Control of dynamic sp³-C stereochemistry

Abstract

Stereogenic sp³-hybridized carbon centres are fundamental building blocks of chiral molecules. Unlike dynamic stereogenic motifs, such as sp³-nitrogen centres or atropisomeric biaryls, sp³-carbon centres are usually fixed, requiring intermolecular reactions to undergo configurational changes. Here we report the internal enantiomerization of fluxional carbon cages and the consequences of their adaptive configurations for the transmission of stereochemical information. The sp³-carbon stereochemistry of the rigid tricyclic cages is inverted through strain-assisted Cope rearrangements, emulating the low-barrier configurational dynamics typical for sp³-nitrogen inversion or conformational isomerism. This dynamic enantiomerization can be stopped, restarted or slowed by external reagents, while the configuration of the cage is controlled by neighbouring, fixed stereogenic centres. As part of a phosphoramidite–olefin ligand, the fluxional cage acts as a conduit to transmit stereochemical information from the ligand while also transferring its dynamic properties to chiral-at-metal coordination environments, influencing catalysis, ion pairing and ligand exchange energetics.

Main

The hugely varied three-dimensional (3D) structures—and therefore the hugely varied properties—of many organic molecules emerge from combining just a few types of atomic building blocks. For example, 19 of the 22 proteinogenic amino acids are formed solely from sp²- or sp³-hybridized carbon, nitrogen and oxygen atoms, capped by hydrogen substituents. Of this small array of elemental building blocks, it is tetrahedral sp³-carbon^1,2,3,4 and sp³-nitrogen^5,6,7,8,9,10 atoms that have the potential to form stereogenic centres, creating chiral structures.

Chirality also arises in organic molecules by virtue of motifs other than stereogenic atoms. However, although stereochemical inversion of some planar chiral motifs^11,12,13, helices^14,15,16 and stereogenic sp³-nitrogen centres^7,8,9,10 can occur rapidly and reversibly through low-barrier conformational isomerism, sp³-carbon centres cannot generally undergo spontaneous stereochemical changes. For example, the energy barrier to pyramidal inversion of methane is greater than its C–H bond dissociation energy^17,18,19. Accordingly, unlike other stereogenic motifs^{11,12,13,14,15,16}, sp³-carbon centres cannot generally adapt to surrounding chiral moieties and cannot be controllably switched by the application of external stimuli.

Instead, intermolecular reactions are usually necessary^20,21 to invert individual stereogenic carbon centres, proceeding through mechanisms involving high-energy bond-breaking and bond-making steps²² with pentavalent transition states²⁰ (for example, S_N2 reactions) or trigonal intermediates²¹, such as carbocations, carbanions or radicals. Of course, it is this stability of sp³-carbon’s tetrahedral geometry that makes it essential to the chiral skeletal diversity of organic compounds. It allows for predictable synthesis of configurationally stable molecules. Yet, the stability also limits the extent to which the complex 3D connectivity of aliphatic structures can exhibit dynamic, adaptive stereochemistry²³.

There have been impressive, but rare, examples of small covalent systems^{24,25,26,27,28} capable of sp³-carbon enantiomerization by low-barrier intramolecular processes. However, they do so without external control of their rate or direction to a single stereoisomer. Only multicomponent interlocked molecules, in which a ring shuttles along a prochiral axle^29,30, have been amenable to external control. There have been no compact and controllable dynamic sp³-carbon building blocks. Therefore, it has not been possible to investigate the transmission of stereochemical information through such systems^7,9,10.

In this Article we report a series of chiral fluxional carbon cages^26,27,28,29 that exhibit responsive sp³-carbon-centred stereochemistry, adapting to and transmitting surrounding stereochemical information. By applying density functional theory (DFT) calculations and solution- and solid-state NMR spectroscopy, in combination with X-ray crystallography, we establish the extent to which their dynamic Cope rearrangements^31,32 are controlled by neighbouring, fixed stereogenic centres. We have found that a substantial energetic bias of more than 10 kJ mol⁻¹ can be exercised over the stereochemical equilibria by a single fixed stereocentre. The rearrangements proceed rapidly at rates more commonly associated with low-barrier conformational changes of aliphatic systems (for example, a cyclohexane ring-flip energy barrier of ~43 kJ mol⁻¹) rather than a configurational change. We show that these rapid constitutional dynamics can be halted by covalent modification of the cage through a [2 + 2 + 2] cycloaddition reaction, then subsequently restarted after a cycloreversion. The rearrangement rate is also attenuated upon coordination of the fluxional cage to Pd(II) or Ru(II) as part of a phosphoramidite–olefin ligand. By its inclusion in the simple ligand design, the fluxional cage transmits stereochemical information to the metal ion—either through the covalent ligand backbone or by ion pairing with a chiral counterion. This property is exploited in enantioselective catalysis of an allylic substitution reaction, as well as in creating chiral-at-metal stereogenic centres that adopt the configurational dynamics of the cage.

Results

The Cope rearrangement of barbaralane (BB) is an example (Fig. 1a) of a narcissistic^25,33 automerization—it gives rise to a degenerate structure through a transition state (TS-BB) bearing an internal mirror plane (${\upsigma}_{{{\mathrm{v}}}}^{\prime}$) that is not present in the minimum energy structure. We noted that by desymmetrizing BB (Fig. 1b) using either a 9-BB, 3-BB or 2,4-BB substitution pattern, the mirror plane present at the energy minimum (${\upsigma}_{{{\mathrm{v}}}}^{\prime\prime}$) is lost, while the mirror plane formed in the transition state (${\upsigma}_{{{\mathrm{v}}}}^{\prime}$) is retained. As a result, the Cope rearrangement inverts simultaneously some, or all, of the four or five stereogenic centres present in the structure. Given that the rearrangement of BB is known to proceed with a remarkably low free energy of activation, ΔG^‡, of 32.3 kJ mol⁻¹ (Supplementary Table 7)^{34,35,36,37,38}, chiral 9-BB, 3-BB or 2,4-BB derivatives should undergo rapid enantiomerization.

Fig. 1: Multiple dynamic sp³-carbon centres.

a,b, Fluxional sp³-carbon stereochemistry arises in BBs when the structures interchanged by their Cope rearrangements (a) are desymmetrized with any of the three substitution patterns shown in b. Cahn–Ingold–Prelog priorities are chosen to be R¹ > C > R² for the assignment of absolute configuration. When assigning a descriptor to position 9 of 9-BB, the cyclopropyl bridgehead C1 has precedence over the divinyl bridgehead C5 (Supplementary Fig. 1). 3-BB and 2,4-BB each have four chirotopic (R/S) centres, whereas the 9-BB pattern gives rise to five stereogenic centres, of which three are chirotopic and two are achirotopic (r/s).

Diastereomeric adaptation

We targeted 9-BB 1 (Fig. 2) as a convenient example of the 9-BB substitution pattern that bears a hydroxyl group for synthetic elaboration. The Cope rearrangement involving positions 2–8 of 1 (Fig. 2a) causes enantiomerization of the whole cage and formally inverts the stereochemistry of position 9 by effectively ‘swapping’ the cyclopropyl and alkene substituents connected to the stereocentre. Compound 1 was synthesized (Supplementary Scheme 1) by a three-step route from ethynyl magnesium bromide and tropylium tetrafluoroborate, using a gold-catalysed enyne cycloisomerization^39,40 to form the BB backbone. When labelling 1 and subsequent compounds, a single stereochemical descriptor is included to indicate the configuration at position 9 of the BB (Fig. 2a), for example, (R)-1 and (S)-1, omitting the additional stereochemical labels of positions 1, 2, 5 and 8 for simplicity (Fig. 1). Treatment of 1 with Mosher’s acid chloride (Fig. 2a) produces a set of Mosher’s esters 2 in which the configurationally fixed stereocentre is introduced at a distance of three covalent bonds from the dynamic BB unit. An additional descriptor for the configuration of the Mosher’s ester group is included in the labels for 2. Derivatization with (S)-Mosher’s acid gives a dynamic mixture of two diastereomers, (R,S)-2 and (S,S)-2, whereas (R)-Mosher’s acid gives (Fig. 2a) the antipodal mixture, (S,R)-2 and (R,R)-2. Solutions of the two antipodal dynamic mixtures give opposite circular dichroism spectra (Fig. 3a), as would be expected.

Fig. 2: Diastereomeric adaptation and manipulation of chiral BBs.

a,b, The dynamic sp³-C stereochemical equilibrium of the BB cage is degenerate in 1 but becomes biased towards one stereoisomer upon attaching a chiral auxiliary (a) or by dimerization through a spirocyclic bridge (b). c, The position of the stereochemical equilibrium changes (and inverts) upon modifying the structure of a chiral auxiliary, remote from the BB unit. d, Further control of the sp³-C stereochemistry is exerted by a cycloaddition reaction, which freezes and symmetrizes the structure, before subsequent cycloreversion re-establishes the dynamic stereochemical equilibrium. Reagents and conditions: (i) 1. (S)-MTPA, (COCl)₂, hexanes, DMF, room temperature (r.t.) to −20 °C, 16 h. 2. 1, DMAP, Et₃N, CHCl₃, r.t., 5 d, 58%. (ii) 1. (R)-MTPA, (COCl)₂, hexanes, DMF, r.t. to −20 °C, 16 h. 2. 1, DMAP, Et₃N, CHCl₃, r.t., 3 d, 79%. (iii) 3, Lawesson’s reagent, PhMe, 110 °C, 18 h, 13%. (iv) 1. 3, (S)-1-phenylethylamine, AcOH, MeOH, r.t., 30 min. 2. NaBH₃CN, 100 °C, 16 h, 89%. (v) 5, PCl₃, Et₃N, CH₂Cl₂, 0 °C, 3 h. 2. 2,2′-methylenediphenol, CH₂Cl₂, 0 °C to r.t., 16 h, 44%. (vi) 6, PTAD, CH₂Cl₂, 50 °C, 24 h, 85%. (vii) 7, NaOH, ⁱPrOH, 85 °C, 24 h, taken on crude. (viii) CuCl₂, HCl_(aq), 0 °C, 4 h, 48% from 7. X-ray structures are shown in stick representation. Compound 4 crystallizes in a centrosymmetric space group, that is, (S,S)-4 and (R,R)-4 are both present, but only (R,R)-4 is shown for clarity. Diffraction data for crystals of (R,S)-5 allow only assignment of relative stereochemistry. MTPA, α-methoxy-α-trifluoromethylphenylacetic acid; DMF, N,N-dimethylformamide; DMAP, 4-(dimethylamino)pyridine; PTAD, phenyl-1,2,4-triazoline-3,5-dione.

Fig. 3: Spectroscopic evidence of sp³-carbon adaptation to covalently tethered chiral auxiliaries.

a, Normalized circular dichroism spectra of 2 (115 μM in MeCN) and 5 (210 μM in MeCN) confirm that antipodal equilibrium mixtures give equal and opposite absorbances. b, A comparison of partial solution ¹H NMR spectra (CDCl₃, 298 K) shows the reduced symmetry of the chiral 9-BB motif: top, 3 (700 MHz); middle, (R)/(S)-1 (600 MHz); bottom, (R,S)/(S,S)-2 (600 MHz). Resonances are labelled according to the numbering in Fig. 2. c, Comparison of partial ¹³C{¹H} NMR spectra: top, solid-state chemical shifts calculated from the X-ray crystal structure of (R,R)-2 in CASTEP v17.2⁶⁵ using the Perdew–Burke–Ernzerhof functional⁶⁶ and on-the-fly generated pseudopotentials; middle, (R,R)-2 as a powder at ambient temperature (105 MHz); bottom, (S,S)-2 as a solution in 5:1 CS₂–CD₂Cl₂ at low temperature (125 MHz, 159 K). The asterisk indicates the resonance of residual acetone. d, The Boltzmann distribution of isomers shifts towards a single stereoisomer at low temperatures; for example, a free energy difference of ~5 kJ mol⁻¹ would give an ~90:10 equilibrium mixture at room temperature, but >98:2 at 159 K, so NMR data would be expected to show a single, major species, as is apparent when comparing the three spectra in c.

DFT modelling using the ωB97X-D functional⁴¹, 6-311++G(d,p) basis set^42,43 and a CS₂ polarizable continuum solvent model using the integral equation formalism variant⁴⁴ was employed to compare (Supplementary Table 7) the stereoisomerization energetics of BB, 1 and 2. Using these parameters, the automerization of BB is predicted to proceed with a calculated activation free energy, ${{{\mathrm{{\Delta}}}}}{G}_{\rm{calc}}^{\ddagger}$, of 38.5 kJ mol⁻¹, which is ~6 kJ mol⁻¹ higher than the experimentally measured³⁵ activation free energy, ${{{\mathrm{{\Delta}}}}}{G}_{\rm{exp}}^{\ddagger}$, of 32.3 kJ mol⁻¹, in keeping with previous DFT investigations^36,37. DFT methods systematically overestimate the energy barrier to Cope rearrangement of BBs, but nevertheless allow useful comparisons of trends in activation energies and are known to predict accurately the relative energy minima of isomers^36,37. The computationally predicted ${{{\mathrm{{\Delta}}}}}{G}_{\rm{calc}}^{\ddagger}$ values for 1 (38.0 kJ mol⁻¹) and 2 (35.5 kJ mol⁻¹) are very similar to BB, indicating that the hydroxyl or ester group substitutions at position 9 do not appreciably change the rapid kinetics.

The absence of the ${\upsigma}_{{{\mathrm{v}}}}^{\prime\prime}$ mirror plane in 1 is evident (Fig. 3b) in its solution-state ¹H NMR spectrum—H3 and H7 are magnetically inequivalent, for example. However, the rapid enantiomerization induces a ${\upsigma}_{{{\mathrm{v}}}}^{\prime}$ mirror plane to the time-averaged structure of 1, so only six distinct methine resonances are observed overall. The additional, fixed stereocentre of 2 breaks this ${\upsigma}_{{{\mathrm{v}}}}^{\prime}$ symmetry. Consequently, nine distinct signals corresponding to the BB methine groups are observed (Fig. 3b).

An energy difference, ΔG_calc, of 4.5 kJ mol⁻¹ is computed (Supplementary Table 7) for the rearrangement of 2. The influence of the (S)-Mosher’s ester group moulds the configuration of the cage unit, which preferentially adopts its S form, biasing the equilibrium towards (S,S)-2. Consistent with this prediction, a single crystal (Fig. 2a) obtained from the dynamic (S)-Mosher’s ester mixture was found to contain (S,S)-2 as a frozen³⁴, single stereoisomer. An equal and opposite outcome is observed from the (R)-Mosher’s ester mixture, giving the enantiomeric (R,R)-2 solid-state structure.

To establish the nature of the dynamic solution-state mixtures, we compared (Fig. 3c) the solid-state ¹³C{¹H} NMR spectrum of enantiopure (R,R)-2 crystals to a spectrum obtained using a sample of the (S,S)-2 crystals dissolved in 5:1 CS₂–CD₂Cl₂, generating a dynamic mixture of (R,S)-2 and (S,S)-2. Cooling the solution to 159 K causes the BB ¹³C{¹H} NMR resonances to enter the slow exchange regime. As 159 K is only ~20 K below the observed coalescence temperature (Supplementary Fig. 66) for this low-barrier process, some resonances exhibit exchange broadening. At this low temperature, the decrease of available thermal energy causes the Boltzmann distribution to shift (Fig. 3d) further towards the lowest-energy isomer³⁴. The solid-state chemical shifts of the BB sp³-carbons 1, 2, 5, 8 and 9 are assigned (Fig. 3b) by comparison to the calculated chemical shifts of (R,R)-2. The resonances of the solution sample match up well with those of (R,R)-2 in the solid state, allowing us to assign the resolved solution-state diastereomer as (S,S)-2. The solution-state analysis is thus consistent with the diastereomeric preference predicted by DFT and observed in the solid state. The 9-BB cage undergoes dynamic diastereomeric adaptation under the influence of the configurationally fixed Mosher’s ester group.

The dynamic stereochemical equilibrium can also be biased in the absence of a fixed stereogenic element. Dimerization of two 9-BB-type cages through a spirocylic linkage breaks the degeneracy of the equilibrium. By treating (Fig. 2b) barbaralone 3 with Lawesson’s reagent, we isolated trithiolane 4, which undergoes dynamic rearrangements between an achiral isomer, meso-4, and a pair of enantiomers, (S,S)-4 and (R,R)-4. A small ΔG_calc of 0.7 kJ mol⁻¹ is predicted (Supplementary Table 7) to favour the pair of enantiomers over the meso form in the solution state. Single crystals grown from a solution of 4 contain a racemic mixture of the two chiral stereoisomers.

Further chemical modification to the substituent at position 9 can substantially influence, and even invert, the cage’s equilibrium distribution. The chiral phosphoramidite–olefin^45,46,47 ligand L_BB1 was synthesized (Fig. 2c) by first subjecting 3 to reductive amination with (S)-1-phenylethylamine to afford a mixture of (S,S)-5 and (R,S)-5. Sequential treatment of the amine with PCl₃ then 2,2′-methylenediphenol affords L_BB1. The 2,2′-methylenediphenol functionality was selected as it lacks fixed stereochemistry but has been shown to adopt dynamically chiral conformations as part of phosphoramidite ligands⁴⁵. Comparing L_BB1 to 5 reveals that the differing size and shape of the substituent at position 9 drives the dynamic stereochemical equilibria of the fluxional cage towards opposite configurations. The solution-phase equilibrium of the secondary amine is weighted (Supplementary Table 7) towards the (R,S)-5 diastereomer by a ΔG_calc of 3.8 kJ mol⁻¹, matching the structure observed by X-ray analysis (Fig. 2c) of a single crystal. By contrast, the (S,S)-L_BB1 diastereomer of the phosphoramidite is favoured with a ΔG_calc of 20.2 kJ mol⁻¹. The large magnitude of ΔG_calc for L_BB1 highlights that the configurational dynamics of the 9-BB motif (Fig. 1) correlate with notable changes in its 3D shape³⁴ and, therefore, its energy. At the same time, the opposing stereochemistry of cages 5 and L_BB1 demonstrates that the malleable sp³-carbon configuration adapts to changes in the nearby steric environment.

Manipulating rates and transfer of stereochemistry

To exert further control over the fluxional enantiomerization, we sought to exploit the reactivity of the BBs’ skipped diene units. The fluxional rearrangements can be stopped entirely by engaging the alkene units in covalent bonding, whereas coordination of the π electrons to a transition-metal ion³⁸ modulates the rearrangement rate instead.

An enantiomerizing mixture of 9-(p-tolyl)barbaralol 6 engages (Fig. 2d) in a [2 + 2 + 2] cycloaddition reaction with phenyl-1,2,4-triazoline-3,5-dione⁴⁸, giving rise to 7. This reaction halts the rearrangement while also symmetrizing the structure by forming a second cyclopropyl group. Subsequently, the fluxional cage can be regenerated (Fig. 2d) in a two-step transformation through diazinane 8 (Supplementary Fig. 62), which undergoes cycloreversion with loss of N₂ upon oxidation with CuCl₂. Alternatively, coordination of Pd(II) (Fig. 4) or Ru(II) (Fig. 5) to L_BB1 causes a reduction in the rate of the Cope rearrangement, as discussed below.

Fig. 4: Transfer of dynamic sp³-carbon stereochemistry in Au(I) and Pd(II) complexes.

a, Bidentate coordination of L_BB1 to PdCl₂ leads to cc-Cope rearrangement in which the Pd ‘walks’ along the side of the BB cage, modulating the rearrangement rate. For comparison, monodentate ligand coordination is observed with AuCl. Reagents and conditions: (i) L_BB1, PdCl₂(NCMe)₂, CDCl₃, r.t., 15 min, 98%. (ii) L_BB1, Me₂S·AuCl, CDCl₃, r.t., 10 min, 93%. X-ray crystal structures are shown in stick representation, with a ball for metal ions. Solvent molecules are omitted for clarity. Two structurally similar conformers of each L_BB1PdCl₂ stereoisomer are present in the unit cell, but only one of each is shown for clarity. b, Partial ¹H NMR (CDCl₃) spectra: top, L_BB1 (599 MHz, 298 K); second row, L_BB1AuCl (599 MHz, 298 K); third row, L_BB1PdCl₂ (499 MHz, 298 K); bottom, L_BB1PdCl₂ (499 MHz, 240 K). Resonances are labelled according to the numbering for L_BB1 in Fig. 2. The spectrum at 240 K shows the two L_BB1PdCl₂ complexes in slow exchange in a ratio of 3:4. c, A partial ¹H-¹H EXSY NMR spectrum (499 MHz, CDCl₃, 240 K, mixing time τ_m = 200 ms) showing exchange peaks (red) between resonances of the minor (H_11′) and major (H₁₁) diastereomers as well as COSY peaks (blue) of geminal proton pairs. d, Free energy diagram for the cc-Cope rearrangement. e, Ball-and-stick representation of the DFT-calculated (ωB97X-D/6-311++G(d,p)/SDD/CS₂)⁶⁷ geometry of L_BB1PdCl₂, showing the BB cage at the transition state, TS-Pd. A truncated structure omitting phosphorus and nitrogen substituents is shown for clarity with selected bond lengths given in ångstroms.

Fig. 5: Transfer of dynamic sp³-carbon stereochemistry in chiral-at-Ru(II) complexes.

a, Four diastereomeric square pyramidal complexes are linked by cc-Cope rearrangements and exchange of an MeCN ligand, which proceeds through two intermediate tetrahedral complexes. A non-coordinated PF₆⁻ counterion is omitted from the structural formula of each complex for clarity. Reagents and conditions: (i) L_BB1, CpRu(NCMe)₃·PF₆, CDCl₃, r.t., 5 min, 69%. b, Comparison of the partial ¹H NMR (CDCl₃, 298 K) spectra: top, L_BB1 (599 MHz); middle, a sample of L_BB1RuCp(NCMe)·PF₆ analysed immediately after dissolving a crystalline sample (400 MHz); bottom, the same sample after allowing it to equilibrate for 4 h (400 MHz), revealing that an initially observed single isomer reaches a 4:1 equilibrium mixture. Resonances are labelled according to the numbering for L_BB1 in Fig. 2. c, (C,R,S)-L_BB1RuCp(NCMe)·PF₆ is identified in the solid-state X-ray crystal structure, which is shown in stick representation with a ball for the Ru(II) ion. Solvent molecules and the PF₆⁻ counterion are omitted for clarity. d, Integration of the ¹H NMR (400 MHz, CDCl₃, 298 K) resonance corresponding to H9′ of (A,S,S)-L_BB1RuCp(NCMe)·PF₆ upon dissolving a crystalline sample of (C,R,S)-L_BB1RuCp(NCMe)·PF₆ reveals a first-order increase in concentration with k_obs = 2.56 × 10⁻³ s⁻¹. e, A potential energy surface for isomerization for the cc-Cope processes. Values of ${{{\mathrm{{\Delta}}}}}{G}_{\rm{calc}}$ and ${{{\mathrm{{\Delta}}}}}{G}_{\rm{calc}}^{\ddagger}$ (ωB97X-D/6-311++G(d,p)/SDD/CS₂)⁶⁷ are given except where ¶ indicates experimentally measured (b,d) equilibrium (ΔG_exp) and ligand exchange (${{{\mathrm{{\Delta}}}}}{G}_{\rm{exp}}^{\ddagger}$) energies (in units of kJ mol⁻¹).

L_BB1 and PdCl₂ form (Fig. 4a) a chiral-at-metal^49,50 complex, L_BB1PdCl₂, linking the sp³-carbon configurational inversion to the A/C isomerism⁵¹ of the distorted trigonal bipyramidal (TBPY-5-12) coordination environment (Supplementary Fig. 2). Both possible stereoisomers, arising from coordination of (S,S)-L_BB1 or (R,S)-L_BB1 through their phosphorus centre and an alkene, are observed (Fig. 4a) in the X-ray crystal structure of the L_BB1PdCl₂ complex. The alkene coordination is also evident (Fig. 4b) in the solution state by ¹H NMR spectroscopy. For comparison, a monodentate L_BB1AuCl complex (Fig. 4a) was prepared, which shows only small changes in the ¹H NMR chemical shifts of its alkene signals H3 and H7 relative to the free ligand (Fig. 4b). The room-temperature spectrum of L_BB1PdCl₂, on the other hand, reveals a large change in the chemical shift of H7, consistent with coordination of Pd(II) to the alkene on the same face as the phosphoramidite group.

At 240 K, the ¹H NMR spectrum reveals (Fig. 4b) the two L_BB1PdCl₂ isomers in slow exchange. Two sets of signals are observed in a 3:4 ratio, corresponding to a small free energy difference, ΔG_exp, of 0.5 kJ mol⁻¹ between the two isomers. Consistent with this observation, DFT calculations predict (Supplementary Table 7) a small ΔG_calc of 1.8 kJ mol⁻¹ in favour of (C,R,S)-L_BB1PdCl₂.

Further NMR and DFT analyses elucidate the mechanism by which the L_BB1PdCl₂ complex isomerizes. Depending on the placement and nature of substituents around the Cope substrate, metal coordination can either stabilize a charged, intermediate species as part of a stepwise associative rearrangement mechanism, or it can increase the rate of a concerted rearrangement pathway by stabilizing the transition state⁵². Consequently, Pd(II) salts and other cationic metal ions are known to accelerate Cope rearrangements^53,54. Remarkably, coordination of the Pd(II) to one face of the fluxional cage in L_BB1PdCl₂ has the opposite effect, slowing down the Cope rearrangement. Unlike 2, for example, whose ¹H NMR resonances (499 MHz) enter the slow exchange regime below 160 K (Supplementary Fig. 64), the slower rearrangement of L_BB1PdCl₂ is resolved by ¹H NMR spectroscopy at the higher temperature of 240 K. Using 2D ¹H–¹H exchange spectroscopy (EXSY) at 240 K (Fig. 4c), we measured a rate of exchange, k, of 6.48 s⁻¹, indicating a ${{{\mathrm{{\Delta}}}}}{G}_{\rm{exp}}^{\ddagger}$ of 54.6 kJ mol⁻¹ for L_BB1PdCl₂ (Fig. 4d). The DFT-calculated transition-state structure (Fig. 4e) shows pairs of equidistant C–C bonds, as would be expected for a coordination-coupled Cope (cc-Cope) rearrangement (Fig. 4a) in which the Pd(II) remains bound to the cage through a concerted rearrangement step. The DFT-predicted ${{{\mathrm{{\Delta}}}}}{G}_{\rm{calc}}^{\ddagger}$ of 60.6 kJ mol⁻¹ for this cc-Cope mechanism matches well with the ${{{\mathrm{{\Delta}}}}}{G}_{\rm{exp}}^{\ddagger}$ of 54.6 kJ mol⁻¹.

These data indicate that the metal ion ‘walks’ along one side of the BB cage as the Cope rearrangement proceeds, moving back and forth in sync with the pericyclic reaction⁵². Consequently, the BB not only transmits the stereochemical information from the fixed sp³-carbon stereocentre through its dynamic sp³-carbon framework, biasing the chiral-at-metal configuration, but it also imparts a novel mechanism of intramolecular configurational change at a pentavalent stereocentre, which differs from the established pseudorotation and turnstile mechanisms⁵⁵.

The dynamic sp³-carbon stereochemistry of L_BB1 can also be linked to an intermolecular ligand exchange process. The cyclopentadienyl (Cp) half-sandwich Ru(II) complex⁵⁶ L_BB1RuCp(NCMe)·PF₆ (Fig. 5) has a stereogenic, distorted square pyramidal Ru(II) centre (Supplementary Fig. 91) coordinated to a labile MeCN ligand. While cc-Cope rearrangements interconvert the SPY-5-21 and SPY-5-23 configurational isomers⁵¹ (Fig. 5a), MeCN dissociation forms the distorted tetrahedral (T-4) chiral-at-metal species L_BB1RuCp·PF₆, which mediates A/C stereochemical inversion.

Ru(II) coordination slows the Cope rearrangement sufficiently for a single stereoisomer to be resolved as a metastable species under ambient conditions (Fig. 5b). Upon dissolving single crystals of L_BB1RuCp(NCMe)·PF₆, obtained by slow evaporation, the ¹H NMR spectrum shows the presence of a single complex (Fig. 5b) with resonances distinct from non-coordinated L_BB1. After allowing the sample to fully equilibrate at room temperature for four hours, a new set of peaks is observed (Fig. 5b) at a ratio of 4:1 in favour of the initially observed isomer, equivalent to a ΔG_exp of 4.0 kJ mol⁻¹. X-ray analysis (Fig. 5c) of the crystalline sample reveals the identity of the energetically favoured isomer to be (C,R,S)-L_BB1RuCp(NCMe)·PF₆.

We measured the isomerization rate of (C,R,S)-L_BB1RuCp(NCMe)·PF₆ by monitoring (Fig. 5d) the first-order growth in intensity of the resonance at 3.1 ppm corresponding to the H9′ signal of (A,S,S)-L_BB1RuCp(NCMe)·PF₆—the isomer calculated (Fig. 5e) to be the next most stable stereoisomer. The k_obs of 2.56 × 10⁻³ s⁻¹ at 298 K allows us to determine a ${{{\mathrm{{\Delta}}}}}{G}_{\rm{exp}}^{\ddagger}$ of 87.8 kJ mol⁻¹. Comparison of this value to (1) maxima of the computed potential energy surface (Fig. 5e), (2) a CD₃CN exchange experiment (Supplementary Fig. 70) and (3) literature measurements of MeCN dissociation from Cp half-sandwich Ru(II) complexes⁵⁵ suggests that the cc-Cope and MeCN exchange processes occur at similar rates. To achieve the (C,R,S)-to-(A,S,S) isomerization observed by NMR, the complex must undergo both ligand exchange and cc-Cope steps (Fig. 5e). Overall, the energetic bias towards (C,R,S)-L_BB1RuCp(NCMe)·PF₆ and observation of its stepwise stereomutation to (A,S,S)-L_BB1RuCp(NCMe)·PF₆ illustrate that the fluxional sp³-carbon cage mediates the transfer of stereochemical information with high fidelity from the single, fixed benzylamino stereocentre through its rigid, tricyclic structure.

Dynamic stereocontrol by ion pairing

Having observed transmission of stereochemical information within the covalent frameworks of the L_BB1 complexes, we investigated the influence of chiral counterions⁵⁷ on the degenerate enantiomerization (Fig. 6a) of the cationic L_BB2RuCp(NCMe)⁺ complex, which lacks a fixed stereocentre in its ligand structure. The complex was synthesized (Supplementary Scheme 3) as its hexafluorophosphate salt, L_BB2RuCp(NCMe)·PF₆, in a manner analogous to its permanently chiral homologue L_BB1RuCp(NCMe)·PF₆ (Fig. 5). X-ray analysis of single crystals confirmed (Supplementary Fig. 94) the expected structure of L_BB2RuCp(NCMe)·PF₆ and revealed the presence (Supplementary Fig. 95) of both the (C,R)- and (A,S)-isomers in the crystal unit cell.

Fig. 6: Chiral counterion-directed sp³-carbon and chiral-at-metal stereochemistry applied to enantioselective catalysis.

a, The enantiomerization of L_BB2RuCp(NCMe)⁺ is degenerate in the presence of an achiral counterion. In the presence of chiral anions, the ion-pair interactions break the degeneracy and bias the equilibrium towards one stereoisomer of the BB–metal complex, that is, K ≠ 1. Note: only the two lowest-energy stereoisomers are shown for clarity. Reagents and conditions: (i) L_BB2RuCp(NCMe)·PF₆, BuN₄·Δ-TRISPHAT, CDCl₃, r.t. (ii) L_BB2RuCp(NCMe)·PF₆, Na·(S)-BORBIN, CDCl₃, r.t. b, Comparison of the partial ¹H NMR spectra (400 MHz, CDCl₃, 298 K) of L_BB2RuCp(NCMe)·PF₆ before (top row) and after (other rows) the addition of 1 equiv. of BuN₄·Δ-TRISPHAT. Integration of the MeCN resonances shows that the initially racemic L_BB2RuCp(NCMe)⁺ ion gradually becomes enriched in one stereoisomer (K = 1.2). c,d, The stereoinduction (c) arising from three phosphoramidite ligands was compared in an iridium-catalysed allylic substitution (d). We hypothesize that the chiral anion formed in situ biases the stereochemical equilibrium of the cationic L_BB2Ir π-allyl intermediate, leading to its improved performance relative to the control ligands lacking dynamic sp³-C stereochemistry. Reagents and conditions: (iii) 1. [Ir(1,5-cod)Cl₂] (4 mol%), ligand (16 mol%), THF, 30 min, r.t. 2. (R)-BDHP (10 mol%), r.t., 24 h. Δ-TRISPHAT, Δ-tris(tetrachloro-1,2-benzenediolato)phosphate(V); (S)-BORBIN, bis[(S)-1,1′-bis-2-naphtholato]borate; 1,5-cod, 1,5-cyclooctadiene; (R)-BDHP, (R)-1,1′-binaphthyl-2,2′-diyl hydrogen phosphate.

In the absence of chiral anions, a single ¹H NMR signal is observed for the coordinated MeCN ligand of L_BB2RuCp(NCMe)·PF₆ in CDCl₃ solution (Fig. 6b). However, the addition of one molar equivalent of a chiral shift reagent, Bu₄N·Δ-TRISPHAT⁵⁸ (Fig. 6b) or Na·(S)-BORBIN⁵⁹ (Supplementary Fig. 71), splits the signal in two. Rapid and reversible counterion exchange in the presence of Bu₄N·Δ-TRISPHAT establishes an equilibrium mixture that includes diastereomeric ion pairs, for example, (C,R)-L_BB2RuCp(NCMe)·Δ-TRISPHAT and (A,S)-L_BB2RuCp(NCMe)·Δ-TRISPHAT, which give distinct NMR resonances. Therefore, by tracking the relative intensities of these resonances over time (Fig. 6b and Supplementary Fig. 72) we can monitor changes in sample composition as the chiral anion biases the L_BB2RuCp(NCMe)⁺ complex towards one stereoisomer. Under these conditions, the Δ-TRISPHAT sample evolves from a 1:1 mixture of stereoisomers to a 1:0.81 mixture over several hours, corresponding to an equilibrium constant, K, of 1.2, and ΔG_exp ≈ 0.5 kJ mol⁻¹. The timeframe of the sample’s evolution is consistent with the slow kinetics of isomerization measured (Fig. 5) for L_BB1RuCp(NCMe)·PF₆, suggesting that the stereomutation is again proceeding by cc-Cope and MeCN ligand exchange. The (S)-BORBIN sample reaches (Supplementary Fig. 72) a K of 1.3 over a similar time period. Overall, these experiments establish a means of noncovalent control of the BB stereochemistry. In principle, K could be further increased by removal or omission of any competing achiral anions (such as PF₆⁻) from the reaction mixture and optimization of solvent and concentration.

Based on these results, we hypothesized that this counterion-directed stereochemistry of cationic L_BB2 complexes could be exploited in enantioselective ion-pair catalysis⁶⁰. Unusually, the fixed stereochemistry of the chiral anion would be passed to the catalytically active, fluxional metal complex to transiently generate an enantioenriched ligand framework in situ. To probe this concept, we screened L_BB2 and two control ligands (Fig. 6c) in the enantioselective synthesis of 9 (Fig. 6d) through iridium-catalysed allylic substitution of alcohol 10 by hydroxycoumarin 11⁶¹. We used chiral phosphoric acid (R)-BDHP, which we expected to protonate 10 and induce formation of the iridium-stabilized allylic cation intermediate while simultaneously generating an equivalent of a chiral phosphate anion. The optimized literature conditions⁶¹ for this allylic substitution employ an achiral Lewis acid (Yb(OTf)₃) rather than a Brønsted acid to generate the allylic cation, in conjunction with enantiopure Carreira’s⁴⁶ phosphoramidite–olefin ligand, CL, to impart enantioselectivity. Pleasingly, replacing these reagents with (R)-BDHP and racemic (±)-CL leads to the formation of 9, albeit in just 14% isolated yield. Importantly, however, there is essentially no enantioinduction under the influence of (±)-CL. The product is obtained with a negligible enantiomeric excess (e.e.) of just 2%. It appears that the chiral phosphoric acid alone does little to override the stereochemical preference (or overall lack of it) arising from the racemic ligand. Using achiral phosphoramidite–olefin ligand L_CP leads to a similar outcome. L_CP bears many of the same structural features of L_BB2, but with an achiral cyclopentene unit in place of the dynamically chiral 9-BB substructure. Compound 9 is produced in 49% yield and just 2% e.e. using L_CP. Conversely, our fluxionally chiral ligand, L_BB2, delivers an improved e.e. Using L_BB2, we isolated 9 in 36% yield and 30% e.e. Contrasting this result with the outcome of the reactions using L_CP and (±)-CL supports the idea that the chiral phosphate counterion biases the covalent L_BB2 ligand stereochemistry of the cationic intermediate complex (Supplementary Scheme 5), which in turn improves the enantioinduction in the key bond-forming step. Although the resulting e.e. is moderate for this particular set of reaction conditions, it suggests that the use of fluxional sp³-carbon units may enhance the design of ligand frameworks for ion-pair catalysis⁶⁰ and other forms of enantioselective synthesis.

Conclusions

Cope rearrangements of chiral 9-BB cages simultaneously invert every stereogenic sp³-carbon centre of their structures. These configurational rearrangements occur rapidly and reversibly, achieving the uncommon property of dynamic sp³-carbon stereochemistry—one that has remained surprisingly rare since Le Bel¹ and van’t Hoff² first identified tetrahedral carbon as a source of molecular chirality in 1874. Both the rate of sp³-carbon inversion and the equilibrium distribution of isomers are sensitive to changes in the 9-BB structure. On the one hand, the dynamics of the rearrangement processes are controlled through manipulation of covalent bonding or metal coordination of the 9-BB olefin groups, providing convenient functional handles. On the other hand, the cage adapts its configuration to minimize steric interactions with nearby fixed stereogenic elements and, in so doing, is able to transmit the stereochemical information across its rigid, tricyclic backbone. When interfaced with transition-metal complexes, the dynamic cage conveys a stereochemical preference to the chiral-at-metal^49,50 centre. Controllable and adaptable sp³-carbon stereochemistry of this kind can be exploited in enantioselective synthesis^{7,9,10,30,45,62,63} and chiral functional materials⁶⁴.