Steering the Shape‐shifting of Bullvalene‐Pd^II Complexes Through Steric and Geometric Strain

Abstract

Di‐substituted bullvalenes are fluxional molecules possessing up to fifteen possible isomers. Their shape‐shifting nature has been harnessed in host‐guest chemistry, drug design, polymers, and electromechanical systems. Despite this, principles for controlling the fluxional pathways of bullvalene, including access to isomers that are not normally populated, remain poorly understood. In this study, we elucidate the restricted shapeshifting of bis‐pyridyl bullvalene ligands in ethylenediamine (en) cis‐capped Pd^II complexes. The energetic landscape of 3‐substituted bis‐pyridyl bullvalene remains open to A and B isomers in [Pd₂(en)L₂] complexes, whilst the coordination of 4‐ and 2‐substituted bis‐pyridyl derivatives significantly restrict bullvalene's isomerization pathway. In one case, chelation of bis‐2‐pyridyl bullvalene enables access to an otherwise energetically inaccessible D‐isomer. Exclusive formation of this isomer is achieved by increasing the sterics of the cis‐capping ligand.

This study explores the use of bis‐pyridyl substituted bullvalenes as dynamic shape‐shifting ligands within cis‐capped Pd^II complexes. The ring position of the pyridyl nitrogen is critical, leading to a range of Pd₂L₂ metallacycles and simple PdL complexes, each with distinctive bullvalene isomer preferences (represented in the figure by A‐D isomer notation).

Introduction

Bullvalene is the archetypal shapeshifting molecule.^{[ 1} , ² , ³ , ^{4 ]} This hydrocarbon undergoes rapid and endless degenerate Cope rearrangements which, through time, render each CH position in the molecule exchangeable with every other. The addition of substituents generates a dynamic ensemble of non‐degenerate isomers. The interconversion of these isomers can be represented as network graphs, whereby the complexity of the network rapidly escalates with substituent number and pattern.

While bullvalene has been known since the 1960s,^{[ 5} , ⁶ , ⁷ , ⁸ , ^{9 ]} initial limitations in its synthesis gave way to a period of dormancy in literature. This was broken by a series of papers by Bode beginning in 2006 which demonstrated the potential of this system as the basis of sensors, probes, and supramolecular components.^{[ 10} , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ , ^{16 ]} However, synthetic access remained a challenge.^{[ 17} , ^{18 ]} In 2018, we introduced a two‐step method for the synthesis of mono‐ and di‐substituted bullvalenes,^{[ 19 ]} which was further advanced in 2020 through the incorporation of boronic ester synthetic handles.^{[ 20 ]} This has facilitated studies from our team and others, towards shape‐shifting polymers, medicinal chemistry, sensors, and metal complexes, based on bullvalene.^{[ 21} , ²² , ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ , ^{28 ]}

The use of bullvalene as a shapeshifting ligand in transition metal complexes and supramolecular systems carries unique appeal. Bullvalene is a molecular explorer, transversing an energetic landscape of permutable isomers. These may uniquely alter both their regio and stereochemistry. Through complexation with transition metals, bullvalene ligands can select for certain isomers, to the exclusion of all others—a feature previously demonstrated by us^{[ 29 ]} (Figure 1a) and Ihmels^{[ 28 ]} (Figure 1b). Critically, through isomerization bullvalene alters its projection in dramatic ways (Figure 2b). What has yet to be realized in either of these works is the population of multiple isomers or isomers found in disfavored regions of the network graph. Bullvalene ligand complexation may “unlock” otherwise disfavored isomers, or even allow the system to switch between supramolecular architectures. To systematically explore these concepts, this study reports the combination of a series of regioisomeric bullvalene bis‐pyridyl compounds (2⋅L, 3⋅L, and 4⋅L, Figure 2) and their coordination to “cis‐capped” Pd²⁺ salts Supporting Information.

Figure 1.

Literature examples of N‐substituted bullvalenes in coordination complexes.

Figure 2.

a) Suzuki cross coupling of 1 with an aryl halide to afford bis‐substituted aryl bullvalenes 2⋅L, 3⋅L and 4⋅L. The theoretical isomer distribution was calculated using DFT at ωB97X‐D/Def2TZVPP//ωB97X‐D/Def2SVP. b) The principle populated isomers A, B and C. [Correction added on 13 April 2026, after first online publication: Figure 2 has been updated.]

Stang's and Fujita's landmark studies laid the conceptual foundation for the rational design of metallosupramolecular polygons from directional ligands and protected convergent metal centers.^{[ 30} , ³¹ , ³² , ³³ , ³⁴ , ³⁵ , ^{36 ]} Square planar Pd²⁺ cis‐capped by a diamine ligand is a classic example, restricting ligand coordination to two mutually orthogonal sites. This dynamic coordination environment is error correcting, selecting for the thermodynamic product—a marked feature of both bullvalene and Pd^II‐N coordination chemistry.^{[ 30} , ³¹ , ³² , ³³ , ³⁴ , ³⁵ , ^{36 ]}

Our recent study explored a fluxional coordination cage, assembled from four bis‐3‐pyridyl bullvalene (3⋅L) ligands and two “naked” Pd²⁺ metal centers (Figure 1a). The structural complexity of the fluxional metallosupramolecular system was enormous but could be reduced through exchanging the bound tetrafluoroborate anion with a halide guest (X^‒). This simplified the mixture to a defined [X@Pd₂L₄]³⁺ coordination cage, locked into a limited isomer B to B′ transition. Due to the complexity of the initial system, it was difficult to distinguish how coordination and host‐guest chemistry individually affect the energetic landscape of bullvalene. To pinpoint the effect of the first step (coordination to Pd^II), we focus here on cis‐capped Pd²⁺ complexes of bullvalene—systems bearing only one or two ligands, thus considerably reducing structural complexity.

Results and Discussion

To expand the limited family of bis‐pyridyl bullvalenes, we utilized a synthetic protocol developed in one of our groups.^{[ 29 ]} The ligands bis‐2‐pyridyl (2⋅L), bis‐3‐pyridyl (3⋅L), and bis‐4‐pyridyl (4⋅L) bullvalene, were synthesized from bis‐Bpin‐bullvalene (1) under standard Suzuki cross‐coupling conditions, in good yields (Figure 1a). Bis‐pyridyl bullvalenes exist as a dynamic ensemble of fifteen possible isomers. In solution, three major isomers are populated; denoted as A, B, and C (Figure 2b), with A and B predominating over C.

With this set of ligands in hand we proceeded in examining their self‐assembly with cis‐capped Pd^II complexes; 1.0 equivalent of the respective bis(pyridyl) bullvalene ligand was combined with 1.0 equivalent of Pd(Y)(BF₄)₂ (Y: ethylene diamine (en) or tetramethyl ethylene diamine (tmeda) in CD₃CN, followed by sonication for 15 min under ambient conditions. This afforded the respective complexes in virtually quantitative yield.

Treatment of 4⋅L with [Pd(en)(MeCN)₂](BF₄)₂, in CD₃CN afforded 4⋅en—a complex assembled from only isomer A (Figure 3a). The presence of a [Pd₂(en)L₂ + nBF₄]^{4− n} complex was confirmed by ESI‐HRMS (See SI for full details). Both bullvalene ligands 4⋅en are as their A isomer—denoted AA—but exist as two interconverting well‐resolved diastereoisomers. These complexes can be differentiated by the relative orientation of the two bullvalene cores. Of these two complexes, here defined as AA‐syn and the C₂‐symmetric AA‐anti, there is a major (70%) and minor (30%) diastereoisomer. The experimental assignment of the major and minor complexes as either AA‐syn or AA‐anti in solution was not possible. 4·en was also prepared in a range of solvents, as well as employing nitrate as the counterion. In all cases analogous results were obtained (See SI for full details).

Figure 3.

a) Complexation of 4⋅L with Pd(Y)(BF₄)₂ affording 4⋅en or 4⋅tmeda b) AA‐syn and AA‐anti variant observed for 4⋅en, and only one variant for 4⋅tmeda. c) Truncated ¹H NMR spectra (CD₃CN, 400 MHz, −35 °C) stacked comparison of 4⋅L, 4⋅en and 4⋅tmeda. Despite 4⋅tmeda only presenting as one diastereoisomer, restricted rotation of the pyridyl signals leads to additional splitting of the signals. d) X‐ray structures of 4·en and Fujita's square complex, with anions and solvent molecules omitted for clarity (CCDC ID number 753 499). e) Isomer network diagram for 4·en for a stepwise isomerization mechanism between the syn and anti diastereoisomers, with the lowest energy pathway highlighted in bold. Nodes represent distinct isomers and the lines/edges represent the transition state structures required to interconvert those isomers. Nodes and pathways marked in grey represent geometrically distorted structures that may be discounted.

When the bulky cis‐capping ligand tmeda was employed, we observe the exclusive formation of a single AA macrocycle diastereoisomer, 4·tmeda (Figure 3c). A feature of this complex is the restricted bond rotation around the pyridyl groups as evident by the resolution of four chemically non‐equivalent proton signals in the aromatic region (at −35 °C).

In the ¹H NMR spectrum of 4·en, the signals relating to the bullvalene core are distinctly resolved at room temperature, relative to 4·L (See SI for full details). This indicates that the rate of bullvalene isomerism is greatly reduced compared with the free ligand. Despite this, dynamic exchange between the AA‐syn and AA‐anti macrocycle diastereoisomers does proceed, as revealed through selective 1D EXSY ¹H NMR experiments with an estimated barrier of 73 kJ.mol⁻¹. Synthesis of the analogous platinum complex, gave 4·en·Pt as a mixture of AA diastereoisomers, which displayed near identical kinetic exchange parameters (see SI for full details). This result provides strong evidence for a non‐dissociative exchange mechanism.

Single crystals of 4⋅en were obtained by slowly diffusing diisopropyl ether into a MeCN solution of the complex. X‐ray crystallography confirmed the structure of the [Pd₂(en)₂(4·L)₂](BF₄)₂ complex as the AA‐anti isomer. The square complex is formed by two A‐isomer ligands, which offer a bite‐angle of 93° similar to the 94° angle between palladium centers observed in Fujita's Pd₄L₄ square complex (Figure 3d).^{[ 32 ]}

The almost right angle offered by isomer A in the macrocyclic complex appears to be the major entropic driving force in the formation of the observed complex. The more obtuse bite angles offered by isomers B and C (∼120°) would likely require higher‐nuclearity complexes (incurring an entropic penalty), which are not observed in the ESI mass spectrum. Instead, the system maintains a pseudo‐square Pd₂(en)₂L₂ geometry whilst shapeshifting between the two AA‐syn and AA‐anti isomers.

To explore the reaction graph and properties of this system we turned to computational analysis. All calculations were performed using Density Functional Theory (DFT) employing the ωB97X‐D method and Def2SVP basis set for geometry optimization and frequency analysis, with final single point calculations at ωB97X‐D/Def2TZVPP (see SI for full details). Initially we compared the ground state energies between AA‐syn and AA‐anti. For 4⋅en, the AA‐anti isomer is predicted to be stabilized relative to AA‐syn by only 0.8 kJ mol⁻¹, consistent with the experimentally observed mixture. For 4⋅tmeda, the AA‐syn isomer is calculated to be the more stable relative to the AA‐anti at 5 kJ mol⁻¹. Based on this, we can tentatively assign 4⋅tmeda as the AA‐syn diastereoisomer.

We next considered the pathway dependence of the 4⋅en AA‐syn and AA‐anti interconversion. A network graph of 4⋅en is presented in Figure 3e, which charts the isomerization of one of the bullvalene units, with the other held static as isomer A in a Pd₂(en)₂L₂ complex. The anti and syn diastereomeric regions of the reaction graph are highlighted. These may only exchange through isomers or transition structures that vertically bisect the graph.

The DFT calculations reveal a plausible low energy isomerism pathway connecting AA‐anti to AA‐syn: AA‐anti ↔ AB‐anti ↔ AC‐anti ↔ AC‐syn ↔ AB‐syn ↔ AA‐syn. The highest energy barrier along this path is 83 kJ mol⁻¹, in rough agreement with the experimentally observed kinetics.

We next turned to the ligand 3⋅L and explored its complexation with cis‐capped [Pd(en)(MeCN)₂](BF₄)₂ in CD₃CN (Figure 4). The assembly resulted in a complex mixture. The ¹H NMR spectrum displayed broad room temperature signals, indicative of rapid isomerism, and was resolved by variable temperature (VT) ¹H NMR analysis (Figure 4c). Despite this, ESI‐HRMS revealed only a single complex, with peaks detected at 330, 538 and 1163 m/z, assigned to the binuclear complex [Pd₂(en)L₂ + nBF₄]^{4− n} (n = 1 – 3), which herein is referred to as 3·en. Only a single BF₄ ⁻ resonance was detected in the ¹⁹F NMR spectrum, suggesting the BF₄ ⁻ anion is not tightly encapsulated by a host species in the complex mixture. To investigate the effect of solvent, 3⋅en was prepared in DMSO. However, a similar degree of complexity was observed in the room temperature ¹H NMR spectrum.

Figure 4.

a) Complexation of 3⋅L with Pd(Y)(BF₄)₂ affording 3⋅en or 3⋅tmeda. b) 3⋅en populates a complex mixture of isomers B, B’ and A. 3⋅tmeda also populates isomer B and A. c) Truncated ¹H NMR spectra (CD₃CN, 400 MHz, ‐35 °C) stacked comparison of 3⋅L (top), followed by 3⋅en, a selective 1D TOCSY exert of isomer B of 3⋅en and 3⋅tmeda (bottom). d) X‐ray structure of 3·tmeda showing the BF₄ ^‒ in the central cavity. Other anions and solvent molecules omitted for clarity. e) AA, AB, and BB 3⋅en diastereoisomers. f) Possible macrocycle topologies, combining to form 82 possible topo‐isomeric 3⋅en macrocycles.

Time‐resolved selective 1D TOCSY was used to elucidate the major isomer of 3⋅en in CD₃CN (see SI for full details). Signals attributable to isomer A, as well as two sets of signals attributable to isomer B were observed. Intriguingly, self‐assembly of 3⋅en with the more bulky tmeda ligand in place of en led to a greater preference for an isomer B complex, followed by isomer A (see SI for full details). ESI‐HRMS supported the presence of the [Pd₂(tmeda)₂L₂]²⁺ complex, and we were ultimately able to elucidate its structure through X‐ray crystallography (Figure 4d).

3⋅tmeda crystallizes in the monoclinic space group C2/c with half of the [Pd₂(tmeda)₂L₂] macrocycle in the asymmetric unit. Indeed, the ligand adopts isomer B with both apical R and S enantiomers present in the structure. C–H···π interactions between the methyl groups of tmeda and the pyridine ring of the ligand (closest distance: 2.78 Å) may explain why the bulky cis‐capping ligand influences the isomer distribution, as these interactions are expected to be absent in the 3⋅en complex. Interestingly, the macrocycle resembles the shape of the previously reported Pd₂L₄ lantern cage^{[ 29 ]} and indeed, a disordered BF₄ ^‒ anion could be located in the cavity, stabilized by multiple C–H···F hydrogen bonds ranging from 2.1–2.5 Å.

The ¹H NMR spectra of 3⋅tmeda strongly suggest a complex mixture of isomers whereby BF₄ ⁻ is not encapsulated. In our previous study on Pd₂L₄ (L =  3·L) cages we observed tight binding and encapsulation of halide counterions in solution, and a dramatic clarification of the ¹H NMR spectra (compared to BF₄ ⁻). However, when solutions of  3⋅tmeda were treated with tetrabutyl ammonium salts of Cl^‒ and I^‒, no significant changes to the ¹H NMR spectra were observed (see SI for full details) indicating no tight binding of the counter‐ion. This also highlights the necessity for a defined cavity for guest inclusion.^{[ 37} , ^{38 ]}

The 3⋅en and 3⋅tmeda complexes represent stereochemically complex macrocycles, which populate both isomer A and B. The shape of isomers A and B has inherent directionality in the macrocycle, embodied by the apical bridgehead. Furthermore, isomer B is itself chiral thereby possessing two stereochemical degrees of freedom within the macrocycle. By just considering the A and B bullvalene isomers of 3⋅L, we might expect to observe AA, AB, and BB macrocycles. Of these the AA would have AA‐syn and AA‐anti diastereoisomers, the AB system would have AB‐syn and AB‐anti diastereoisomers, and the BB system would have B(R/S)B(R/S)‐syn B(R/S)B(S/R)‐syn, B(R/S)B(R/S)‐anti and B(R/S)B(S/R)‐anti diastereoisomers (Figure 4e). Interconversion of any of these eight isomers necessitates bullvalene isomerism. However, overlayed on this picture are topological features of the macrocycle itself and the stereochemical relationship between the two Pd‐en‐3⋅L square planes. We identify 5 distinct macrocycle topologies as shown in Figure 4f, and ascribe these as Pd‐cis, Pd‐trans, Pd‐halftwist, Pd‐C₂ ^h , and Pd‐helical. Combining all these features we can enumerate 82 possible topo‐isomers of 3⋅en and 3⋅tmeda macrocycles. Simple DFT geometry optimization of these structures (ωB97X‐D/Def2SVP) indicates that a multitude have low laying energies, in broad agreement with the complex ¹H NMR spectra of these compounds (see SI for full details). The computational modelling also indicates an apparent rigidity of many of these isomers, raising the possibility that their interconversion might by necessity be concomitant with bullvalene isomerism. However, given the complex NMR spectra in this case, this remains a speculation. Regardless of the multitude of isomeric possibilities in solution, crystallization of 3⋅tmeda selects for the B(R/S)B(R/S)‐syn‐Pd‐cis with both bullvalenes endo to the macrocycle curvature. Notably, in the analogous 3⋅en series, this isomer is calculated to be the lowest energy isomer.

Treatment of 2⋅L with [Pd(en)(MeCN)₂](BF₄)₂ in CD₃CN under ambient conditions, afforded the complex 2⋅en (Figure 5). In contrast to 3⋅en and 4⋅en, ESI‐HRMS revealed a mononuclear complex with prominent peaks at 225 and 537 m/z, assignable to a [Pd(en)L + nBF₄]^{2− n} species (n = 0, 1). VT NMR analysis of this complex revealed a distinct preference for two isomers—A (24%) and D (76%), shown in Figure 5b. This later isomer has two adjacent alkene substituents, a pattern usually disfavored, anpreviously only observed in annulated bullvalenes.^{[ 39 ]} Further investigation of the room temperature ¹H NMR spectrum revealed a mixture of sharp signals assigned to isomer A and broad signals assigned to isomer D. These sets of resonances did not coalesce upon heating to higher temperatures (40°C–80°C, see SI for full details). 1D selective EXSY experiments showed slow exchange of isomer A–D at 25 °C (see SI for full details). Despite this, Isomer D's dynamic behavior at room temperature is unsurprising. Analysis of the network reveals its engagement in a degenerate D–D isomer interconversion. Interestingly, the treatment of 2⋅L with sterically bulky [Pd(tmeda)(MeCN)₂](BF₄)₂ led to a similar mononuclear complex, albeit with isomer D exclusively populated (see SI for full details). Single crystal X‐ray diffraction unambiguously confirmed the structure of this complex, with the D‐isomer forming a seven‐membered chelate complex with Pd^II(tmeda) (Figure 5d). C–H···π interactions (shortest contact = 2.68 Å) between the tmeda methyl groups and the bullvalene alkene bridge provide evidence for the increased stabilization of this particular isomer, as well as the splitting of the methyl resonances in the ¹H NMR spectrum.

Figure 5.

a) Complexation of 2⋅L with Pd(Y)(BF₄)₂ affording either 2⋅en or 2⋅tmeda. b) 2⋅en populates isomer D as the major species followed by isomer A as the minor. 2⋅tmeda exclusively populates isomer D. c) Stacked truncated ¹H NMR spectra (CD₃CN, 400 MHz) of 2⋅L, 2⋅en and 2⋅tmeda with the populated isomers annotated. d) X‐ray structure of 2⋅tmeda, with the C–H···π interaction shown as a red dotted line. e) Isomer network diagram of 2⋅en. Nodes represent distinct isomer and the lines connecting the transition state structures required to interconvert those isomers. Upon complexation, only isomer A or isomer D may be populated depending on the cis‐capping ligand.

A computational survey of the reaction graph of 2⋅en is broadly consistent with the experimental observations. Isomers A and D are predicted to be very close in energy, with isomer A stabilized by only a 4 kJ mol⁻¹. A comparison of isomer A and D in 2⋅tmeda predicts isomer D to stabilized by 14 kJ mol⁻¹, consistent with its exclusive experimental preference. However, the calculated transition state energies interconnecting 2‐en isomer A and D are somewhat overestimating the absolute barriers in this case, compared with their experimentally observed exchange.

Conclusion

In summary, the restricted shapeshifting of bis‐pyridyl bullvalene ligands 4⋅L, 3⋅L, and 2⋅L was elucidated in binuclear and mononuclear cis‐capped Pd^II complexes. Whilst the free ligands exclusively populate isomers A, B and C in varied ratios, complexation with Pd(en/tmeda) results in a significant change in this isomer distribution. Coordination‐driven self‐assembly of 4⋅L afforded Pd₂L₂ 4⋅en/tmeda exclusively favoring isomer A by ¹H NMR spectroscopy, with the bulkier tmeda complex narrowing the complexity down to a single AA diastereomer. By contrast 3⋅L formed a complex mixture of Pd₂L₂ complexes consisting of isomer A and B, with tmeda further amplifying isomer B in this mixture. Complexation of 2⋅L with Pd(en/tmeda) afforded a mononuclear PdL complex. The complex capped by en (2⋅en) populated isomer A and an otherwise disfavored isomer D, with this isomer exclusively favoured in the 2·tmeda complex.

Engaging bullvalene as a ligand in metallosupramolecular chemistry invites enormous complexity. Whilst host‐guest chemistry was previously shown to influence bullvalene's isomerization,^{[ 29 ]} here, in the absence of host‐guest interactions, we have demonstrated that both the substitution pattern of the binding heteroatom (N) and the steric bulk about the metal center can be harnessed to dictate or reduce the number of isomer(s) observed. Complexation can both “lock down” parts of bullvalene's shape‐shifting network and “unlock” previously inaccessible isomers. The ability to predict and control shape switching within coordination complexes provides a foundation for advancing research into dynamic catalysts,^{[ 40 ]} functional materials,^{[ 24 ]} and catch‐and‐release supramolecular systems.^{[ 41 ]}

Supporting Information

Please include SI references with consecutive numbering directly after the last manuscript reference: 1, 2, 3,…30, 31.

Conflict of Interests

The authors declare no conflict of interest.

Supporting information

Supporting Information

Data Availability Statement

The data that support the findings of this study are available in the Supporting Information of this article.