|Home | About | Journals | Submit | Contact Us | Français|
The method of continuous variation in conjunction with 6Li NMR spectroscopy was used to characterize lithium phenolates solvated by tetrahydrofuran and 1,2-dimethoxyethane. The strategy relies on the formation of ensembles of homo- and heteroaggregated phenolates. The symmetries and concentration dependencies of the heteroaggregates attest to the aggregation numbers of the homoaggregates. The structurally diverse phenols afford substrate- and solvent-dependent combinations of lithium phenolate monomers, dimers, trimers, tetramers, and pentamers. We discuss the refinement of protocols for characterizing O-lithiated species. Computational studies examine further the substituent and solvent dependencies of aggregation.
Characterizing O-lithiated species in solution is particularly important because of their role in organic chemistry,1 but it is a daunting task.2,3,4,5,6,7 Problems stem from a combination of both the high inherent symmetry of the possible aggregates and the absence of measurable Li-X scalar couplings that have in other studies proven indispensable for characterizing alkyllithiums8 and lithium amides.9 As a result, methods of characterization are indirect and often specific to the salt/solvent combination.
We have recently found the method of continuous variation10,11 to be a powerful and widely applicable method for characterizing lithium enolates.6,7 The strategy finds its origins in studies by Chabanel,12 Gagne,13 Günther,14 and others.15 We have now extended these studies to include a broad range of lithium phenolates (ArOLi) solvated by two commonly used ethereal solvents, tetrahydrofuran (THF) and 1,2-dimethoxyethane (DME). It is tempting to underscore the importance of alkali metal phenolates in the synthesis of pharmaceutical agents,16–18 but that would be misleading because only the more reactive sodium and potassium phenolates are prominent.19 In reality, we followed the lead of Jackman and coworkers5 and turned to lithium phenolates as proxies for lithium enolates.20 Lithium phenolates offer desirable stability as well as considerable substrate diversity that allow for a systematic study of substrate-dependent structures. An unanticipated benefit of using lithium phenolates is the remarkable substrate-dependent array of monomers, dimers, trimers, tetramers, and pentamers (Chart 1, 1–5). (Hexamers (6) are noticeably absent.)21,22,23 Such structural diversity and complexity offer opportunities for refining and stress-testing experimental protocols.
This manuscript follows two distinct themes. The results section describes strategies and tactics—a road map of sorts—for readers who wish to understand how the structures are assigned. The discussion section focuses primarily on the principles of solvation and aggregation.
Table 1 summarizes the lithium phenolates studied and their resulting aggregation states and affiliated 6Li chemical shifts. In lieu of a detailed description of each substrate-solvent combination, we summarize the types of experiments used to ascertain structure. Details are archived in supporting information. A later section illustrates how these methods can be amalgamated to assign structures to components of complex mixtures of homoaggregated phenolates.
An ensemble of homo- and heteroaggregated phenolates (eq 1) was monitored as a function of the mole fractions of phenolate subunits (denoted generically as A and B) using 6Li NMR spectroscopy.6,7 The number of aggregates and their symmetries reflect the aggregation state, n (Table 2; Chart 2).13 Resolution of the often large number of 6Li resonances is maximized by pairing An and Bn homoaggregates with widely separated chemical shifts. (The resonance counts for pentamers and hexamers assume slow Li-Li site exchange; vide infra.) A second parameter that influences the choice of the An-Bn pair stems from the tendencies (or lack thereof) of the phenolates to form well-behaved ensembles. In short, pairs of structurally related forms (such as similarly hindered phenolates) typically provide statistical distributions7 that offer compelling structural assignments. To this end, pairs of phenolates differing only in the presence or absence of a para halogen proved especially useful at times.
6Li NMR spectra recorded on An-Bn mixtures offer structural insights, as illustrated in Table 2 and Chart 2. When An and Bn are monomers (n = 1), A-B mixtures show resonances corresponding to A and B to the exclusion of additional resonances. By contrast, 6Li spectra recorded on mixtures of A2 and B2 display resonances of the homodimers along with a single resonance corresponding to the AB heterodimer (Figure 1a). The AB resonance is typically of intermediate chemical shift consistent with Brown’s “local environment” model proffered decades ago.15c,25 Spectra derived from A4-B4 mixtures show single resonances for A4 and B4 along with 3:1, 2:2, and 1:3 resonance pairs characteristic of the A3B1, A2B2, and A1B3 heteroaggregates (Figure 1b). Once again, the 6Li chemical shifts are dominated by the proximate phenolates (local environment).
Trimers are rarely observed: only in DME and invariably as components of mixtures.26 Although special characterization methods required by mixtures are discussed below, a few comments about the symmetries of cyclic trimer 3 versus ladder 33 are appropriate here.27
Cyclic homotrimer 3 would display a single resonance (Chart 2), whereas the corresponding ladder 33 would display a 2:1 pair of resonances provided that the exchange of lithium nuclei within the aggregates—so-called intraaggregate exchange—is slow.28 In theory, A2B1 and A1B2 cyclic trimers would appear as two pairs of 2:1 resonances (four total), whereas ladder forms of A2B1 and A1B2 would afford, by virtue of positional isomerism (depicted below), as many as ten distinct resonances. In practice, single resonances observed for A3 and B3 homoaggregated trimers seem to implicate the cyclic trimer 3 motif. (Cyclic trimer derived from 19 appears as a single sharp resonance even when higher resolution 19F NMR spectroscopy is used.) A2B1 and A1B2 heterotrimers, however, also appear as single, sharp resonances, indicating that intraaggregate exchange is rapid on 6Li NMR time scales.28 The high fluxionality, previously noted in DME-solvated lithium enolates,7 is attributed to hemilability of DME as posited in Scheme 1. Thus, the cyclic and ladder motifs for trimers are indistinguishable by the apparent symmetry.
Pentamers are prevalent only at low DME concentrations (<2 equiv/Li). Homoaggregated ladder 5,27,29 the only pentameric form we considered seriously, should appear as a 2:2:1 triad of resonances. In practice, however, pentamers appear as very broad downfield 6Li resonances at −90 °C that sharpen to single resonances and shift markedly upfield when the probe is warmed to > −50 °C. The broadness appears to derive from partially decoalesced intraaggregate exchange, and the temperature-dependent chemical shift may result from changing solvation numbers (vide infra). Poor solubilities at low DME concentrations precluded using the lower probe temperatures required to observe pentamers by 6Li NMR spectroscopy in the slow exchange limit. Fortunately, 19F NMR spectra of fluorinated phenolates 11 and 12 revealed a well-resolved 2:2:1 triad in each case. Pentamer-derived Job plots (vide infra) are also compelling.
Ensembles of pentamers derived from A5-B5 mixtures could have posed serious problems because positional isomerism (illustrated emblematically for the A4B1 form below) would afford 84 magnetically inequivalent 6Li nuclei in an A5-B5 ensemble. Indeed, spectra recorded at the lowest attainable temperatures are useless. Warming the probe above −50 °C, however, elicits rapid intraaggregate exchange,28 causing the six stoichiometrically distinct AnB5-n pentamers to appear as six well-resolved resonances (Figure 1c). This exchange removes the inherent asymmetry of the ladders and the problems created by positional isomerism.6
Plots of aggregate concentrations versus mole fractions of subunits A and B—so-called Job plots10—representing the five observed aggregation states (monomers through pentamers) are illustrated in Figures 2–6. The parametric fits show that all are nearly statistical distributions; non-statistical ensembles usually attest to deep-seated structural anomalies (vide infra). The math underlying the parametric fits has been discussed previously.6,7
The protocols used to measure aggregate concentrations for the Job plots require comment. The actual mole fractions of the A and B subunits can deviate from the intended mole fractions because of experimental error, nonquantitative lithiation, impurities, selective formation of mixed aggregates24 with base (especially in DME),30,31 formation of by-products, or multiple homoaggregates (vide infra). (Some of these deviations are, admittedly, unlikely with lithium phenolates.) Although such deviations do not impair the qualitative analysis of aggregate numbers and their symmetries, they certainly distort (sometimes profoundly) the concentrations of the available A and B subunits from the intended concentrations. Consequently, the mole fractions of the A and B subunits were measured (rather than inferred from the concentrations of added phenols) by integrating the 6Li resonances within an ensemble. Even in the cleanest and simplest cases wherein the intended and the measured mole fractions are comparable, we believe the latter to be more accurate. Job plots for monomers, however, must use the intended mole fraction because the measured mole fraction would necessarily afford perfect lines (no scatter).
Deviations from statistical behavior can occur for a variety of reasons. For example, a mixture of 25 and 27 in DME affords a single new resonance consistent with formation of a heterodimer, but with a nonstatistical (unusually high) preference. We suspect that heteroaggregation is promoted by severe steric interactions in the homodimer of 27.7,32 When distributions deviate significantly from statistical predictions, pairing substrates that are similar is constructive. Indeed, mixtures of 25 and 26, differing only in the absence or presence of a halogen moiety in the para position, afford statistical distributions.33
One also observes homoaggregates that resist heteroaggregation. The obvious case occurs when two monomers are paired (30 and 31, for example). In a more instructive example, mixtures of 11 and 14 in DME show only traces of a new species assigned by symmetry as a heterodimer. Pairing 7 and 11 shows that 11 is tetrameric (a statistical ensemble), whereas pairing 14 and 24 reveals 14 to be dimeric. Thus, heteroaggregation is precluded when the homoaggregates are of two different aggregation states. In the most potentially deceptive case, a mixture of tetrameric 9 and dimeric 16 in 2.9 M DME afforded an ensemble that could easily be confused with an all-dimer A2-AB-B2 ensemble. The tetramer assignment of 9 was made using independent pairings and solvent swaps (vide infra).
The most striking and provocative deviations from statistical behavior center around DME-solvated dimers. Pairing 2,6-diphenyl-substituted phenolate 28 with isostructural 4-bromo analog 29 affords a statistical A2-AB-B2 ensemble; phenolate 28 in DME is clearly a dimer. Nonetheless, 28 fails to heterodimerize with seemingly similar dimeric phenolates 25 or 27, suggesting fundamental structural differences (vide infra). Similarly, dimeric 16 and other polyhalogenated phenolates do not statistically heterodimerize with seemingly analogous dimeric hindered phenolates (such as 27) or orthophenylated lithium phenolate 28. There appear to be three types of dimers; these are discussed further below.
A simple protocol can be used to show whether a change in solvent is accompanied by a change in aggregation.7,34 This protocol serves as a control experiment to confirm conclusions suggested by other data and can provide structural assignments to even minor components of mixtures.
The experiment is based on the observation that solvent-solvent exchanges (ligand substitutions) occur much faster than aggregate-aggregate exchanges.35 By recording a series of spectra in which one coordinating solvent is incrementally replaced (swapped) by a second, either of two limiting behaviors can be observed: (1) If the observable forms in the two coordinating solvents differ only by ligating solvent, the incremental solvent swap will cause the resonances to exchange by time-averaging (Figure 7a); (2) if the observable forms in the two solvents differ by aggregation number, an incremental solvent swap causes one aggregate to disappear and the other to appear (Figure 7b). The experiment requires a 6Li chemical shift difference in the two solvents of as little as 0.05 ppm. Experimental data are illustrated with DME- and THF-solvated dimers of 23 (Figure 8) and with a DME-solvated dimer and THF-solvated tetramer of 14 (Figure 9).36
To the extent that an An-Bn ensemble represents a single aggregation number (all tetramers, for example), serial dilution should confirm that the proportions of aggregates within the ensemble remain unchanged. In practice, we find that the concentration range needed to provide compelling affirmation of the assignment presents technical challenges. We carry out such control experiments but consider them of minor importance relative to more compelling data supporting aggregation state assignments. Concentration changes can, however, uncover and offer insights into homoaggregate mixtures.
Varying the absolute lithium phenolate concentration or, more consequential, the donor solvent concentration often reveals additional homoaggregates and their relative solvation and aggregation numbers. Full characterization of such mixtures usually requires a combination of methods. Both qualitative and quantitative concentration dependencies are discussed in the next section. First, we describe Job plots on mixtures of homoaggregates.
Mixing a previously characterized aggregate such as an A2 dimer with a Bm-Bn binary mixture in which one of the homoaggregates is a dimer typically affords a statistical A2-AB-B2 ensemble with one homoaggregate, leaving the other homoaggregate untouched. The success of the Job plot does not demand structural homogeneity. In some instances, mixing an Am-An binary mixture with an analogous Bm-Bn binary mixture affords independent Am-Bm and An-Bn ensembles (Figure 10). Such multiple ensembles can be engineered by pairing structurally similar phenolates, especially those differing only in the absence or presence of a para halogen substituent.
The power of solvent swapping becomes evident when deconvoluting mixtures of homoaggregates. Combining a tetramer-dimer mixture of 13 with dimeric 27 in DME, for example, affords a statistical dimer-based ensemble, leaving the second aggregate of 13 (suspected to be tetramer) isolated. Swapping DME for THF (Figure 11) in the tetramer-dimer mixture of 13 shows that the putative tetramer time-averages with the fully characterized THF-solvated tetramer. Solvent swapping can even be used to identify individual components of ternary mixtures of homoaggregates.37
The crystallographic literature of lithium enolates and phenolates is replete with tetrasolvated tetramers.5,21 If one is willing to use these as benchmarks (which we are), quantitative monitoring of tetramer-dimer ratio versus THF concentration can reveal the solvation number of the dimers. Results using lithium phenolate 18 (Figure 12; eq 2) show that the dimers are tetrasolvated.
THF-solvated monomers are only observed using acutely hindered lithium phenolates 30 and 32. The extreme steric demand attenuates the importance of the studies because (a) the substrates are not emblematic of normal phenolates, and (b) the solvation numbers of the dimer to which the monomer assignments are anchored are not firmly established. With that said, the solvation numbers are remarkably high. Varying the THF concentrations in solutions containing monomers and dimers of 32 reveals 3.0 additional THF ligands per lithium on the monomer relative to the dimer (Figure 13), which implicates tetra- or pentasolvated monomers, depending on the solvation number of the dimer (eqs 3 and 4). We know that the dimer contains at least one solvent because of a marked shift that occurs when low concentrations of THF are added to toluene solutions of 30.
Varying the DME concentrations in solutions of structurally diverse DME solvates offers insights and surprises. DME-concentration-independent dimers, trimers, and tetramers attest to one solvent per lithium (eq 5).
Solvation numbers for DME-solvated monomers are assigned by monitoring the dimer-monomer ratio of phenolate 32. Once again, measured solvation numbers are remarkably high, but the results are also somewhat quirky. Varying the DME concentrations shows that the monomer of 32 differs from the dimer by having 1.5 additional DME ligands per lithium on the monomer relative to the dimer. We know that the dimer is solvated by DME because of a marked shift that occurs when low concentrations of DME are added to toluene solutions of 32. If this dimer is monosolvated—a structural motif that could result from severe buttressing—then the monomer is a disolvate (eq 6).38 If the dimer is a disolvate, then the monomer solvation number is suggested to be 2.5 (eq 7). Such a fractional solvent order could stem from experimental error or a mixture of solvated forms. High solvation numbers are considered in the discussion.
The preference for DME-solvated pentamers only at very low DME concentrations attests to a low per-lithium solvation number, quite logically containing two DME ligands capping the ladder ends as chelates (see below). A significant (≈0.3 ppm) upfield shift of the time-averaged pentamer resonance with increasing temperature, however, suggests facile desolvation39 of at least one internal lithium. Solvent concentration dependencies of pentamer-tetramer mixtures of phenolate 8 at −30 °C are quantitatively consistent with pentamers containing an average of three DME ligands (eq 8). The concentration dependencies suggest that tetrasolvation at low temperature (−90 °C) is giving way to trisolvation on warming. Computations proved both supportive and instructive (vide infra).
Computational studies of lithium enolates, phenolates, and simple alkoxides have been reported.40,41 We have carried out additional density functional theory (DFT) computations at the B3LYP/6-31G(d) level42 addressing issues of particular interest to the structural studies described herein. Me2O was used as a model for both THF and η1-DME to reduce the degrees of freedom. Free energies (ΔG) are reported on a per-lithium basis in kcal/mol at −90 °C. The results for the generic deaggregations in Scheme 2 are listed in Table 3. The solvation numbers and importance of chelation are consistent with the computational data archived in supporting information and seem to match experiment reasonably well. We largely refrain from detailed discussions of the absolute free energies (ΔG1-ΔG6), instead focusing on the influence of substituents and solvents (trends). The substituent dependencies of ΔG1-ΔG6 in Table 3 are considered in the discussion.
We experimentally detected what appears to be three classes of lithium phenolate dimers manifesting (1) intermediate steric demands, (2) high electron deficiency (polyfluorination), and (3) 2,6-diphenylation. Experimentally, dimers within one class afford decidedly nonstatistical heterodimer ensembles with phenolates from another class. We examined this finding computationally using lithium phenolates 25 and 28 (eq 9). In short, the computations showed no significant deviations of the homodimer-heterodimer equilibrium from statistical, underscoring an uncontestable disagreement between theory and experiment. In this context, we re-evaluated the experimental data and stand behind it.
Cursory computational studies of the rarely observed DME-solvated trimers showed triply chelated cyclic form 43 to be more stable than ladder 42 bearing an η1 solvate. We proffered the ladder as an intermediate in the rapid intraaggregate exchange (Scheme 1).
Computations evaluated the geometry and stabilities of DME-solvated pentameric ladders, suggested by experiment to be a temperature-sensitive mixture of tri- and tetrasolvates. Three geometries—W, U, and a hybrid (44–46)—are stable minima in their unsolvated, doubly chelated forms; the U form (47) is the most stable. Me2O was used to model the solvation of the internal lithiums by monodentate (η1) DME ligands (Scheme 3). Tetrasolvate 51 is calculated to be preferred at temperatures below −74 °C, whereas trisolvate 50 is favored above −74 °C. Thus, the computations successfully mimic the putative temperature-sensitive desolvation detected experimentally. We were also struck by the allosteric effects, phenomena we have referred to as correlated solvation.43
Lithium salts are centrally important in such disparate fields as organic synthesis,44 polymer science,45 electrochemistry,46 medicine,47 and automotive lubricants.48 We have been refining protocols that are based on the method of continuous variation—the method of Job—for characterizing O-lithiated species.10,11 Our interest in the phenolates stems from our desire to develop and refine methods for characterizing their more synthetically important lithium enolate counterparts. Nonetheless, lithium phenolates stand on their own merits because of their stability, purity, availability, substrate flexibility, and aggregate diversity. Examination of a broad array of lithium phenolates in THF and DME solutions revealed a remarkable range of aggregates from monomers through pentamers (1–5) and underscored the versatility of these methods.
The problems associated with characterizing LiX species in which the X group is magnetically opaque have been long-standing and acute.49 The seemingly most general and certainly most popular methods involve measuring colligative properties of solutions, which afford average molecular weights.2,4,5,19,45,49 Although colligative measurements can provide valuable supporting evidence of structure, as stand-alone methods they are fraught with risk because of problems caused by undetected impurities and mixtures of structural forms. Methods that are also germane to our tactic involve quantitatively monitoring LiX-LiY mixed aggregates constituted from the LiX salt of unknown structure and an LiY salt of known structure.3,12–15,50,51 In general, progress has been made—enolate studies of Streitwieser and coworkers are certainly notable3—but the strategies used are often quite specific to the LiX-solvent combination.
Of particular importance to the work described herein, Jackman and coworkers investigated the structures of lithium phenolates using a combination of methods in which 7Li quadrupolar splitting constants and predicted chemical shifts of 13C resonances derived from the para carbons were of paramount importance.5 Obviously their method applies only to lithium phenolates and not other O-lithiated species. They also focused on a smaller and different subset of lithium phenolates as well as a number of solvents not described herein. Their emphasis on enthalpic and entropic differences in aggregates to ascertain relative solvation numbers appears to be challenging at best. But the congruence of our results and conclusions and those of Jackman and coworkers are notable despite the use of distinctly different approaches.
Our strategy for characterizing lithium phenolates entails breaking the high inherent symmetry (spectral simplicity) of (ROLi)n aggregates by mixing two homoaggregates (An and Bn) and generating ensembles containing both homo- and heteroaggregates (eq 1).6,7 The numbers and symmetries of the heteroaggregates, as shown by 6Li NMR spectroscopy, offer considerable insights into the structures of the homoaggregates (Table 2; Chart 2). By systematically varying the mole fractions of the A and B subunits, we can use parametric fits to afford the Job plots illustrated emblematically in Figures 2–6. Statistical distributions observed for structurally similar phenolates (some varying by as little as a para halogen substituent) confirm the structural assignments, whereas marked deviations from statistical usually foreshadow different aggregation numbers for Am and Bn. Once An is characterized in one solvent (such as DME) and an aggregate of unknown aggregation number is generated in a different solvent (THF), one can trivially ascertain if the aggregation number differs in the two solvents by a simple experiment illustrated in Figures 7–9 in which DME is incrementally replaced (swapped) by THF. Lithium phenolates distinguished only by the coordinated ligands display a single time-averaged 6Li resonance, whereas phenolates of differing aggregation number appear as discrete resonances.
Varying the solvent and phenolate concentrations often reveals additional homoaggregates and provides qualitative insight into their relative aggregation and solvation numbers. In some instances, combining Am-An mixtures with Bm-Bn mixtures afford two fully separate Am-Bm and An-Bn ensembles that are easily and independently quantified. In other cases, solvent swaps reveal the structures of homoaggregates irrespective of the total number of homoaggregates. In fact, one of the more satisfying outcomes is that essentially any mixture can be characterized provided that discrete resonances are observed. One cannot, however, overstate the importance of using complementary methods as well as multiple combinations of aggregates. Overdetermination of structure is important.
THF and DME were used to examine the influence of mono- and bifunctional ligands on lithium phenolate aggregation. THF affords tetrameric, dimeric, and monomeric lithium phenolates depending on solvent and substrate concentrations and on the aryl substituents (discussed below). By contrast, DME affords five aggregation states, a diversity that may stem in part from its capacity to serve as either a monodentate or bidentate (chelating) ligand.52 The occurrence of cyclic trimers (3) and pentameric ladders (5) uniquely in DME suggests that chelation is mandatory. Stabilization by chelation is supported by computational studies (Scheme 2; Table 3). Qualitative data from the solvent swaps suggest that the computations may, however, overestimate the relative stability of the DME-solvated dimers (37).36
THF solvation numbers were assigned by monitoring how the tetramer, dimer, and monomer concentrations varied with THF concentration (Scheme 4). The tetramer and dimer numbers enjoy full computational support. The monomer solvation numbers, by contrast, are a little more difficult to interpret because only the extraordinarily hindered 2,6-di-tert-buyl derivatives afford monomers. Certainly phenolate 30 is not representative of phenolates in general. The data, however, support tetra- or pentasolvates 54 and 55, respectively. Tetrasolvated monomer (Me3Si)2NLi(THF)4,54 as well as other evidence of high-coordinate lithium, support the tetrasolvate 54.53–57 Putative pentasolvate 55 is supported by crystallographically characterized +Li(THF)5 and +Li(THF)6 cations.54 Nonetheless, such high solvation numbers were unanticipated and certainly are inconsistent with conventional wisdom.
Analogous studies of DME-solvated aggregates afford results comparable to those found with THF (Scheme 5). Equivalent per-lithium solvation numbers of the tetramers (56) and dimers (57) are consistent with η1 and η2-coordinated DME, respectively. The η1-DME ligands on 56 may seem a little odd on first inspection, but their presence is supported by analogy to crystallographic and solution structural studies of other DME-solvated organolithiums.58 Chelation in dimer 57 has ample crystallographic and computational support, requiring little further justification.21
Once again, the monomers give us pause because of the extraordinary steric demands of phenolate 30 required to observe them. Moreover, the solvation numbers of the monomers are measured relative to the dimers, but we are not fully confident that such hindered dimers are doubly solvated. Consequently, doubly-chelated monomer 58 and ion-paired monomer 59 are both consistent with the data. Five-coordinate LiX(η2-DME)2 complexes support the proposed structure of monomer 58,55,56 and the well-precedented +Li(η2-DME)3 cation supports that of 59.57 There is a significant argument against 59: A mixture of putative ion pairs derived from lithium phenolates 30 and 31 show two distinct lithium cations. The origin of such a slow gegenion exchange is unclear. Regardless of the details, the high solvation numbers are surprising.
Trimers are observed only in DME, for a limited number of phenolates, and never as major components. The per-lithium solvation number of DME-solvated trimers is the same as those of dimers and tetramers (one per lithium). Rapid 6Li intraaggregate exchange (speculatively presented in Scheme 1) precluded distinction of a ladder versus a cyclic form of the trimers; computational studies support the cyclic form (eq 10). Based on analogy with other lithium derivatives (lithium amides in particular) in which solvated trimers appear to be sterically more problematic than their cyclic dimer counterparts, we suspect the low steric demands of DME chelates may be important.59
The exotic and unanticipated pentameric ladders are prevalent at low DME concentrations. Rapid intraaggregate exchange obscures their asymmetry using 6Li NMR spectroscopy, but 2:2:1 triads of resonances were observed using 19F NMR spectroscopy on two fluorinated substrates. At the outset, the low per-lithium solvation number suggested that the ends of the ladders are capped by chelating DME ligands. Concentration and temperature dependencies evidence weak solvation of one of the internal lithiums. Computations support a ladder capped by chelating DME ligands and indicate a marked preference for U-shaped rather than W-shaped ladders (45 and 44, respectively). Moreover, the computations support partial solvation of the internal lithiums (Scheme 3). It is not obvious, however, why pentameric ladders are the only observable ladder forms (assuming, of course, that the trimers are cyclic). Although U-shaped (concave) hexameric ladders would be sterically problematic, tetrameric ladders seem plausible at low DME concentration but are not observed.60
Lithium phenolates are ideally suited for examining the influence of electronic effects on aggregation. Monohalogenated phenolates 11–13 are unperturbed enough to elicit measurable aggregation state changes. By contrast, computations of multiply halogenated phenolates 14–17 clearly show that halogens promote deaggregation (Table 1).61 Moreover, comparing phenolate 15 and 16 under similar conditions shows that the meta fluoro moieties promote deaggregation more so than para moities do. There are parallels with the pKAs of 3- and 4-phenols.62 The meta substituent and its W relationship to the anionic phenolate oxygen appears optimally oriented to influence aggregation. Computational studies comparing monofluorinated lithium phenolates with the parent phenolate 7 show that ortho>meta>para at promoting deaggregations. (Apparent Li-F contacts in THF-solvates obscure insights into the electronic effects of ortho fluoro moieties.)
Considerable empirical evidence suggests that high steric demands promote deaggregations of lithium salts because of the severe van der Waals interactions within the aggregates.32 Phenolates 18–32 are sterically demanding to some extent, and all show a greater penchant for forming dimers than do simple phenolates. Only the most severely congested phenolates buttressed by two tert-butyl groups (30–32) afford monomers. Computations comparing parent phenolate 7 and 2,6-dimethyl phenolate 25 (Table 3) confirm that congestion promotes dimers relative to tetramers and monomers relative to dimers.
It seems self-evident—a truism—that LiX salts are most likely to modify the reactivity and selectivity of organolithium reactions if they form mixed aggregates.24,44 One of the most interesting and potentially important observations, therefore, is that heteroaggregation is promoted by pairing species that are similar.
A4 and B4 mixtures usually afford ensembles of tetramers that are nearly statistical. The same is true for other analogous An-Bn pairs. By contrast, A4 tetramers and B2 dimers resist heteroaggregation. It seems logical in retrospect that a homoaggregate would resist heteroaggregation if it meant forfeiting a preferred aggregation number. For this reason, we place great importance on statistical ensembles. Of course, there are exceptions. We observed that A4-B2 mixtures derived from 9 and 16 (respectively) in THF afford no heterotetramers but do afford heterodimers. Homoaggregates that are on the cusp of shifting aggregation number are likely to be more promiscuous at heteroaggregation.
There appear to be three albeit loosely defined classes of dimers: (1) sterically congested lithium phenolates (18–27), (2) highly electron-deficient lithium phenolates (14–17), and (3) 2,6-diphenylated lithium phenolates (28 and 29). Pairs chosen from two different classes tend to form nonstatistical ensembles (sometimes markedly so). Classes 1 and 2 form heteroaggregates that are prone to deviate from statistical distributions in both directions. 2,6-Diphenylated substrates 28 and 29 present the most vexing category.63 They seem to have steric demands akin to dimeric phenolates 25–27 and electronic properties that could be similar to polyfluorinated derivatives 14–17, yet they form heterodimers only with each other, not with class 1 or 2 dimers. It was somewhat disappointing and even somewhat disconcerting that computational studies completely failed to mimic these results, affording statistical distributions for all combinations (eq 9).
There may be a niche market for insights into the structures of lithium phenolates. The work described herein, however, has broader implications. Characterizing lithium salts with an NMR-inactive gegenion has been a long-standing problem of considerable importance. The strategies and tactics described in this manuscript appear to offer general solutions, especially for O-lithiated species.6,7 It became increasingly apparent as the work progressed that the view of substrate-dependent aggregation provided by lithium phenolates may be unique.5 Those interested in understanding the solution structures of lithium halides have a very limited choice.49 Similarly, we have studied lithium amides for almost 25 years,39 yet we have examined in detail less than a dozen variations of the dialkylamido moiety, which is hardly systematic. The breadth and number of alkyllithiums studied to date is impressive,8 but an RLi dependence is difficult to systematize. By contrast, lithium phenolates allow for a reasonably systematic study of how changing X in LiX salts influences aggregation state.
Our interest in lithium phenolates as vehicles for studying salt effects in solution has been piqued and will persist. The advantages and synergies of studying several solvents concurrently became evident in these studies and, in conjunction with Jackman’s early studies,5 should gain momentum with each additional solvent. The low basicity of lithium phenolates should allow us to study metal ion solvation by a very broad range of solvents, including many that are incompatible with more reactive organolithium derivatives.
DME, THF, and toluene were distilled from solutions containing sodium benzophenone ketyl. The toluene stills contained approximately 1% tetraglyme to dissolve the ketyl. We prepared and recrystallized [6Li]LiHMDS and [6Li,15N]LiHMDS as described previously.31 Air- and moisture-sensitive materials were manipulated under argon using standard glove box, vacuum line, and syringe techniques. The precursor to 29 was prepared using standard halogenation procedures.64
Individual stock solutions of substrates and base were prepared at room temperature. An NMR tube under vacuum was flame dried on a Schlenk line and allowed to come to room temperature. It was then backfilled with argon and placed in a −78 °C dry ice/acetone bath. The appropriate amounts of the base and substrate were added sequentially via syringe. The tube was sealed under partial vacuum, vortexed for approximately 10 seconds at room temperature, and cooled to −78 °C. Each NMR sample contained 0.10 M total phenolate and 0.11 M LiHMDS.
6Li NMR spectra were typically recorded at −90 °C (unless stated otherwise) on a 400 or 500 MHz spectrometer with a delay between scans set to >5 × T1 to ensure accurate integrations. Chemical shifts are reported relative to a 0.30 M 6LiCl/MeOH standard at the reported probe temperature. The resonances were integrated using standard software accompanying the spectrometers. After weighted Fourier transform with 64,000 points and phasing, line broadening was set between 0 and 0.2, and a baseline correction was applied when appropriate. Deconvolution was performed in the absolute intensity mode, with application of a drift correction using default parameters for contributions from Lorentzian and Gaussian line shapes. For poorly resolved spectra, the resonances were indicated using the mark and use-mark commands. The math underlying the parametric fits has been described in detail.6,7a
We thank the National Institutes of Health (GM 077167) for direct support of this work and Pfizer, Boehringer-Ingelheim, and Sanofi-Aventis for indirect support.