18	2.0 equiv NaHMDS 25 °C, 2 h toluene	19	85%
14	20	2.0 equiv NaHMDS 25 °C, 1.0 h toluene	21	77%
15	22	2.0 equiv NaHMDS 25 °C, 1.0 h toluene	23	83%
16	24	2.0 equiv NaHMDS 110 °C, 2 h toluene	25	55%
17	26	2.0 equiv NaHMDS 110 °C, 3 h toluene	27	76%
18	28	2 equiv NaHMDS 60 °C, 24 h toluene	(>50:1) 29	86%
19	30	2 equiv NaHMDS 25 °C, 24 h THF	31	76%

^aForms a gel-like mixture during the reaction.

For context, the aminolysis of methyl cinnamate to form carboxamide (entry 1) rather than alternatives such as 1,4-addition,^11,12 enolization,⁵ or generalized destruction caught our attention. Ester aminolyses by LiHMDS and NaHMDS may populate pharmaceutical notebooks, but they are poorly represented in the published literature. In 1963, Kruger et al. aminolyzed an aryl ester with NaHMDS to form an O-methyloximino ether, an intermediate en route to carboxamides (as in entry 2). One is detected in mechanistic studies below.¹³ In 1998, Hwu and coworkers reported a NaHMDS-mediated conversion of aryl esters to nitriles with NaHMDS at 110–185 °C (as in entry 4), attributing a central importance to phenolic groups.¹⁴ One-pot conversions of aryl esters to amidines (entries 11 and 12) or carboxamides (such as entry 2) by NaHMDS are unreported. Additions to nitriles to form amidinates (as in entry 12) are well known for LiHMDS.¹⁵ Mechanistic studies below offer some thoughts on why analogous additions of NaHMDS to nitriles are rare.¹⁶

The potentially useful selectivities for the formation of carboxamides, nitriles, and amidines directly from esters have complex mechanistic underpinnings dependent on solvent, temperature, and equivalents of NaHMDS (vide infra). The 3- and 4-picoline methyl esters form gels in weak solvents (entries 9 and 11), presumably owing to head-to-tail oligomeric substrate-NaHMDS dimer complexes but without negative consequences.

S_NAr reactions (entries 15–19) represent a reaction class of unquestioned importance.⁸ We find it somewhat confounding that S_NAr reactions seem to be more effective with soft nucleophiles. Malonates, for example, are far superior nucleophiles¹⁷ than unstabilized enolates.¹⁸ We can find only one report of direct (uncatalyzed) S_NAr substitution by LiHMDS¹⁹ and none for NaHMDS, while RNH₂-based aminolyses facilitated by LiHMDS are legion albeit with an unknown role of the LiHMDS.²⁰ We expended considerable effort trying to glean mechanistic insights (entry 15–17) only to be thwarted by deeply colored debris that appears to be diazo derivatives arising from NaHMDS reacting with the nitro moiety noted years ago (eq 1).²¹ Nitroarenes are spartan in the voluminous literature of lithium amide-mediated ortholithiations,²² possibly for similar reasons.

(1)

A few problematic substrates are illustrated in Chart 1. Arenes 32–34 gave complex products suggestive of orthometalations and possibly pyridyne intermediates. Halopyridines 35 and 36 and chloroisoquinoline 37 aminated but were prone to transfer a silyl group to the 3-position by a Fries-like process.²³ Nitrobenzene 38 underwent a clean halogen dance,²⁴ forming 2-fluoro-3-iodonitrobenzene.²⁵ We replicated the highly regioselective epoxide openings reported by Withnall and coworkers in 2008²⁶ (entries 20 and 21) to evaluate them as candidates for mechanistic work and include these potentially useful results because they have gone largely unnoticed.²⁷

Chart 1.

Failed and troublesome substrates

To assist the reader, Scheme 2 represents a roadmap to what is forthcoming in the mechanistic studies. We explore the reaction coordinates to convert NaHMDS and methyl picolinate 3 to carboxamide 4 in DMEA (Part 1), to nitrile 6 in THF (Part 2), and conversion of nitrile 6 to amidine 7 in DMEA (Part 3) and THF (Part 4). Each part delves into the underlying organosodium and organosilicon chemistry by examining spectroscopically observable structural events, kinetic studies to evaluate solvation and aggregation in the rate-limiting transition structures,²⁸ and computational probes of solvation, aggregation, critical transition structures, and elusive details along the reaction coordinates.^29,30

Scheme 2.

Focus of the four studies examined spectroscopically, kinetically, and computationally.

Monitoring the reactions by ¹H and ²⁹Si NMR spectroscopies showed the reactions to be very clean and provided detailed structural assignments. [¹⁵N]NaHMDS⁷ distinguished O–Si and ¹⁵N–Si species owing to ¹⁵N–²⁹Si coupling whereas combinations of [¹⁵N]NaHMDS and NaHMDS were exploited to examine reversible steps. DFT computations using Me₃N in place of Me₂NEt to reduce unnecessary degrees of freedom revealed an exothermic substitution of DMEA on dimer 39. We take the liberty of using the original numbers assigned to DMEA solvates in the schemes showing computed Me₃N solvates. The shorthand of general form A_mS_n alludes to the hexamethyldisilazide fragment (A) and solvent (S).

Part 1: Mechanism of aminolysis of methyl-2-picolinate (3) with NaHMDS/DMEA.

We focused structure-reactivity studies on additions to methyl-2-picolinate in DMEA and THF. Toluene acts similarly to DMEA but affords broader resonances. The reaction coordinate for the addition of NaHMDS to picoline 3 in DMEA observable by NMR spectroscopy is summarized Scheme 3. Mono- and dichelated dimers 40 and 41 observable at −80 °C reflect the behavior of NaHMDS with other bifunctional ligands studied previously.⁷ Both dimers display coupling constants (¹J_N-Si = 8.5 and 8.7 Hz, respectively) and chemical shifts (−15.8 and −16.4 ppm, respectively) characteristic of [¹⁵N]NaHMDS dimers.⁷ Addition of excess 3 affords computationally viable doubly chelated monomer 42 with an upfield ²⁹Si chemical shift and large coupling constant (−21.41 ppm, ¹J_N-Si = 13.1 Hz) characteristic of a NaHMDS monomer.⁷ Rate studies (below) place dimer 40 on the reaction coordinate. The stability of 43 allows carboxamide 4³¹ to be isolated in high yield (Table 1, entry 5).³²

Scheme 3.

Spectroscopically observed species on the reaction of methyl-2-picolinate with NaHMDS/DMEA.³³

The computed structure of 40 (Figure 1) reveals a geometry that more closely approximates square planar than tetrahedral sodium. Serial substitution from 39 to 41 is computed to be mildly exothermic for both steps (ignoring translational entropy).³⁴ DFT supports monomer stereoisomer 42 with the alternative stereoisomeric monomer (not drawn) to be 10.4 kcal/mol less stable.

Figure 1.

Ball and stick depictions of DFT-computed dimer 40 displaying an approximate 10° rotation of the chelate and Na₂N₂ planes from coplanarity.

The only observable species during the aminolysis in DMEA are mixed aggregate 43 (Figure 2A) and Me₃SiOMe with the latter confirmed by comparison with an authentic sample (²⁹Si, δ 17.49 ppm). Two ²⁹Si resonances of 43 at −16.0 ppm (¹J_N-Si = 9.7 Hz) and −16.6 ppm (¹J_N-Si = 10.3 Hz) are similar to [¹⁵N]NaHMDS dimer 39 (−15.2 ppm, J_N-Si = 8.5 Hz) and coalesce into a single broad resonance at −20 °C. The imino ²⁹Si resonance is markedly downfield (−9.70 ppm, J_N-Si = 11.2 Hz). Addition of excess unlabelled NaHMDS to pre-formed [¹⁵N₂]43 at −80 °C shows immediate incorporation of unlabelled silazide fragments (Figure 2B). DFT computations illustrate the magnetic inequivalence of the two silazide-derived Me₃Si moieties in 43 (Figure 3).

Figure 2.

Spectrum A is ²⁹Si NMR spectrum of 0.05 M ester 3 with 0.10 M added [¹⁵N]NaHMDS in DMEA after reaction containing only mixed aggregate [¹⁵N]43 and residual NaHMDS dimer [¹⁵N]39 (Scheme 3). Spectrum B is the sample from spectrum A with 1.0 equiv (relative to ester 3) of unlabelled NaHMDS.

Figure 3.

Ball and stick depictions of DFT-computed and artist’s rendition of mixed aggregate 43 with Me₃N as a DMEA surrogate.

We probed the mechanism for the addition of NaHMDS to picolinate 3 in DMEA-pentane mixtures using in situ IR spectroscopy³⁵ under pseudo-first-order conditions (excess NaHMDS) following the loss of monochelated dimer 40 (1737 cm⁻¹) to form mixed dimer 43 (1584 cm⁻¹). Clean first-order decays (Figure 4) afford pseudo-first-order rate constants, k_obsd, that are independent of the initial concentration of 40. Plotting k_obsd versus DMEA concentration with pentane cosolvent (Figure 5) and versus NaHMDS dimer 39 concentration (Figure 6) both show zeroth-order dependencies. The resulting rate law (eq 2) is consistent with a rate-limiting addition via a transition structure of stoichiometry A₂S(substrate). (Recall that A an amide subunit and S is solvent.) The dimer-based reaction coordinate is a hallmark of alkali metal amides in poorly coordinating solvents.^2,7c,36 Figure 7 depicts the computed transition structure TS-1 displaying a developing Na₂(O)(N) mixed dimer core.

Figure 4.

Addition of 0.10 M NaHMDS to methyl-2-picolinate (3, 0.005 M) at 25 °C measured with IR spectroscopy (1737 cm⁻¹). The curve depicts an unweighted least-squares fit to y = ae^−bx (a = 8.1 × 10⁻³ ± 0.1 × 10⁻³; b = 1.5 × 10⁻³ ± 0.1 × 10⁻³).

Figure 5.

A plot of k_obsd vs [DMEA] (M) in pentane for the addition of NaHMDS (0.10 M) to methyl-2-picolinate (3, 0.005 M) at 25 °C measured with IR spectroscopy (1737 cm⁻¹). The curve depicts an unweighted least-squares fit to y = ax + b (a = 2.4 × 10⁻³ ± 0.2 × 10⁻³; b = 8.3 × 10⁻⁶ ± 0.3 × 10⁻⁶).

Figure 6.

A plot of k_obsd vs [NaHMDS] (M) for the addition of NaHMDS to methyl-2-picolinate (3, 0.050 M) in 3.8 M (neat) DMEA at 25 °C measured with IR spectroscopy (1737 cm⁻¹). The curve depicts an unweighted least-squares fit to y = ax + b (a = 2.7 × 10⁻³ ± 0.2 × 10⁻³; b = −1.3 × 10⁻³ ± 0.1 × 10⁻³).

Figure 7.

DFT-computed transition structure TS-1 for the NaHMDS-based addition to methyl-2-picolinate (3) with Me₃N as a DMEA surrogate.

$$-d[\mathbf{40}]/d\mathrm{t}=k^{\prime }{\left[\mathbf{40}\right]}^1{\left[{\mathrm{A}}_2{\mathrm{S}}_2\right]}^0{\left[\mathrm{S}\right]}^0$$ (2)

Probing the reaction coordinate en route to observable product 43 (carboxamide 13 after workup) and Me₃SiOMe using DFT computations served up a nuanced and quite complex story (Scheme 4). To reiterate, Me₃N is used as a surrogate of DMEA to reduce unnecessary degrees of freedom while taking the liberty of reusing the original numbers for 40, 43, and 46. Transition structures TS-1, TS-2, TS-3, and TS-4 all have single negative frequencies. Intrinsic reaction coordinate (IRC) calculations³⁷ were used to determine the minima flanking each transition structure. Minima connected without connecting transition structures represent structural minima corresponding to reorganizations of insufficient interest to probe further. There are likely more minima and barriers resulting from other minor adjustments. These same caveats also apply to the reaction coordinate depicted in Scheme 6 (below).

Scheme 4.

DFT-computed reaction coordinate for the aminolysis of methyl-2-picolinate (3) by NaHMDS using Me₃N as a DMEA surrogate.

Scheme 6.

DFT-computed reaction coordinate for the aminolysis of methyl-2-picolinate (3) with NaHMDS in THF with the ⁺Na(THF)₆ omitted for all structures.

Following the formation of open dimer 45,³⁶ the 1,2-addition proceeds via rate-limiting transition structure TS-1 (depicted in Figure 7) to form the tetrahedral adduct as mixed aggregate 46. Structures beyond TS-1 represent kinetically and spectroscopically invisible post-rate-limiting events. The collapse of the adduct 46 via transition structure TS-2 affords NaOMe–NaHMDS mixed dimer 47. Following a readjustment via TS-3 to position the methoxy as a MeO–Si Lewis acid-base complex 48, silyl transfer via transition structure TS-4 extrudes Me₃SiOMe via complex 49 to form mixed dimer 50 (same structure compared to mixed aggregate 43 instead of utilizing Me₃N as a solvent surrogate in DFT-computation) as the spectroscopically observable product (albeit with an exothermic addition of a second solvent.)

As drawn, the NaOMe silylation is an intra-aggegate reaction. We cannot exclude the possibility that NaOMe is released into solution before silyl transfer but have no reason to invoke it either. We also probed for a methoxy-silyl coupling directly from the tetrahedral adduct 46 without first expelling NaOMe (see 51, path a) but were unable to locate such a transfer. Similarly, a silyl transfer to extrude NaOSiMe₃ from 46 (see 51, path b) was located but had an unreasonably high (23 kcal/mol) barrier consistent with our failure to observe Me₃SiONa spectroscopically.

Part 2: Mechanism of aminolysis of methyl-2-picolinate (3) with NaHMDS/THF.

The aminolysis in THF follows a distinctly different pathway. The spectroscopically observable forms summarized in Scheme 5 bely mechanistic complexity lurking beneath the surface. At the outset, we note that the excess NaHMDS undergoes a THF-dependent deaggregation of disolvated dimer 52 to provide tetrasolvated monomer 53 described previously,^7a,b which proves to be mechanistically important. Picolinate-complexed dimer 54 (analogous to 9) and THF-complexed monomer 55 can be observed at −110 °C at both low and high THF concentrations, respectively, using pentane cosolvent. Evidence of complexation is gleaned from distinct chemical shift differences in the ¹H and ²⁹Si NMR when compared with the uncomplexed substrate 3 and free NaHMDS (Supporting Information). The assigned THF solvation numbers derive from DFT computations.

Scheme 5.

Spectroscopically observed species for the reaction of NaHMDS to methyl-2-picolinate 3 in THF.

Warming mixtures of 3 and NaHMDS to room temperature affords imino ether 56, analogous to that isolated previously,¹³ within 1.0 min. The concomittant formation of Me₃SiONa was confirmed by comparison with an authentic sample (17.70 ppm). In the absence of excess NaHMDS, 56 gives nitrile 6 over 24 h at RT or in 30 minutes at 50 °C in good yield (Table 1, entry 7) with concomitant formation of Me₃SiOMe, confirmed with an authentic sample. The conversion of 56 to 6 is somewhat mysterious in that it does not appear to require NaHMDS and is not accelerated by excess NaHMDS, yet it does not occur by heating a crude isolated sample of 56.

Reaction of 3 with excess NaHMDS renders nitrile 6 unobservable because it is quickly scavenged to give amidinate 57³⁸ with noteworthy spectroscopic properties. The ²⁹Si resonance of [¹⁵N₂]57 appears as a clean triplet owing to second-order effects referred to as “virtual coupling.” The two magnetically equivalent ²⁹Si nuclei experience equivalent coupling by the ¹⁵N nuclei despite one being proximal and one being distal to each silicon nucleus (Figure 8). Similar virtual coupling is observed in bridging transition metal phosphides (M₂P₂ core)³⁹ and has been observed in a Li₂P₂ lithium phosphide dimer.⁴⁰ In theory, we could have observed this in NaHMDS (Na₂N₂) dimers as well but have not.⁷ If the symmetry is broken by generating [¹⁵N₁]57 from unlabelled nitrile 6 and [¹⁵N]NaHMDS, the ²⁹Si resonance appears as a doublet of doublets owing to a large ¹J_N-Si coupling and small ³J_N-Si coupling with a slight isotopic perturbation visible.⁴¹ The requisite silyl transfer from the initially formed unsymmetric N,N-disilylamidine to N,N’-disilylamidine 56 is discussed in the context of DFT computations below. The depicted 6–57 nitrile-amidinate equilibrium cannot be observed directly, but adding [¹⁴N]NaHMDS to labelled amidinate 57 results in label exchange (eq 3).

Figure 8.

²⁹Si NMR spectra of [¹⁵N₂]57 showing virtual coupling (¹J_N–Si = 8.8 Hz) and standard coupling in [¹⁵N₁]57 (¹J_N–Si = 8.9 Hz and ³J_N–Si = 2.8 Hz).

(3)

An analogous addition of NaHMDS to PhCO₂Me (Table 1, entry 13) to generate phenyl-substituted amidinate 58 was approximately 60-fold (see in supporting information) slower than for 3. It also revealed the elusive nitrile-amidinate exchange as a temperature-dependent equilibrium (eq 4). Heating excess NaHMDS (3.0 equiv) and PhCO₂Me to 80 °C followed by rapid thermal quenching to −80 °C affords predominantly benzonitrile (17). Alternatively, cooling the reaction to 25 °C and letting it stand for 12 h generates amidinate 58. Heating-cooling cycles show the changes are temperature dependent and reversible. Adding [¹⁵N]NaHMDS to unlabelled benzonitrile, 17, shows label incorporation in both 17 and 58. Thus, benzonitrile can be isolated in 86% yield using 1.1 equiv of NaHMDS and elevated temperatures (Table 1, entry 13), while amidine 58 can be isolated in 98% yield using 3.0 equiv NaHMDS at 80 °C for 3 h, then 25 °C for 12 h. This odd temperature dependence in which heating accelerates the aminolysis but retards addition to the intermediate nitrile might be why nitrile-to-amidine conversions largely exploit LiHMDS¹⁵ rather than NaHMDS.¹⁶ Alternatively, it could just be a cultural preference for LiHMDS. There are no extraneous silazide signals in the ²⁹Si spectrum that would implicate a mixed aggregate.

(4)

Rate studies for the addition of NaHMDS to picolinate 3 to afford imino ether 56 provided yet another nuanced story. We anticipated that the first-order dependence—the first-order decay—on complexed monomer 55 measured in neat THF would be accompanied by a zeroth-order NaHMDS dependence. To the contrary, a plot of k_obsd vs NaHMDS shows a half-order dependence on the free NaHMDS monomer (Figure 9). The partial rate law (deferring discussion of solvent-concentration dependencies momentarily) is described by eq 5. To be clear on a critical point: A dimer-based addition involving contributions of a standard NaHMDS monomer would manifest a first-order NaHMDS dependence.⁴² The half-order on a monomer demands a full ionization of the sodium cation.⁴³

$$–d[\mathbf{55}]/dt=k^{\prime }{[\mathbf{55}]}^1{\left[{\mathrm{A}\mathrm{S}}_4\right]}^{1/2}$$ (5)

such that $\mathbf{55}={\mathrm{A}\mathrm{S}}_2(\mathbf{3})$

Figure 9.

The plot of k_obsd vs [NaHMDS] (M) for the addition of NaHMDS to methyl-2-picolinate (3, 0.005 M) in 12.8 M (neat) THF at 0 °C measured with IR spectroscopy (1731 cm⁻¹). The curve depicts an unweighted least-squares fit to y = axⁿ (a = 2.4 × 10⁻³ ± 0.1 × 10⁻³; n = 0.44 ± 0.01).

We may not yet fully understand the role of solvation. The forthcoming discussion may even seem excessive, but we are still groping to understand how organosodium solvation and aggregation differ from those of organolithiums. A plot of k_obsd vs THF in pentane (Figure 10, curve A) manifests a maximum in the rates at intermediate THF concentrations. By contrast, analogous data using 2,5-dimethyltetrahydrofuran (2,5-Me₂THF) or the decidedly more expensive 2,2,5,5-tetramethyltetrahydrofuran (2,2,5,5-Me₄THF) as poorly coordinating polar cosolvents^44–47 eliminate the maximum. The solvent dependencies suggest that medium effects are in play. We remind the reader that two deaggregations are occurring concurrently. At low THF concentrations, the dimeric reactant A₂S(substrate) (54) exists concurrently with A₂S₂ (NaHMDS dimer 52), and the rise in rates indicates they are collectively under-solvated relative to the rate-limiting transition structure. At high THF concentrations in polar cosolvents, the observable reactants AS₂(substrate) (54) and AS₄^7c (53) appear optimally solvated—six THF ligands total—but under-aggregated relative to the dimeric rate-limiting transition structure.

Figure 10.

A plot of k_obsd vs [THF] (M) for the addition of NaHMDS to methyl-2-picolinate (3, 0.005 M) in 0.10 M NaHMDS at 0 °C measured with IR spectroscopy (1732 cm⁻¹). Curve A derives from pentane cosolvent and depicts an unweighted least-squares fit to y = [axⁿ/(1+bxⁿ)](1/1+cx²) (a = 2.4 × 10⁻³ ± 0.1 × 10⁻³; b = 1.3 × 10⁻¹ ± 0.1 × 10⁻¹; c = 0.6 × 10⁻³ ± 0.1 × 10⁻⁴; n = 5). Curve B derives from 2,5-Me₂THF cosolvent and depicts an unweighted least-squares fit to y = axⁿ/(1+bxⁿ) (a = 1.6 × 10⁻¹ ± 0.1 × 10⁻¹; b = 1.6 × 10⁻¹ ± 0.1 × 10⁻¹; n = 5). Curve C derives from 2,2,5,5-Me₄THF cosolvent and depicts an unweighted least-squares fit to y = axⁿ/(1+bxⁿ) (a = 8.2 × 10⁻⁴ ± 0.1 × 10⁻¹; b = 7.9 × 10⁻² ± 0.1 × 10⁻²; n = 2.2 ± 0.2).

The problem we confronted in this study was that 2,5-Me₂THF was binding cooperatively with THF to promote monomer formation as illustrated in eq 6. At low THF concentration, an unusually elevated concentration of NaHMDS monomer in 2,5-Me₂THF compared with pentane implicates a

(Me₃Si)₂N(THF)_x(Me₂THF)_y mixed solvate. However, there is almost no chance that 2,2,5,5-Me₄THF cooperatively solvates monomers, and DFT computations support this assertion. We still, however, observe the promotion of the monomer. Thus, the flattening of the curve is at least partially attributable to such an unproductive side equilibrium to form a mixed solvate according to the Principle of Detailed Balance.⁴⁸ We must confess that the function used to fit the THF/pentane data (caption in Figure 10) contains a correction for inhibitory medium effects, 1/(1+cx²), that was derived empirically with no formal molecular basis. The challenge posed by these medium effects rears its ugly head below.

(6)

With all that said, the rate law in eq 5 provides an adequate picture of a mechanism requiring 6 solvents and two silazide subunits—a hexasolvated dimer-based transition structure TS-5 (Figure 11) with a ⁺Na(THF)₆ gegenion (Figure 12).³⁶

Figure 11.

DFT-computed transition structure TS-5 for the NaHMDS-based addition to methyl-2-picolinate (3) solvated by THF. The implicit ⁺Na(THF)₆ cation^7b (Figure 12) is not included.

Figure 12.

DFT computed ⁺Na(THF)₆.

We computationally examined the overall reaction coordinate using a triple-ion-based framework (Scheme 6). The cation (Figure 12) is omitted for all structures in Scheme 6. Triple ions are well precedented,³⁶ including for LiHMDS. Inspection of Figure 11, however, reveals that it is a triple-ion-based reference state (59) because the lowest-energy rate-limiting transition structure TS-5 has lost all semblance of the N–Na–N triple ion connectivity as has the IRC-derived minimum 61 preceding it.

There is plenty of room for alternative interpretation. One could, for example, imagine 61 stemming directly from a chelated monomer and ionized NaHMDS fragment. It is also possible, however, that the conversion of 60 to transition structure TS-5 proceeds via 61 in which the (Me₃Si)₂N⁻ in 61 remains “bound.” (The dissociation of 61 is calculated to be +5.5 kcal/mol.) Transition structure TS-5 gives way to tetrahedral adduct 62. The sequence follows an aza-Brook-like silicon transfer akin to that presumed to occur in an aza-Peterson-like imine formation⁴⁹ via transition structure TS-6, elimination of 63 via transition structure TS-7 to form arene π complex 64, and arene dissociation to afford the observed imino ether 56 and unobserved (fleeting) mixed triple ion 65.^36,50

Part 3: Mechanism of aminolysis of 2-pyridinecarbonitrile (6) with NaHMDS/DMEA.

Motivated as much by an obsessive need to complete the story as by curiosity, we examined the addition of NaHMDS to nitrile 6 to form amidinate 57 (Scheme 5) and were rewarded for our persistence. IR and NMR spectroscopy showed no evidence that nitrile 6 binds to NaHMDS at low or high DMEA concentrations. The addition occurs to give mixed amidinate 66 within seconds at 25 °C (eq 7) manifesting a silazide fragment that is well resolved from NaHMDS homodimer 39 displaying dimer-like coupling (¹J_N–Si = 8.5 Hz; Figure 13). As described above, further heating at 50 °C causes scrambling of the ¹⁴N–¹⁵N isotopes.

Figure 13.

²⁹Si NMR spectra of [¹⁵N₂]NaHMDS dimer (39, ¹J_N–Si = 8.5 Hz) and mixed aggregate [¹⁵N₁]66 (¹J_N–Si = 7.3 Hz, ³J_N–Si = 2.0 Hz).

(7)

Rate studies show an inverse-second-order dependence on DMEA (Figure 14) and first-order dependence on NaHMDS dimer 39 (Figure 15) consistent with a disolvated-dimer-based 1,2-addition. A full computational workup is found in the Supporting Information. An abbreviated version is illustrated in Scheme 7. For example, the conversion of 70 to 66 requires several additional mundane readjustments. Also, we examined the role of solvation of the intermediates and found minima that were higher energy than the unsolvated forms except for 70, whose Me₃N-solvated form is −2.5 kcal/mol more stable.

Figure 14.

A plot of k_obsd vs [DMEA] (M) in pentane for the addition of NaHMDS to pyridinecarbonitrile (6, 0.005 M) at −40 °C measured with IR spectroscopy (1584 cm⁻¹). The curve depicts an unweighted least-squares fit to y = axⁿ/(1+bxⁿ) (n = −1.9 ± 0.1; a = 2.7 × 10⁻² ± 0.1 × 10⁻²; b = 7.1 × 10⁻³ ± 0.1 × 10⁻²).

Figure 15.

Plot of k_obsd vs [NaHMDS] (M) for the addition of NaHMDS to 2-pyridinecarbonitrile (6, 0.005 M) in 3.85 M (neat) DMEA at −40 °C measured with IR spectroscopy (1584 cm⁻¹). The curve depicts an unweighted least-squares fit to y = axⁿ (a = 2.1 × 10⁻² ± 0.1 × 10⁻²; n = 1.1 ± 0.1).

Scheme 7.

DFT-computed reaction coordinate for the aminolysis of 2-pyridinecarbonitrile (6) by NaHMDS using Me₃N as a DMEA surrogate.

Part 4. Mechanism of aminolysis of 2-pyridinecarbonitrile (6) with NaHMDS/THF.

IR and NMR spectroscopy showed no evidence that nitrile 6 binds to NaHMDS (eq 8). The only species observable by ¹H and ²⁹Si NMR spectroscopy was free NaHMDS (52 and 53, Scheme 5) and sodium amidinate 57 to the exclusion of any mixed aggregate. DFT computations suggest 57 is a disolvate.

(8)

Rate studies showed first-order decays and concentration-independent values of k_obsd, consistent with a first-order dependence on nitrile 6. A plot of k_obsd vs THF concentration shows a distinct sigmoidal dependence affording a pronounced inhibition at high THF (Figure 16, curve A). Any doubt that this correlates with the NaHMDS dimer-monomer deaggregation (52 and 53, Scheme 5) was put to rest by superimposing the rate data on a plot of the equilibrium population of dimer (curve B) measured in 2020 by a different experimentalist.^7b Clearly, deaggregation inhibits the reaction. A plot of k_obsd vs NaHMDS in neat THF where NaHMDS is >98% monomer shows a clean second-order dependence (Figure 17), consistent with a requisite monomer-to-dimer aggregation before rate-limiting addition. The crudely defined zeroth-order THF dependence implicates the partial rate law in eq 9, pointing to the overall dimer-based transition structure, [A₂S₂(ArCN)] ^‡.^51,52 The DFT computed transition structure TS-10 is illustrated in Figure 18.

Figure 16.

A plot of k_obsd vs [THF] (M) for the addition of NaHMDS to 2-pyridinecarbonitrile in pentane cosolvent (curve A, red) and ²⁹Si chemical shift (curve B, green) plotted versus [THF] in 2:1 pentane/toluene as cosolvent measured at −20 °C. The latter function fits to a model based on an A₂S₂−AS₄ equilibrium (Supporting Information).^7b,52 The blue data (Curve C) derives from 2,5-Me₂THF rather than pentane as cosolvent (vide infra).

Figure 17.

Plot of k_obsd vs [NaHMDS] (M) for the addition of NaHMDS to picolinonitrile (6, 0.005 M) in 12.8 M (neat) THF at −20 °C measured with IR spectroscopy (1584 cm⁻¹). The curve depicts an unweighted least-squares fit to y = axⁿ (a = 4.3 × 10⁻² ± 0.1 × 10⁻²; n = 2.0 ± 0.1).

Figure 18.

DFT computed transition structure TS-10 for the NaHMDS-dimer-based addition to nitrile 6.

$$–d[\mathbf{6}]/d\mathrm{t}=k^{\prime }{[\mathbf{6}]}^1{\left[{\mathrm{A}\mathrm{S}}_4\right]}^2$$ (9)

The DFT-computed reaction coordinate for addition to nitrile 6 in THF shows many parallels with that in DMEA that are relegated to the supporting information. The notable differences are the exothermic solvation of all key species including transition structure TS-10 in Figure 18, mixed dimer-based transition structure TS-9 for silicon transfer, and unobserved mixed aggregate 71 whose analog (66, eq 7) was fully characterized in DMEA.⁵³

We have accumulated evidence of pre-aggregation-based reactions originating from monomers, but none have ever been so poignant.³⁶ Why is the dimer-based addition to nitriles dominant? A partial answer is that the trajectory of the silazide attack on the nitrile π system in TS-10 (Figure 18) appears optimal. The most stable transition structure for monomer-based addition (TS-11, Figure 19) is less favorable by 4.4 kcal/mol using a monomer reference state, which appears to be a consequence of an inferior trajectory.

Figure 19.

DFT computed transition structure TS-11 for the unfavorable NaHMDS-monomer-based addition to nitrile 6.

We would be remiss to not comment on the influence of 2,5-Me₂THF rather than pentane as the cosolvent for nitrile addition (curve C in Figure 16). It provided further evidence of cooperative solvation. While checking for polarity effects we detected and subsequently documented cooperative monomer solvation, (Me₃Si)₂NNa(THF)_x(Me₂THF)_y, noted in eq 6. The details are obscured in Figure 16 by an overlaying THF concentration dependence. The complete suppression of the rate suggested that the mixed solvated monomers persisted as an unproductive side equilibrium at stunningly low THF concentrations. Indeed, this was caused by a steep temperature dependence on the deaggregation. The message for us is that 2,5-Me₂THF is a far better non-coordinating THF surrogate for lithium than for sodium.

Summary.

The studies described above proved to be four discrete mechanistic studies in which the choice of solvent and conditions markedly influence the outcome of the reactions. We offer an overview in which the results are grouped according to choice of solvent—THF or DMEA. This summary is organosodium centric; there is a considerable body organolsilicon chemistry (largely addressed computationally) that is left to the four parts described above. The arrows in Schemes 8 and 9 encompass many discrete steps including the organosilicon chemistry.

Scheme 8.

Summary of NaHMDS mechanistic studies in THF solution.

Scheme 9.

Summary of NaHMDS mechanistic studies in DMEA solution.

The THF results are summarized in Scheme 8. While NaHMDS in neat THF solution exists exclusively as tetrasolvated monomer 53, it reacts with methyl-2-picolinate, 3, via a dimer-based pathway in which a triple ion motif TS-5 is invoked at the rate-limiting transition structure. The critical chemoselectivity stems from a post-rate-limiting extrusion of NaOSiMe₃ to form spectroscopically observable imino ether 56, which reacts further to give nitrile 6 and sodium amidinate 57 en route to a high-yielding one-pot synthesis of amidine 7. This ester-to-amidine conversion seems potentially important synthetically. Addition of NaHMDS to nitrile 6 is also dimer-based as depicted in computationally viable transition structure TS-10. The dominance of dimer-based reactivity under conditions affording monomeric NaHMDS should pique the interest of those interested in mechanistic alkali metal chemistry.

The chemistry of NaHMDS in poorly coordinating DMEA^7b (Scheme 9) shows some parallels with the results in THF, but the differences are consequential. Although we do not consider a dimeric NaHMDS to be the proximate cause of dimer-based reactivity—many if not most monomer-based organoalkali metal reactions emanate from aggregates—the dominance of a dimer-based transition structure, TS-1, leading to observable mixed aggregate 43 aligns with previous studies of both LiHMDS and NaHMDS in DMEA.^7c,54 In contrast to THF, however, a critical post-rate-limiting extrusion of MeOSiMe₃ rather than NaOSiMe₃ dictates the chemoselective formation of carboxamide 4. Observable mixed aggregate 50 is robust but finds a path to 66 over a month at 25 °C. Although this is clearly an inferior route to amidine 7 we examined the addition to nitrile 6 and, once again, uncovered dimer-based chemistry via an unsolvated transition structure, TS-8.

Conclusion

While casting about for case studies to investigate solvent-structure-reactivity relationships in NaHMDS we happened across several potentially useful reactions that are poorly represented in the literature and provided mechanistic insights that were wholly unexpected. The potential utility stems from C-N bond formations directly from aryl methyl esters, bypassing more activated forms. For us, however, they provided clean examples illustrating the role of solvation and aggregation in the chemistry of sodium amides in general and NaHMDS in particular. The NaHMDS itself serves as a preface to ongoing studies of sodium alkyl(trimethylsilyl) amides with potentially greater utility as strong bases and for C-N bond-forming reactions.

The dimer- and mixed dimer-based reactivity has precedent,^7c,36 but the total dominance of aggregate-based reactivity still came as a surprise. From a mechanistically tactical perspective, the work underscores both the power of ²⁹Si–¹⁵N coupling to determine solution structures and the general merits of ²⁹Si NMR spectroscopy. (We asserted previously that ²⁹Si NMR spectroscopy is underutilized by those outside the organosilicon community.) We offer a final caveat to those who might be tempted to probe primary- versus secondary-shell solvation by using 2,5-Me₂THF as a polar, non-coordinating cosolvent. Despite its success in organolithium chemistry, the larger sodium ion may call for the more expensive 2,2,5,5-Me₄THF. On the truism that an effect is either sterics or electronics, ours’ and others’ experience with metal ion solvation is that sterics dominates.⁵⁵

Experimental

Reagents and solvents.

NaHMDS and [¹⁵N]NaHMDS were prepared as white crystalline solids.^7b Toluene, hexanes, THF, MTBE, cyclopentane, 2,5-Me₂THF, and HMPA were distilled from blue or purple solutions containing sodium benzophenone ketyl. All substrates and products in Table 1 are commercially available.^31,32

General Procedure A: picolinamide 4.

Solid sodium hexamethyldisilazide (NaHMDS, 165 mg, 0.9 mmol) was dissolved in 2 mL of DMEA at 25 °C. 2 mL of the NaHMDS solution was added to a dry 5 mL Kimble vial equipped with a magnetic stir bar. Methyl picolinate (3, 36 μl, 0.30 mmol) was then added to the reaction solution. The reaction was stirred at 25 °C for 0.3 h. DI water (1 mL) was added, and the resulting biphasic mixture was partitioned between water (1 mL) and ethyl ether (2 mL). The aqueous layer was separated and extracted further with three 2 mL portions of ethyl ether. The combined organic layers were dried over anhydrous magnesium sulfate and then concentrated. Purification of the residue by flash column chromatography (80 % ethyl acetate in hexanes) afforded picolinamide 4 as a white solid. ¹H NMR (500 MHz, CDCl3) δ 8.56 (d, J = 4.6 Hz, 1H), 8.20 (d, J = 7.8 Hz, 1H), 7.89 (s, 1H), 7.84 (td, J = 7.7, 1.8 Hz, 1H), 7.43 (dd, J = 7.6, 4.8 Hz, 1H), 6.23 (s, 1H). ¹³C{¹H} NMR (126 MHz, CDCl₃) δ 166.99, 145.59, 148.32, 137.31, 126.47, 122.44. HRMS (DART) Calc. for C₆H₆N₂O (M+H⁺): 123.04746; found: 123.05493.

General Procedure B: amidine 7.

Solid sodium hexamethyldisilazide (NaHMDS, 165 mg, 0.9 mmol) was dissolved in 1.5 mL of THF at 25 °C. 1.5 mL of the NaHMDS solution was added to a dry 5 mL Kimble vial equipped with a magnetic stir bar. Methyl picolinate (3, 36 μl, 0.30 mmol) was then added to the reaction solution. The reaction was stirred at 50 °C for 0.3 h. DI water (1 mL) was then added. 2M HCl aqueous solution was added to adjust pH = 1, which then was stirred at 25 °C for 1 h. Saturated NaOH solution was added until pH = 12, and the mixture was partitioned between water (1 mL) and chloroform (4 mL). The aqueous layer was separated and extracted further with three 4 mL portions of chloroform. The combined organic layers were dried over anhydrous magnesium sulfate and then concentrated to afford amidine 7 as a yellow oil. ¹H NMR (500 MHz, CDCl3) δ 8.56 (m, 1H), 8.11 (d, J = 7.9 Hz, 1H), 7.78 (td, J = 7.7, 1.8 Hz, 1H), 7.35 (dd, J = 7.5, 4.9 Hz, 1H), 5.94 (s, 3H). ¹³C{¹H} NMR (126 MHz, CDCl₃) δ 161.06, 150.79, 148.35, 137.03, 125.21, 120.81. HRMS (DART) Calc. for C₆H₇N₃ (M+H⁺): 122.06345; found: 122.07097.

NMR spectroscopic analyses.

An NMR tube fitted with a double-septum under vacuum was flame-dried on a Schlenk line and allowed to passively cool to room temperature, backfilled with argon, and placed in a dry ice/acetone cooling bath. Individual stock solutions of the substrate and NaHMDS or [¹⁵N]NaHMDS were prepared at room temperature. The appropriate amounts of the substrate, NaHMDS, solvent, and co-solvent were added sequentially to the tube cooled to −78 °C via a gastight syringe. The tube was flame-sealed under a partial vacuum while cold to minimize evaporation in some cases and left unsealed for incremental additions. The tubes were mixed with a vortex mixer for approximately 10 secs to minimize warming. Standard ¹H, ¹³C, and ²⁹Si direct detection spectra were recorded at 500, 125.79, and 99.36 MHz, respectively, and referenced to Me₄Si (0.0 ppm). Integration of the NMR signals was determined using the line-fitting method included in MNova (Mestrelab research S.L.).

Rate Studies.

IR spectra were recorded with an in situ IR spectrometer fitted with a 30-bounce, silicon-tipped probe. The spectra were acquired at a gain of 1 and a resolution of 4 cm⁻¹. All tracked reactions were conducted under a positive flow of argon from a Schlenk line. A representative reaction was carried out as follows: The IR probe was inserted through a Teflon adapter and O-ring seal into an oven-dried, cylindrical flask fitted with a magnetic stir bar and a T-joint. The T-joint was capped with a septum for injections and an argon line. After evacuation under full vacuum, heating, and flushing with argon, the flask was charged with the solvent mixtures of our choice (toluene, DMEA, THF, 2,5-Me₂THF, and 2,2,5,5-Me₄THF) and cooled to −78 °C in a dry ice−acetone bath. A set of 256 baseline scans were collected and IR spectra were recorded every 15 seconds from 30 scans. The reaction vessel was charged stock solutions of NaHMDS and additional cosolvent through the septum. One set of scans was collected before the addition of substrate through the septum, and the baseline is zeroed. The substrate was added neat or in highly concentrated solutions and was tracked to completion (1731 cm⁻¹ for 7, for example). The spectrometer was configured to collect spectra every 5 seconds from 16 scans. The reaction was tracked over 3–5 half-lives monitoring the disappearance of the starting material and the appearance of the product. The former was used for the rate studies.

Density functional theory (DFT) computations.

All DFT calculations were carried out using Gaussian 16.²⁹ Prompted by a recent benchmarking of modern density functionals, all calculations were conducted at the M06–2X level of theory.^30a–c A pruned (99, 590) integration grid (equivalent to Gaussian’s “UltraFine” option) was used for all calculations. The Ahlrichs basis set def2-svp was used for geometry optimizations and the expanded def2-tsvp basis set for single-point energy calculations.^30d Ball-and-stick models were rendered using CYLview 1.0b.^30e A large number of DFT-computed energies are archived in the Supporting Information.