Establishing the Role of Triflate Anions in H₂ Activation by a Cationic Triorganotin(IV) Lewis Acid

Abstract

Cationic Lewis acids (LAs) are gaining interest as targets for frustrated Lewis pair (FLP)-mediated catalysis. Unlike neutral boranes, which are the most prevalent LAs for FLP hydrogenations, the Lewis acidity of cations can be tuned through modulation of the counteranion; however, detailed studies on such anion effects are currently lacking in the literature. Herein, we present experimental and computational studies which probe the mechanism of H₂ activation using iPr₃SnOTf (1-OTf) in conjunction with a coordinating (quinuclidine; qui) and noncoordinating (2,4,6-collidine; col) base and compare its reactivity with {iPr₃Sn·base}{Al[OC(CF₃)₃]₄} (base = qui/col) systems which lack a coordinating anion to investigate the active species responsible for H₂ activation and hence resolve any mechanistic roles for OTf^– in the iPr₃SnOTf-mediated pathway.

Introduction

Frustrated Lewis pairs (FLPs), where bulky Lewis acid (LA) and Lewis base (LB) partners are sterically precluded from forming strong classical adducts, are now well-established systems for small-molecule activation with main group compounds.^1,2 Perhaps most notably, H₂ heterolysis can be effected by FLPs to form H⁺ and H^– equivalents, which can be delivered to unsaturated bonds to form a reduced compound and the original FLP. Accordingly, FLP-catalyzed hydrogenations of organic compounds containing the C=X (X = C, N, and O) bonds have received widespread attention and have been thoroughly studied.³ The LAs used in these reactions are predominantly triarylboranes BAr₃, particularly B(C₆F₅)₃ and similar derivatives⁴ thereof, where the electronic and steric profiles have been tuned by the variation of aryl substituents (although these are predominantly confined to H, F, and Cl),⁵ which can afford a systematic manipulation of BAr₃ Lewis acidity. Accordingly, novel BAr₃ have recently been developed that can catalyze the hydrogenations of substrate classes previously inaccessible by FLPs or display the catalytic activity in the presence of alcohols/moisture, yet accessing such heteroleptic boranes requires laborious multistep syntheses.^5e,6 Although some aspects of the H₂ activation mechanism between the BAr₃ and N- or P-centered bases are still being debated,⁷ it is generally assumed that facile H₂ heterolysis is associated with the initial formation of a noncovalent LA–LB “encounter complex”, which subsequently reacts with H₂ in a cooperative manner (i.e., a concerted mechanism via a single transition state).⁸ Important exceptions are “frustrated radical pairs”, which activate substrates via single electron pathways.⁹ Recently, the range of LAs displaying FLP-like reactivity has expanded to encompass a number of cationic species (or neutral species that act as surrogates for the active cationic LA) based on B(III),¹⁰ Si(IV),¹¹ C(IV),¹² N(III),¹³ P(V),¹⁴ and Sb(V).^15,16

Tuning the LA strength, and hence the reactivity, of these LAs may also be achieved through ligand design within the formally cationic fragment, akin to BAr₃ compounds; however, their anions also provide an additional locus for variation and hence a comparatively simple route to optimize the Lewis acidity of the cationic species. For instance, replacement of [TfO]⁻ for [B(C₆F₅)₄]⁻ in phosphonium or stibonium LAs can switch on otherwise absent reactivity (see Figure 1), which has been rationalized only by reference to the relative sequestration of the cation by more strongly coordinating anions, resulting in weaker LAs. A detailed analysis of any intimate effects the anion may exert during a reaction (e.g. by its coordination within a transition state, which could perturb its energy) have not been previously undertaken for cationic LAs.

Figure 1.

Examples of cationic LAs where reactivity was promoted by using a weakly coordinating counteranion (top) and the stannylium-based LA/base systems investigated for H₂ activation in this study (bottom). R^F = C(CF₃)₃.

We reported previously that iPr₃SnOTf (1-OTf; Tf = O₂SCF₃), which acts as a surrogate for the stannylium ion [iPr₃Sn]⁺ ([1]⁺), activates H₂ in conjunction with LBs (e.g., 2,4,6-collidine; col) and catalyzes the hydrogenation of imines and ketones and the reductive amination of carbonyls.¹⁷ Despite the diversity of hydrogenation substrates accessible by 1-OTf, which can operate in the presence of moisture and strong LBs, productive hydrogenation using 1-OTf/col is rather slow and requires forcing conditions (>120 °C, 10 bar H₂). A thorough understanding of the factors affecting 1-OTf/LB catalytic hydrogenations would clearly aid the development of superior R₃SnX LA hydrogenation catalysts, although it should be borne in mind that this will require a more complicated treatment, cf. B-based FLPs. This is predominantly due to the presence of the (hemi-labile) TfO anion and the fact that the [R₃Sn]⁺ species are known to readily bind two anionic/neutral donors (as opposed to only one for BAr₃); such five-coordinate adducts may also be important off-cycle entities in catalysis. A previous computational study by Pati et al. suggested that the weakly associated complexes of 1-OTf and either DABCO (1,4-diazabicyclo[2.2.2]octane) or col cleave H₂ via an FLP-type concerted mechanism¹⁸ analogous to that seen in the FLP reactivity of the electrophilic [(C₆F₅)₃P-X]⁺ species.^14a−14d Pati et al. assumed that 1-OTf was the only active LA despite the experimental evidence for the formation of [1·(LB)]⁺ adducts (e.g. ¹¹⁹Sn NMR shifts) in 1-OTf/LB mixtures;^17a accordingly, any role of [1·(LB)]⁺ species was not investigated, nor the potential impact of charged hypervalent Sn species upon reductions mediated by 1-OTf, all of which could arise from the displacement of labile [TfO]⁻ during the reaction pathway (see Scheme 1). Clearly, ascertaining the correct catalytic speciation in these more complex systems is crucial if targeted improvements in activity/selectivity are to proceed in a logical and predictive manner. Herein, we report a comprehensive study into the mechanism of H₂ activation using 1-OTf/LB pairs, using complementary experimental and theoretical methods, which reveal the noninnocent nature of [TfO]⁻ in influencing both kinetic and thermodynamic factors. These results provide fundamental insight into how heavier-element cationic Lewis pairs activate H₂, which will help guide the design of new cationic p-block LA catalysts by encouraging further research into the role and tunability of the anion in reactions featuring such LAs.

Scheme 1. Possible Sn-Containing Species in 1-OTf + LB Combinations and Mechanistic Pathways for H₂ Activation.

Results and Discussion

Identifying Ground-State Speciation in 1-OTf/Quinuclidine Lewis Pairs

Because DABCO contains two donor atoms, which in theory could both interact with LAs and accordingly complicate the identification of equilibria speciation in 1-OTf/LB pairs, we chose instead to study the interaction of 1-OTf with quinuclidine (qui) because of its similar basicity [pK_a(MeCN): [DABCO-H]⁺ 18.3;¹⁹ [qui-H]⁺ 19.5]²⁰ and steric profile. 1-OTf exhibits a broad peak in its ¹¹⁹Sn{¹H} NMR spectrum at δ = 166 ppm in 1,2-difluorobenzene (DFB),²¹ which moves to 34 ppm and sharpens upon addition of 1 equiv of qui at room temperature (RT), suggesting a spontaneous association to form (qui)·1-OTf and/or formation of the ion pair [1·(qui)]⁺[OTf]⁻.

Gratifyingly, slow cooling of this solution yielded single crystals that were analyzed by X-ray diffraction; the solved structure (Figure 2) reveals the former species: a five-coordinate adduct in which OTf^– and the qui coordinate trans across the [1]⁺ core, with a trigonal bipyramidal geometry for the Sn atom [bond angles (°): C–Sn–N = 91.41(9), 93.73(9), and 93.76(9); O1–Sn–N = 177.40(7)]. Although this (qui)·1-OTf adduct represents the ground-state structure in the solid phase, it is possible that [1·(qui)]⁺[OTf]⁻ is accessible at RT in solution. In order to spectroscopically identify [1·(qui)]⁺ as a distinct moiety, it was necessary to characterize it in the absence of any exchange equilibria, which was achieved by exchanging [OTf]⁻ for the much more weakly coordinating aluminate {Al[OC(CF₃)₃]₄}⁻ ([Al(OR^F)₄]⁻); this was chosen because it is particularly robust to strongly Lewis and Brønsted acidic media.²² Thus, the reaction of K[Al(OR^F)₄], 1-OTf, and qui (1:1:1) in DFB resulted in the precipitation of KOTf, with the ¹⁹F NMR spectrum of the filtrate showing only a single signal for the aluminate anion, demonstrating clean anion metathesis (Scheme 2). Layering this solution with pentane yielded crystals suitable for X-ray diffraction studies; the structural solution to these data is represented in Figure 3 and corresponds to the salt [1·(qui)]⁺[Al(OR^F)₄]⁻.²³

Figure 2.

Oak Ridge Thermal Ellipsoid Plot (ORTEP) diagram of (qui)·1-OTf. Thermal ellipsoids set at 50% probability. C atoms in black, F atoms in green, N atom in blue, O atoms in red, S atom in yellow, and Sn atom in pink. The Sn–N and Sn–O bond lengths are given in Å with estimated standard deviations (ESDs) in parentheses. H atoms are omitted for clarity.

Scheme 2. Syntheses of [1·(LB)]⁺[Al(OR^F)₄]⁻; LB = qui or col (Collidine, 2,4,6-Trimethylpyridine).

Figure 3.

ORTEP diagram of [1·(qui)]⁺[Al(OR^F)₄]⁻. Thermal ellipsoids set at 50% probability. Al atom in orange, C atoms in black, F atoms in green, N atom in blue, O atoms in red, and Sn atom in pink. The Sn–N bond length is given in Å with the ESD in parentheses. The H atoms are omitted for clarity.

As anticipated, the [Al(OR^F)₄]⁻ ion does not display any close contacts to the [1]⁺ unit and therefore does not coordinate to the LA Sn atom. Consequently, the [1·(qui)]⁺ assembly adopts a distorted tetrahedral geometry around the Sn center [N–Sn–C bond angles (°) = 103.7(2), 102.8(2), 102.9(3)]. The Sn–N bond distance is considerably shorter than that observed for the qui·1-OTf adduct [2.448(2) Å], which is consistent with a Sn–N weakening in the latter because of a trans-labilizing effect of the OTf^– and/or increased steric encumbrance. The ¹H NMR spectroscopic analysis of the DFB solutions of [1·(qui)]⁺[Al(OR^F)₄]⁻ showed that the observed chemical shifts and couplings for the iPr groups are similar to those for 1-OTf in DFB; the relative integration of these with the signals for qui confirms a 1:1 ratio between [iPr₃Sn]⁺ and qui. The ¹¹⁹Sn{¹H} NMR spectrum, however, conclusively shows the formation of a new stannylium adduct with a sharp resonance at δ = 113 ppm, which is upfield from 1-OTf (δ = 166 ppm), but notably downfield of that seen for 1-OTf/qui mixtures (δ = 34 ppm). Taken together, these spectroscopic data confirm that the gross solid-state structure of [1·(qui)]⁺[Al(OR^F)₄]⁻ is retained in the solution. Curiously, the ¹¹⁹Sn{¹H} NMR spectroscopic data for the adduct formed from 1-OTf and qui in situ, or by redissolution of (qui)·1-OTf crystals obtained thereof, are inconsistent with those of either authentic 1-OTf or [1·(qui)]⁺. This suggests that neither compound is present exclusively. It is plausible that a rapid equilibrium containing 1-OTf, (qui)·1-OTf, and [1·(qui)]⁺ exists in the solution. To understand better the speciation, we assessed the relative energies of these species computationally (Scheme 3), with our density functional theory (DFT) calculations²⁴ predicting the formation of a ternary complex (qui)·1-OTf to be clearly favored thermodynamically with respect to qui + 1-OTf, in contrast with the previously obtained results for the related DABCO + 1-OTf system.^18,25 Additionally, dissociation of the adduct (qui)·1-OTf into [1·(qui)]⁺ + [OTf]⁻ is also found to be thermodynamically feasible, with the latter computed to be 4.8 kcal/mol more stable than the original qui + 1-OTf reference state.

Scheme 3. Possible Reactant-State Species in the Solution Phase for the 1-OTf + qui Lewis Pair; Computed Relative Stabilities (in kcal/mol at 298 K) Are Shown in Parentheses²⁴.

We also probed two other possible adducts: one where the dissociated [OTf]⁻ binds to 1-OTf and the other where a further qui molecule coordinates to [1·(qui)]⁺; of these two ternary adducts, [1(OTf)₂]⁻ is stabilized notably with respect to the dissociated form, whereas [1·(qui)₂]⁺ is predicted to be an unstable species (see Scheme 3). The optimized structures of possible binary and ternary adducts are depicted in Figure 4, and the computed bond distances suggest considerably weakened Sn–ligand bonds in the ternary species. The notable difference between the computationally and X-ray crystallographically determined Sn–N/O bond lengths in the (qui)·1-OTf complex is likely due to the intermolecular forces in the crystal structure, which appreciably influences the character of the Sn–N/O dative bonds. Of note, one of the Sn–N bonds in [1·(qui)₂]⁺ is particularly elongated, which results in asymmetry across the equatorial plane and is likely due to steric effects.²⁶ Having predicted that 1-OTf, (qui)·1-OTf, and [1·(qui)]⁺ are thermally accessible in a solution at RT, we sought to identify each species experimentally by manipulating the equilibria between them. Upon portion-wise addition of qui to 1-OTf in DFB, the ¹H NMR iPr methine resonance (δ = 2.15 ppm) moves upfield to a limiting value of δ = 1.94 ppm (5.5 equiv qui), after which further qui results in no further change. Similarly, the ¹¹⁹Sn{¹H} NMR resonance moves upfield from that observed for pure 1-OTf (δ = 166 → 0 ppm; see the Supporting Information). Only a single resonance attributable to the [1]⁺ fragment was observed for any qui:1-OTf ratio in the ¹¹⁹Sn{¹H} NMR spectra (the same is true for the iPr resonances in the ¹H NMR spectra), indicating that the observed shifts are weighted averages of the multiple species, all of which are in rapid exchange.

Figure 4.

Optimized structures of possible binary and ternary adducts in the solution phase for the 1-OTf + qui Lewis pair. Selected bond distances are given in Å. All hydrogen atoms are omitted for clarity.

A variable temperature ¹¹⁹Sn{¹H} NMR experiment of a 1:1 mixture of 1-OTf and qui (CD₂Cl₂)²⁷ was undertaken to slow the exchange between the adducts (Figure 5, spectra A1 and A2),²⁸ which revealed the broad RT resonance (δ = 45 ppm) to split into three distinct peaks (i–iii) below 263 K. Resonance i (δ = 119 ppm, 233 K) is very similar to that seen for authentic [1·(qui)]⁺[Al(OR^F)₄]⁻ (δ = 113 ppm, 293 K; Figure 5, spectrum B), and accordingly, it is assigned to the cation fragment therein. The existence of the latter implies that dissociation of [OTf]⁻ from (qui)·1-OTf has occurred, which is computationally predicted to bind spontaneously to 1-OTf and form [1·(OTf)₂]⁻ (ΔG = −3.2 kcal/mol).²⁹ To verify this, an equimolar solution of [nBu₄N][OTf] and 1-OTf (DFB, 293 K) was prepared, which displayed a broad peak at δ = 62.5 ppm in the ¹¹⁹Sn{¹H} spectrum (Figure 5, spectrum C); this is significantly upfield from the chemical shift of either 1-OTf or [1·(qui)]⁺ and is almost identical to that observed for ii (62.9 ppm, 233 K), so it is attributed to [1(OTf)₂]⁻. Notably at temperatures below 203 K, only one peak at δ = 14 ppm is observed (Figure 5, spectrum A3); DFT calculations predict the 1:1 adduct between qui and 1-OTf to be the most stable species at low temperatures (for details, see the Supporting Information, Section 6.1), and accordingly, we ascribe resonance iii to (qui)·1-OTf.

Figure 5.

(A1–A3): Stacked ¹¹⁹Sn{¹H} NMR spectra of a 1:1 mixture of 1-OTf + qui in CD₂Cl₂ at selected temperatures. (B) ¹¹⁹Sn{¹H} NMR spectrum of authentic [1·(qui)][Al(OR^F)₄] in DFB. (C) ¹¹⁹Sn{¹H} NMR spectrum of 1-OTf + [nBu₄N][OTf] in DFB.

Mechanism for the Activation of H₂

Admission of 4 bar H₂ to [1·(qui)][Al(OR^F)₄] in DFB led to no reaction by ¹H or ¹¹⁹Sn{¹H} NMR spectroscopy at RT after several days, nor after heating at 100 °C for 10 h. Furthermore, this salt did not scramble HD to H₂, D₂, and HD, which would detect rapidly reversible heterolysis. These results exclude a simple mechanism by which the Sn–N bond can elongate (with or without complete cleavage) to activate H₂ in a manner akin to that reported for the related classical adduct [iPr₃Si–PtBu₃][B(C₆F₅)₄].^11a In support of this conclusion, computations predict the activation energy for H₂ cleavage via this mechanism to be prohibitively high (ΔG^⧧ = 36.4 kcal/mol, see Supporting Information Section 6.4 for the transition-state structure).

Because 1-OTf and [1·(qui)]⁺ are both present at RT in solutions of 1-OTf/qui LP, either could feasibly engage with qui to mediate H₂ cleavage; accordingly, we sought to elucidate whether they could act as LAs for the FLP-mediated H₂ activation. [1·(qui)][Al(OR^F)₄] is a valid model compound to study the reactivity of [1·(qui)]⁺ as a LA fragment, so using [1·(qui)][Al(OR^F)₄] + qui (hereafter referred to as the “[1·(qui)]⁺/qui” LP) allows the investigation of H₂ activation, where [OTf]⁻ cannot be involved at any point in the reaction.

When one equivalent of qui is added to [1·(qui)][Al(OR^F)₄], the evidence of H₂ activation is observed immediately after H₂ admission by the appearance of NMR spectroscopic resonances attributable to 1-H [SnH: δ (¹H) = 5.10 ppm, ¹J(¹H–¹¹⁷Sn) = 1404 Hz, ¹J(¹H–¹¹⁹Sn) = 1470 Hz; δ (¹¹⁹Sn{¹H}) = −47.3 ppm] and [qui-H]⁺ [qui-H⁺: δ (¹H) = 13.84 ppm] (Scheme 4).^17a The ²⁷Al and ¹⁹F NMR spectra remain unchanged, indicating that [Al(OR^F)₄]⁻ remains intact and does not participate in H₂ activation, as expected. Based on these observations, we propose the mechanism shown in Scheme 5, which is supported computationally.³⁰ H₂ is cleaved via a termolecular transition state as illustrated in Figure 6, which requires the addition of free qui, akin to H₂ activation mediated by Ar₃B and N/P-centered FLPs. Here, the qui coordinated in [1·(qui)]⁺ acts as a leaving group, while the external qui is the LB for H₂ activation.

Scheme 4. H₂ Activation by [1·(qui)]⁺ Only Occurs When Free qui LB Is Present; R^F = C(CF₃)₃.

Scheme 5. Proposed Mechanism for H₂ Activation with the [1·(qui)]⁺/qui Lewis Pair via a Ternary TS; [Al(OR^F)₄] Counteranion Omitted; R^F = C(CF₃)₃.

Figure 6.

Transition-state TS_K-qui identified computationally for H₂ activation with [1·(qui)]⁺/qui and the species corresponding to the most stable product state (1-H + [qui-H·(qui)]⁺). Relative stabilities (in kcal/mol) are shown in parentheses, and selected bond distances are given in Å. All hydrogen atoms except those from H₂ are omitted for clarity.

Computations reveal that H₂ splitting and dissociation of the departing qui molecule take place concertedly, yet in an asynchronous manner (after TS_K-qui, the dissociation of qui lags behind the H₂ cleavage). Thus, a weakly bound termolecular intermediate [qui-H]⁺·(1-H)·(qui) is produced in this process (lying at 9.4 kcal/mol in free energy, see Supporting Information Section 6.2 for the ternary intermediate structure); however, the separation of 1-H and the formation of the H-bonded [qui-H·(qui)]⁺ species result in a significant stabilization on the product side (the reaction is exergonic by 4.3 kcal/mol). Computations predict the free energy barrier of this H₂ cleavage reaction to be 19.5 kcal/mol, which is consistent with the observed rate at the applied reaction conditions.

Interestingly, while the qui-mediated hydrogenolysis of [1·(qui)][Al(OR^F)₄] with H₂ is quantitative after 48 h (RT), the analogous reaction with 1-OTf reaches only ca. 50% conversion to 1-H, albeit much more quickly (2 h). Figure 7 shows the time evolution of the respective ¹H NMR spectra for these reactions to illustrate their differing rates. Initially, as H₂ activation by 1-OTf/qui proceeds, the ¹¹⁹Sn{¹H} NMR spectra show the expected resonance at δ = −47 ppm for 1-H growing in intensity and a peak at δ = 40 ppm corresponding to the aforementioned fast equilibrium between 1-OTf, (qui)·1-OTf, [1·(qui)]⁺[OTf]⁻, and [1(OTf)₂]⁻ species; the latter moves upfield and sharpens as the reaction approaches completion, reaching a limiting value of δ = 19 ppm, which we attribute to the generation of [1(OTf)₂]⁻.

Figure 7.

¹H NMR spectra at selected time points (shown in parentheses in h) during the activation of H₂ with 1-OTf/qui and [1·(qui)]⁺/qui LPs at RT. The initial rate of H₂ activation with 1-OTf/qui is much faster than [1·(qui)]⁺/qui; however, the former reaches only ca. 50% conversion to 1-H. Contrastingly, the comparable reaction with [1·(qui)]⁺/qui LP is slower, although it eventually forms 1-H quantitatively. * denotes the resonances associated with 1-H.

This result implies that as H₂ activation proceeds, [qui-H]⁺[OTf]⁻ accumulates, and [OTf]⁻ is concomitantly sequestered by 1-OTf to produce inactive [1(OTf)₂]⁻ (Scheme 3); this association quenches the electrophilicity of 1-OTf because the five-coordinate anionic adduct is not Lewis acidic enough to participate in H₂ heterolysis. Therefore, the attainable conversion of H₂ cleavage is impaired by the increasing concentration of [1(OTf₂)]⁻. In support of this hypothesis, the incremental addition of [nBu₄N]⁺[OTf]⁻ to 1-OTf in DFB reproduces the upfield movement of the ¹¹⁹Sn{¹H} NMR shift for the 1-OTf/qui/H₂ reaction exactly, confirming that the binding of [OTf]⁻ to 1-OTf occurs in situ during H₂ activation (see the Supporting Information). Conversely, during H₂ activation by [1·(qui)]⁺/qui, the noninteracting counteranion in the [qui-H]⁺[Al(OR^F)₄]⁻ product cannot coordinate and quench the [1·(qui)]⁺ LA, so the reaction can proceed to quantitative conversion.

It is notable that for this latter LP system, the concentration of uncoordinated qui should remain constant throughout the reaction with H₂ (Scheme 5); indeed, it is found that the [1·(qui)]⁺/qui LP is actually catalytic with respect to qui, and H₂ activation proceeds with substoichiometric amounts of added qui, albeit at correspondingly reduced rates. For the 1-OTf/qui LP, however, the concentration of free qui is reduced by its association with 1-OTf and decreases further as the reaction proceeds. The concentration of free qui is therefore expected to be lower at all times for H₂ cleavage reactions employing 1-OTf/qui versus [1·(qui)]⁺/qui; this difference makes a direct quantitative kinetic comparison extremely difficult. Nevertheless, the initial H₂ activation rate is clearly higher for the 1-OTf/qui LP in spite of the lower concentration of free qui available, which supports the conclusion that H₂ activation is inherently faster at the beginning of the reaction for 1-OTf/qui than [1·(qui)]⁺/qui. In order to rationalize our observations regarding the rates and yields of H₂ heterolysis induced by these two qui-dependent systems, we additionally examined the basic mechanistic features of the latter reaction, computationally (Figure 8).

Figure 8.

Transition-state TS_OTf-qui identified computationally for H₂ activation with 1-OTf/qui and the species corresponding to the most stable product state (1-H + [qui-H]⁺[OTf]⁻). Relative stabilities (in kcal/mol) are shown in parentheses, and selected bond distances are given in Å.³³

Consistent with our experiment results, the transition state identified for the H₂ splitting process (TS_OTf-qui) possesses a significantly lower barrier with respect to the 1-OTf + qui + H₂ state³¹ (13.5 kcal/mol) when compared to that obtained for the [1·(qui)]⁺ + qui + H₂ reaction (19.5 kcal/mol). When the ground state of the reactants in the former case is factored in, corresponding to [1·(qui)]⁺ + [OTf]⁻ + H₂, the overall kinetic barrier is actually 18.3 kcal/mol (see Scheme 3 and Figure 8). These computed barriers are in agreement with the measured reaction times.³² Our computational analysis suggests that the differing energies of TS_OTf-qui and TS_K-qui relative to their corresponding reactant states could be associated with solvent effects (for a more detailed analysis of solvation effects, see Supporting Information Section 6.6). The reactant state of

involves only neutral components, whereas the H₂ heterolysis yields two ionic species. The charge separation is notable already in TS_OTf-qui, which has a relatively late character, so it is stabilized more by solvent polarity than the reactant state. On the other hand, no charge variation occurs in the reaction

Interestingly, on the product side of the reaction with the 1-OTf/qui Lewis pair, the 1-H + [qui-H]⁺[OTf]⁻ state is predicted to be the most stable form lying at −4.9 kcal/mol in free energy, which is close to that of the reactant state (−4.8 kcal/mol) and suggests that, even in the absence of subsequent capture of [OTf]⁻ by 1-OTf to form [1(OTf)₂]⁻ (predicted to be exergonic by 3.2 kcal/mol; vide supra), the reaction would in fact still only reach approximately 50% conversion.

Employing Collidine as the LB Partner

To provide further context, we also studied the analogous system using col as a LB partner for [1]⁺-based LAs because it was found to be the optimally performing base surveyed for catalytic hydrogenations using 1-OTf and also to resolve any reactivity effects because of its different steric profile versus qui.^17a,17b All attempts to isolate samples (col)·1-OTf or [1·(col)]⁺[OTf]⁻, from mixtures of the LPs by analogy with qui experiments, were unsuccessful. This indicates that the association between 1-OTf and col is weaker compared with qui, which is consistent given that col exhibits a lower basicity [pK_a(MeCN) col-H⁺ = 14.98]³⁴ and that its N center is hindered by the flanking 2,6-Me substituents. The datively bound adduct (col)·1-OTf, analogous to (qui)·1-OTf, could be identified computationally; however, an appreciable Sn–N bonding interaction is sterically prevented with only a weak col···1-OTf association evident from DFT calculations (see Supporting Information Section 6.7 for a ternary complex structure). A mixture of 1-OTf and col (1:1, DFB) displays a broad signal in the RT ¹¹⁹Sn{¹H} NMR spectrum at δ = 142 ppm, which is significantly downfield from that observed for 1-OTf/qui (δ = 34 ppm) and moves upon warming to a limiting value at 353 K that is identical to authentic 1-OTf. When the solution is cooled, the peak moves upfield (δ = 116 ppm at 243 K), indicating that col is binding to 1-OTf more strongly.³⁵ Our computations suggest that the small change in chemical shift (24 ppm) can be ascribed to the [1·(col)]⁺ complex formed in low concentrations. Although several different species might be contributing to the observed average chemical shift, these could not be resolved; nonetheless, it is clear that the association between 1-OTf and col is significantly less favored than for qui.

Fortunately, [1·(col)]⁺[Al(OR^F)₄]⁻ can be prepared by an analogous route to [1·(qui)]⁺[Al(OR^F)₄]⁻ (Scheme 2), as a yellow oil that unfortunately resisted attempts at crystallization. Its ¹¹⁹Sn{¹H} NMR spectrum exhibits a broad resonance at δ = 128 ppm, which is slightly downfield from the qui counterpart (δ = 113 ppm). Similar to [1·(qui)]⁺, [1·(col)]⁺[Al(OR^F)₄]⁻ does not react with H₂ unless additional col is added, resulting in the appearance of resonances attributable to 1-H + [col-H]⁺ by ¹H and ¹¹⁹Sn{¹H} NMR spectroscopy. This strongly suggests that [1·(qui/col)]⁺ LAs activate H₂ with their corresponding LB partners via the same mechanism. Notably, however, an elevated pressure (10 bar H₂) is required for H₂ activation to proceed effectively by [1·(col)]⁺/col, and the rate to reach quantitative conversion is considerably slower (28 days) under analogous equimolar conditions. These latter results are consistent with previous observations^17a,17b and are commensurate with the lower basicity of col relative to qui. For H₂ activation with the 1-OTf/col LP, once again only partial conversion to 1-H and [col-H]⁺ is observed (ca. 50% in 4 days at RT) by ¹H NMR spectroscopy because of the formation of [1(OTf)₂]⁻; however, the initial reaction rate is faster when 1-OTf is the LA compared with [1·(col)]⁺.

To shed light on the similarities and differences between the qui and col systems thus far, the energetics of H₂ activation for the [1·(col)]⁺/col and 1-OTf/col LPs were assessed computationally; the key structures and their relative stabilities are shown in Figures 9 and 10. The principal mechanistic features of these reactions are quite similar to those of the analogous reactions with qui as the LB (concerted asynchronous H₂ splitting and base dissociation processes, formation of H-bonded species with the protonated base), yet there are profound differences in the computed energetics. For instance, H₂ activation with [1·(col)]⁺/col is predicted to be more favored thermodynamically as compared to [1·(qui)]⁺/qui, which appears counterintuitive in light of the relative basicities; the computed proton affinities (kcal/mol) are −160.9 and −155.4 for qui and col, respectively.

Figure 9.

Transition-state TS_K-col identified computationally for H₂ activation with [1·(col)]⁺/col and the species corresponding to the most stable product state (1-H + [col-H·(col)]⁺). Relative stabilities (in kcal/mol) are shown in parentheses, and selected bond distances are given in Å.

Figure 10.

Transition-state TS_OTf-col identified computationally for H₂ activation with 1-OTf/col and the species corresponding to the most stable product state (1-H + [col-H]⁺[OTf]⁻). Relative stabilities (in kcal/mol) are shown in parentheses, and selected bond distances are given in Å.

However, the strained nature of the [1·(col)]⁺ binary complex implies substantial reactant-state destabilization with a dissociation free energy into [1]⁺ + col of only 17.6 kcal/mol, so the reaction becomes more exergonic versus qui (for a detailed energy decomposition analysis, see Supporting Information Section 6.8). Destabilizing steric effects are even more enhanced in the transition state for H₂ splitting (TS_K-col in Figure 9), which in combination with the reduced basicity of col results in a higher barrier compared to that obtained for [1·(qui)]⁺/qui + H₂ (25.7 vs 19.5 kcal/mol, respectively).

For the 1-OTf/col system, the computational results predict the 1-OTf + col state to be clearly favored over the ionic [1·(col)]⁺[OTf]⁻ alternative (Figure 10), and the H₂ activation reaction is found to be less favored (both kinetically and thermodynamically) than the analogous reaction with 1-OTf/qui. The latter trend for the predicted energetics of these two reactions follows the difference in the proton affinity of the two bases, which is also in line with our experimental observations. Importantly, H₂ activation by 1-OTf/col is predicted to have a much lower activation energy (20.2 kcal/mol) than the [1·(col)]⁺/col LP (25.7 kcal/mol), which is consistent with the observed reaction rates. We attribute this to the stabilization of TS_OTf-col by the polar solvent relative to the neutral ground state (1-OTf + col + H₂), and the destabilizing steric effects in TS_K-col, which are weaker in TS_OTf-col.

Relevance to Catalytic Hydrogenations Using 1-OTf/col

Our previous observations in the catalytic hydrogenation of imine and carbonyl compounds by 1-OTf + col (our optimum LB for catalysis) showed that the reduction steps involved the reaction of 1-H with protonated imines and (carbonyl)·1-OTf species (i.e., activated substrates), respectively. These were much more rapid than H₂ activation, and notably, 1-H was observed neither by ¹H nor by ¹¹⁹Sn{¹H} NMR spectroscopy at any point during catalysis; hence, it was proposed that the H₂ cleavage was the rate-determining step.^17a,17b The current results show that [1(OTf)₂]⁻ forms as H₂ heterolysis by 1-OTf/col proceeds and because of the sequestration of active LA 1-OTf, it would therefore be expected to affect the reaction rate deleteriously in catalytic reactions. It should be noted, however, that in catalytic reactions, the direct H₂ cleavage products 1-H and [col-H]⁺[OTf]⁻ are almost immediately consumed during proton and hydride transfers to the substrate; because H₂ cleavage is the rate-determining step, the accumulation of [1(OTf)₂]⁻ will be suppressed and/or it will form reversibly (Scheme 6). The latter postulate is reinforced by the calculated small free energy difference between the formation of 1-H + [col-H][OTf] from 1-OTf + col + H₂ (+1.7 kcal/mol) and binding of 1-OTf with [OTf]⁻ (−3.2 kcal/mol); this implies that all of these species would coexist in a readily perturbed equilibrium, as 1-H + [col-H][OTf] is consumed in substrate hydrogenation. Indeed, [1(OTf)₂]⁻ is not observed by ¹¹⁹Sn{¹H} NMR spectroscopy during catalytic reactions,^17a,17b verifying that this species plays an insignificant role under the conditions of hydrogenation catalysis. Taking the results together, we propose that lowering the activation energy of H₂ cleavage is the most important factor to improve the rate of hydrogenation by triorganostannylium-based catalysts. Because [OTf]⁻ explicitly interacts with [1]⁺ in the reactants and the lowest energy TS possible for H₂ heterolysis by 1-OTf/col, this anion affects their stabilities; variation of this anion should consequently affect the activation energy as long as it remains associated in such a manner.

Scheme 6. Simplified Catalytic Cycle for Hydrogenations Mediated by 1-OTf/col. The Reduction Step Is Rapid Relative to the H₂ Cleavage, Making 1-OTf + Col the Resting State under Catalytic Conditions; Because [OTf]⁻ Does Not Accumulate in Situ, 1-OTf Is Not Sequestered (as [1(OTf)₂]⁻) during the Catalytic Hydrogenations; X = NR″, O.

Conclusions

In conclusion, we have elucidated that 1-OTf is the LA involved in the lowest energy TS for H₂ activation with both coordinating (qui) and noncoordinating (col) LBs. 1-OTf and qui associate to form the dative adduct (qui)·1-OTf, which was characterized by single-crystal X-ray crystallography; solution-phase NMR studies and DFT calculations reveal that this adduct is in equilibrium with 1-OTf, [1·(qui)]⁺ and [1(OTf)₂]⁻ at RT. By contrast, 1-OTf and col show a far lesser degree of association because of the greater steric protection of the LB. Ionic salts [1·(LB)]⁺[Al(OR^F)₄]⁻ permitted the examination of the reactivity of an alternative LA adduct in the absence of a coordinating anion, where it was found that an additional extraneous LB was necessary to promote H₂ activation. This indicates that these reactions proceed via a (LB)·H₂·[1·(LB)]⁺ TS, which is supported by computational results that also predict analogous mechanisms for H₂ heterolysis by 1-OTf/LB pairs. Although the [1·(LB)]⁺/LB systems do activate H₂, their initial reaction rates are significantly slower than the corresponding 1-OTf/LB pairs, which is attributed to differing degrees of charge separation along the respective reaction pathways; in the latter case, this is more pronounced, resulting in the relative stabilization of the TS by highly polar solvent media (e.g., DFB), and hence a lower barrier to the H₂ cleavage. The activation energies for H₂ cleavage for both 1-OTf/LB and [1·(LB)]⁺/LB pairs are found to be significantly higher when LB = col, which is due to its lower basicity and the higher steric strain in the TS, versus qui. A surprising finding is that while [1·(LB)]⁺/LB LPs activate H₂ quantitatively, the catalytically relevant 1-OTf/LB pairs achieve only ∼50% conversion. This is rationalized by the liberation of [OTf]⁻ during H₂ activation, which is subsequently sequestered by 1-OTf to form [1(OTf)₂]⁻, as well as by the modest thermodynamic driving force of H₂ activation.

Taken together, these results reveal that the triflate ion in 1-OTf/LB LPs exerts a considerable effect on both the thermodynamics (the relative stabilities of ground-state species prior to H₂ activation and the products thereafter) and kinetics (the TS energy barriers) of H₂ activation. Our present work highlights the importance of how the counteranion can impact the reactivity of cationic LAs beyond a simple modulation of Lewis acidity, and we anticipate that these findings will translate to the reactivity of other systems. Encouragingly, the ability to vary the anion partner of LA cations offers a design principle unavailable to neutral LAs. We expect that this will encourage further research into anion tunability for the tailoring of LA reactivity, especially toward the reactions of important small molecules.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acscatal.0c02023.

• Full synthetic and computational details, NMR spectra of the reported reactions, and crystallographic data are contained therein (PDF)

• X-ray structure for (qui)·1-OSO₂CF₃ (CIF)

• X-ray structure for [1·(qui)]⁺{Al([OC(CF₃)₃]₄}⁻ (CIF)

Author Contributions

The manuscript was written through contributions of all authors.

We wish to thank Imperial College for a President’s Ph.D. Scholarship award (R.C.T.-R.); the EPSRC for PhD (J.S.S. and A.D.P.) and grant funding (EP/N026004), including a Doctoral Prize Fellowship (EP/N509486/1, D.J.S.); and the Royal Society for a University Research Fellowship (A.E.A.; UF/160395). This work was also supported by the NKFIH (K-115660). Computational resources provided by the NIIF Supercomputer Center are acknowledged.

The authors declare no competing financial interest.