Introduction

While completely translating host–guest complementarity to the design and synthesis of a perfectly preorganized and rigid host is highly challenging, engineering flexibility into host framework enables conformational readjustment upon guest binding and strengthens the host–guest interactions1. However, structural flexibility often inevitably imposes an additional energy cost for host reorganization upon formation of host–guest complex2. A delicate balance of structural rigidity and flexibility is hence of critical importance in the design of molecular receptors, especially for multitopic hosts that need to change their conformations when binding to guests of vastly different steric and electronic features, and meanwhile minimizing the energy penalty for reorganization3,4. Selecting suitable covalent linkers of just enough flexibility and connecting binding motifs using these linkers at a right position are therefore key to a successful host design (Fig. 1). In this context, mechanical interlocking of rigid molecular components as mechanically interlocked molecules (MIMs) will be an alternative strategy to introduce structural flexibility into the host design5,6. In contrast to hosts with covalently linked binding motifs, the extent of preorganization in MIM-based hosts will be regulated by their co-conformational flexibility, which in turn can be fine-tuned by additional parameters that govern the nature of mechanical bond (e.g., ring tightness, strength of inter-component interactions)7,8. Moreover, the large-amplitude co-conformational changes unique to MIMs could allow the binding motifs to better sample the three-dimensional space around the guest while undergoing rearrangement, which serve to optimize the host–guest complementarity and allow the hosts to respond to changes in the external environment9,10,11,12.

Fig. 1: Designs and features of partially flexible hosts.
figure 1

a Schematic illustration of a macrocyclic host with rigid and flexible covalent linkers. b Schematic illustration of a [2]catenane host with mechanically flexible but covalently rigid macrocycles.

In this work, a heteroditopic [2]catenane composed of two interlocked macrocycles containing both bipyridine (bpy) and urea motifs for respective cation and anion binding is described. In addition to copper(I) ion that templates its synthesis, the host also displays an unexpectedly strong and selective binding to sulfate anion (log K1 ~ 4.3 and log K2 ~ 3.6 in 5% aqueous DMSO), despite no specific sulfate binding pocket has been included in the design of the catenane host. Although the covalent backbone of the individual macrocycles is relatively rigid, the [2]catenane is co-conformationally flexible that the interlocked rings can undergo a “180°-turn” to switch from the achiral, copper(I)-bound co-conformer to the mechanical chelating sulfate complex in a chiral co-conformation. In addition to the significance in the transport13,14, sensing15,16,17,18, delivery, and recovery of ionic species and resources19,20,21, this work also highlights mechanical interlocking as an efficient element in designing molecular receptors with versatile structures, properties and guest binding behavior.

Results and discussion

To obtain catenane-derived receptors, the Cu+-templated, dynamic urea formation from building blocks 1 and 2 containing respectively t-butylamine and isocyanate functional groups was explored to simultaneously introduce the cation binding bipyridine (bpy) and anion binding urea groups during the catenane synthesis22,23. Catenane C is obtained in three simple steps without any chromatographic purification (Fig. 2). By virtue of the strong Cu+-bpy template and the effective ring-closing from the error-checking dynamic urea formation, the catenane topology was efficiently created. Simple acid treatment removed the t-butyl group after the catenane formation, which locked the dynamic urea and activated their anion binding potential24. Catenane C was then obtained by using ethylenediamine to extract the Cu+ template followed by aqueous washing25. Control [2]catenane and macrocycle M featuring only one bis-urea bpy-based macrocycle were also prepared to evaluate the effects of mechanical interlocking on the binding (Fig. 2 and Supplementary Figs. 1 and 2). Of note, because of the different possible ways of connecting the reactive ends of [Cu(1)2]+ when diisocyanates of different structural features are used, formation of the catenane topology is not necessarily efficient, and attempts to synthesize related catenanes were either unsuccessful or only gave the target [2]catenane in a low yield. For example, the related catenane was obtained only in 8% yield along with other non-interlocked products by following the same synthetic procedure (see Supplementary Fig. 3 for more details)26.

Fig. 2: Synthesis of tetra-urea catenanes and structures of control bis-urea hosts.
figure 2

The tetra-urea [2]catenanes C and , and bis-urea control macrocycle M and [2]catenane were synthesized from dynamic urea formation from the corresponding t-butylamine and isocyanate precursors.

HR-ESI-MS spectrum of C showed molecular ion peaks at m/z 1375.5458 [C + Na]+ and 677.2873 [C + 2H]2+ that are consistent to the expected molecular formula C82H72N12O8 of the catenane, and the interlocked nature is evidenced by the direct fragmentation of the singly charged molecular ion peak to the constituting macrocycle at m/z 677.2837 [M + H]+ in the MS/MS experiment (Supplementary Fig. 5). The 1H NMR spectrum (600 MHz, 298 K, DMSO-d6) of C showed that the catenane adopts a highly symmetrical structure with three Hbpy and two Hdpm resonances, suggesting that the two interlocked rings are equivalent and the [2]catenane possesses two C2 axis. The two urea NH were found at 6.52 ppm and 8.25 ppm, which are similar to other reported urea-based macrocycles23,27. Only slight spectral changes were observed when the temperature was increased from 298 K to 358 K (Supplementary Fig. 26), which may be explained by a rapid co-conformational exchange of the interlocked host. Attempt to obtain the low-temperature NMR spectrum of C in other solvents such as CDCl3 failed due to its poor solubility. Comparing the 1H NMR spectra of C and the non-interlocked macrocycle M, proton resonances are generally more upfield shifted in the catenane due to a stronger shielding effect as a result of mechanical interlocking (e.g. Δδ ~ 0.52–0.61 ppm for Hbpy and 0.30–0.42 ppm for Hdpm).

Similar to most other metal-templated catenanes, the bis(bpy) cavity crafted in C during the templated synthesis could be complementary for cation binding11,28,29,30. As expected, addition of [Cu(MeCN)4](PF6) to a solution of C led to the formation of a species with a new UV-Vis absorption at 450 nm characteristic to the Cu+-to-bpy MLCT and a 1H NMR spectrum consistent to that of [CuC]+ (Supplementary Fig. 30 and 36). The 1:1 Cu(I) complexation was found to be in slow exchange on the NMR timescale, and no further spectral changes were observed after one equivalent of Cu+ ion had been administered. In the 1H NMR spectrum of [CuC](PF6), resonances of the meta-substituted aryl spacer signals HAr are significantly upfield shifted when compared to that of the free host (Fig. 3), and NOE cross peaks between HAr and Hbpy were also found in the 2D NOESY spectrum (Supplementary Fig. 28). These spectral features are consistent with the reinforced π-stacking interactions between the aromatic units as a result of the Cu+-bpy coordination22. Slight chemical shift changes were found for the urea NH, H1, and H2, and a more significant downfield shift by ca. 0.4 ppm was noticed for the diphenylmethylene (dpm) H3. Because of the interlocked structure, the dpm linkers are likely perpendicularly oriented with respect to the bpy of the other macrocycles, hence placing H3 in the deshielding region of the bpy in the Cu(I) complex.

Fig. 3: 1H NMR studies of the [2]catenane.
figure 3

Partial 1H NMR spectra (600 MHz, DMSO-d6, 298 K) of a macrocycle M, b catenane C, and c Cu(I) complex [CuC](PF6). Aromatic signals from bpy and dpm units are highlighted in blue and pink respectively.

While the Cu+ binding may not be surprising, the bpy-derived [2]catenane showed a very different cation selectivity than those closely related bis(phenanthroline) catenanes. Although catenanes with a bis(phenanthroline) core have been shown to form complexes with a range of other cations31,32, no obvious changes in the 1H NMR and UV-Vis spectra were observed when Li+, Na+, Fe2+, Co2+, Ni2+, Zn2+, and Cd2+ ions were added to a DMSO solution of C (Supplementary Fig. 30 and 31). The lack of apparent binding of other metal ions to C is likely due to a slow coordination kinetics due to the (co)conformational flexibility of the bpy and the catenane, effects from the counter ions and solvent, as well as a mismatch in the coordination geometry. In DMSO, no coordination of 6,6’-dimethyl-2,2’-bipyridine to Zn2+ (as ClO4 salt) was observed (Supplementary Fig. 38), and similar kinetic effects have indeed been reported for the direct metalation of a bpy-derived molecular knot33. The addition of 1 eq. of Cu2+ led to proton signals that are assignable to [CuC]+ and a new absorption peak at 450 nm in the 1H NMR and UV-Vis spectra respectively, showing that the Cu2+ has been reduced with the subsequent formation of the Cu(I) catenane34. On the other hand, broadening and gradual shifting of the proton resonances of C were observed when up to 2 eq. of Ag+ ion was added, suggesting the formation of a new, fast exchanging Ag+-associated species in DMSO. Yet, the Ag+ binding is not as strong as that of Cu+, and the Ag+-associated species can be converted to [CuC]+ when a stoichiometric amount of Cu+ was introduced (Supplementary Fig. 37). The fast exchange of the Ag(I) species may be unusual if the Ag+ ion is fully encapsulated in the catenane cavity35,36, and more extensive structural analysis and studies will be warranted to fully understand the impact of the catenand effect on the coordination properties of these bpy-derived MIMs on different metal ions.

The incorporation of four hydrogen bonding urea motifs in the macrocycle backbone is envisaged to endow the [2]catenane C with potent anion binding capability. To investigate its anion recognition behavior, 1H NMR anion titration experiments were conducted, in which aliquots of different anions (Cl, Br, I, NO3, HSO4, SO42−, CH3COO, H2PO4) as their tetrabutylammonium (TBA) salts were added to the catenane solutions in DMSO-d6 (Supplementary Fig. 40). Only negligible spectral changes were observed when Br, I, NO3, HSO4 were added, suggestive of no significant binding to these anions. In stark contrast, the addition of sulfate resulted in a drastic downfield perturbation to the urea NH signals (Δδ = 1.56 ppm and 2.22 ppm over 2 eq. of sulfate), suggesting the oxoanion is bound via the formation of urea NH hydrogen bonds (Fig. 4). Further addition of sulfate (>2 eq.) did not induce further chemical shift changes, suggesting that the urea anion binding sites have been saturated. Downfield shifts of urea proton signals, in a less pronounced magnitude were also observed upon the addition of 10 eq. of Cl, H2PO4 and CH3COO, indicating that the catenane also binds to these anions but with a weaker affinity. The anion binding stoichiometry was revealed by Job plot (Supplementary Figs. 41 and 42), and the 1:1/1:2 host–guest anion association constants were determined from the titration data and are summarized in Table 1.

Fig. 4: Characterization of sulfate binding to catenane C.
figure 4

a Partial 1H NMR (500 MHz, DMSO-d6, 298 K) spectra of C in the presence of increasing amount of TBA sulfate (signals from urea NH, bpy, dpm and methylene protons are highlighted in green, blue, pink and yellow respectively). Methylene protons H1 and H2 become diastereotopic in the presence of 2 eq. or more of sulfate. b Change in chemical shift of the urea NH proton originally at 8.25 ppm. c Job plot of the binding showing a 1:2 host–guest stoichiometry. d HR-ESI-MS (-ve) spectrum obtained from a sample of C in the presence of 10 eq. of TBA sulfate. e Isotopic distribution of the mass signal at m/z 1789.8.

Table 1 Anion binding constants of catenane C and control compounds and Ma

In particular, the catenane was found to be selective to sulfate with a 1:2 stoichiometry and binding constants of >105 M−1 in DMSO-d6 for both the first and second sulfate association. Encouraged by the strong and selective binding of sulfate anion in DMSO-d6, 1H NMR titration experiment was carried out in the more competitive 5% D2O in DMSO-d6. For comparison, the sulfate association constants of the macrocycle M, and the control [2]catenane consisting of only one bis-urea-based bpy macrocycle, were also determined. Both macrocycle M and catenane displayed comparable 1:1 host–guest binding to sulfate anion with K ~ 400 M−1. In contrast, the tetra-urea catenane C demonstrated significantly stronger 1:2 host–guest association to the oxoanion, with K1 = 2.1 × 104 M−1 and K2 = 3.6 × 103 M−1 at 298 K. The negative cooperativity (K1/K2 ≈ 5.8) of the first and second sulfate binding is likely due to the electrostatic repulsion between the two anions within the neutral catenane37. Van’t Hoff analysis shows the binding is driven by enthalpy with a negative entropy change (ΔH = –23 kcal mol−1; ΔS = –41 cal mol−1 K−1), which could be explained by the decrease in the (co)conformational flexibility of the catenane upon guest binding.

The sulfate complex can also be directly observed by HR-ESI-MS, in which peaks for the 1:1 and 1:2 sulfate complexes at m/z 1449.5154 [C + SO4 + H], 1691.7937 [C + SO4 + TBA] and 1789.7611 [C + 2(SO4) + 2H + TBA], with isotopic distributions corresponding to the respective molecular formula, were found in the MS spectrum (Fig. 4). These findings suggest the mode of sulfate binding by C features a mechanical chelation of a sulfate anion by two urea groups from the two interlocked rings, and more importantly highlights the critical role of the mechanical flexibility for efficient, large-amplitude co-conformational rearrangement of the binding motifs despite the rigidity of the covalent macrocyclic skeleton.

Apart from the strong binding in competitive aqueous environment, sulfate binding of C is also characterized by a distinctive structural change of the host. In addition to the downfield shifting of the urea NH, the 1H NMR spectrum (500 MHz, DMSO-d6, 298 K) of C in the presence of 10 eq. SO42− also showed that the Hbpy resonances are downfield shifted to 8.34 ppm, 7.72 ppm, and 7.58 ppm, while that of the Hdpm are more upfield at 6.69 ppm and 5.98 ppm when compared to that of the free host C and macrocycle M, suggestive of a respectively weaker and stronger shielding effect for the bpy and dpm units in the sulfate complex. A significant upfield shift of the dpm CH2 (H3) was also observed in the sulfate complex, indicating that C adopts a co-conformation in which the bpy are more exposed at the catenane exterior, and concomitantly the dpm are buried in the catenane center upon sulfate binding38. Compared to the co-conformation in [CuC]+ in which the bpy and dpm units locate respectively at the central and peripheral positions of the catenane, the two interlocked macrocycles have each undergone a “180°-turn”, and the bpy and dpm exchanged their relative positions in the sulfate complex. Furthermore, the 2D NOESY spectrum of C obtained in the presence of 10 eq. SO42− showed NOE cross peaks between Hdpm and HAr protons (Supplementary Fig. 29), which is consistent to the interlocked macrocycles rotated inside-out, such that the dpm units are now in a closer proximity to the meta-substituted aryl spacer of the other macrocycle. Noteworthily, literature examples of sulfate-selective receptors frequently invoke a hydrophobic effect from host encapsulation of sulfate anions, which shields the highly hydrophilic anion and assists its desolvation in aqueous environment27,39,40,41,42,43,44,45,46. In contrast, sulfate anion induces a large-amplitude co-conformational change in the interlocked backbone of C, bringing the urea donors in the two macrocycles in proximity for convergent sulfate binding. The high flexibility and ability to undergo large-amplitude co-conformational change thus enable the catenane host to adapt to ionic guests of different size, charge, geometry and binding stoichiometry (i.e., 1:1 binding of monocationic, spherical Cu+ and 1:2 binding of dianionic, tetrahedral SO42−).

The anion binding of C is also studied by DFT modeling, and the computed association constants were found comparable to our experimental data (Supplementary Table 1). Consistent to the NMR structural studies, the energy-minimized structure of the 1:2 sulfate complex features a mechanical chelate connected via NH···O hydrogen bonds between the urea and the anion, and the dpm units are buried in the center of the catenane (Supplementary Fig. 63). Close contacts between the anions and other parts of the catenane were also observed in these simulated anion complexes, suggesting that all the urea hydrogen bonds and these other interactions, albeit some of which are less obvious, are collectively responsible for the overall stability of the anion complexes. The anion selectivity is therefore a result of the delicate balance of all these non-covalent interactions involving the specific anion and the different catenane (co)conformations.

Sulfate binding also leads to a change in the stereochemistry of C. While the time-averaged structures of the free host and Cu(I) complex [CuC]+ display no chiral feature in our spectroscopic studies, diastereotopic splitting of the methylene singlets H1 and H2 to a pair of spin-coupled doublets can be observed in the presence of 10 eq. SO42−, showing that the sulfate ions stabilize a specific chiral co-conformer of C. Weakening the sulfate association by adding D2O to the NMR sample in DMSO-d6 led to the coalescence of H1 and H2 doublets (Supplementary Fig. 58), and in the presence of 5% D2O or above, the resulting 1H NMR spectrum essentially showed no chiral feature although significant sulfate binding can still be measured as discussed previously. Variable temperature 1H NMR studies of the sulfate complex showed the diastereotopic doublets of H1 coalesce at 318 K, corresponding to a racemization barrier of 15.7 kcal·mol−1 in DMSO-d6 (Supplementary Fig. 57). These findings not only confirm that sulfate selects and binds strongly to a chiral co-conformer of the catenane host, but also demonstrate a rare example in which the extent of chiral property expression can be controlled by the specific condition of host–guest binding. For both the Cu(I) and sulfate complexes of C, 1H NMR analysis showed that the two interlocked macrocycles are equivalent, and both complexes are C2 symmetric with respect to the catenane molecular axis with only three bpy and two dpm aromatic signals. Chirality of the sulfate complex is therefore explained by a tilted orientation of the two macrocyclic planes as a result of the rocking of the macrocycles being arrested by the sulfate binding, and such structure has been referred as being mechanically helically chiral (Fig. 5)47,48,49,50,51,52,53,54. On the other hand, the tetrahedral Cu+-bpy coordination brings the bpy to the catenane center and enforces a perpendicular orientation of the macrocyclic planes to give an achiral structure. Other co-conformations of C are unlikely to be C2 symmetric although some of them are also chiral55,56. Finally, the chiral structure induced by sulfate binding is found to be characteristic to C, and the related urea catenane showed no chiral spectral feature although the interlocked host also displayed a strong sulfate binding in DMSO-d6 (Supplementary Fig. 54).

Fig. 5: Co-conformational switching of catenane C upon selective ion binding.
figure 5

Co-conformational exchange of C involving a 180°-turn of the macrocycles upon guest binding to produce achiral Cu(I) and chiral sulfate complexes. Schematic illustrations of complexes in different perspectives are provided (blue: bpy; pink: dpm; gray ball: Cu+; yellow ball: SO42−).

Further structural details of the sulfate complex are revealed by single crystal X-ray diffraction analysis (Fig. 6). Single crystals of the sulfate complex were obtained by slow vapor diffusion of diisopropyl ether into a DMF solution of C in the presence of 10 eq. TBA2SO4. The sulfate complex crystallized in the monoclinic P21/n space group with each asymmetric unit consists of one catenane, one sulfate, two TBA ions and four DMF molecules. Consistent with the NMR structural studies, the catenane was found to adopt a co-conformation in which the dpm and bpy locate respectively in the catenane center and periphery, and an average interplanar separation of 4.1 Å was found between the dpm units from the two macrocycles. The bpy units are also found to adopt a trans configuration that the two pyridine donors are pointing toward opposite directions, probably as a result of relieving any potential ring strain in the interlocked macrocycle. Eight N–H···O hydrogen bonds at a distance of 2.755–3.133 Å were identified between each of the sulfate and four urea groups from two different interlocked rings of two molecules of C, confirming the mechanical chelating mode of the sulfate binding. The chiral co-conformation is evidenced by the tilted arrangement of the macrocyclic planes at an angle of ~62°. The crystal sample is overall racemic and the enantiomeric co-conformers of opposite chirality are bridged by a sulfate ion in an alternate fashion to result in infinite 1D chains, which further align with adjacent strands at a distance of ~4.1 Å (Supplementary Fig. 65).

Fig. 6: Solid-state structures of the sulfate complex of C.
figure 6

a X-ray crystal structure showing the sulfate catenane complex in different perspectives. Color code: gray, C; blue, N; red, O; white, H; yellow, S. b Hydrogen bonding interactions between the sulfate and urea groups from two catenanes. c Arrangement of the enantiomeric co-conformers. The two macrocycles of C are colored in green and orange respectively, and sulfate anions are colored in purple. TBA ions and DMF molecules are omitted for clarity.

The guest-controlled, reversible co-conformational mechanostereochemical switching of C was further demonstrated by the sequential introduction of Cu+ and SO42−, as well as the competitive CN and Ba2+ ions. As shown in Fig. 7, addition of 4 eq. of TBA2SO4 to a 1 mM DMSO-d6 solution of C resulted in the formation of the 1:2 sulfate complex as shown by the downfield shifted urea NH, as well as the diastereotopic H1 and H2 resonances. This chiral, sulfate-bound form of C can be made co-conformationally flexible again by addition of Ba(OTf)2 that led to the formation of BaSO4 and reversed the spectral changes, with a resulting 1H NMR spectrum essentially the same as that of the initial free host C. Further switching to the achiral form of the Cu+ complex was achieved by addition of [Cu(MeCN)4](PF6), which produced a 1H NMR spectrum with the Hbpy and HAr signals characteristic to a Cu+-bis(bpy) in an overall achiral structure. The guest-free catenane was regenerated again by adding [(Me4N)(CN)] that competitively extracted the Cu+ ion from the catenane, which could then switch on its sulfate binding ability and gave the chiral sulfate chelate after the introduction of a second batch of TBA2SO4. The successful use of four ionic guests of different charges, geometry, and properties to control a multitopic host to give two different stereochemical outcomes, by virtue of the exceptional co-conformational flexibility of the mechanical bond, hence highlights the unique potential of exploiting the topologically non-trivial catenane as a receptor skeleton in the design and development of diverse classes of smart and responsive molecular hosts. Of note, although ion-controlled multi-state switching in MIM-based molecular switches has been described, the binding cavities in these examples are usually pre-designed with known binding and structural features favorable for the specific guests57,58,59,60,61,62. Our discovery that catenane C can selectively form a strong mechanical chelate with sulfate in an unexpected, chiral co-conformation thus demonstrates the distinctive advantage of combining rigid macrocycles and a co-conformationally flexible, mechanically interlocked molecular framework as next-generation receptors with the ability to undergo facile host–guest optimization while maintaining a high degree of preorganization.

Fig. 7: Reversible achiral-chiral switching of C.
figure 7

Upper: Schematic illustration of co-conformations of C switching between achiral and chiral by sequential addition and removal of sulfate and Cu(I) guests. Bottom: Partial 1H NMR (500 MHz, DMSO-d6, 298 K) spectra of C and species obtained after additions of different salts. Signals from Hbpy and Hdpm are highlighted in blue and pink respectively. Diastereotopic splitting of H1 and H2 in the sulfate complex is highlighted in orange.

In summary, a novel and efficient approach to prepare heteroditopic [2]catenane has been developed, exploiting the metal cation template-directed strategy and dynamic urea formation as ring-closing reaction. Extensive binding studies revealed that in addition to encapsulating the spherical Cu+ cation at the core via mechanical chelation of the two bpy ligands in an overall 1:1 stoichiometry, the [2]catenane is also capable of strong and selective binding to the tetrahedral SO42− anion at the periphery mediated by urea hydrogen bonding formation in a 1:2 host–guest fashion. 1H NMR analysis and X-ray crystallographic study showed the interlocked macrocycles are related by a “180°-circumrotation” in the Cu(I) and sulfate complexes, and saliently, whilst the former remains achiral, the catenane·sulfate complex is chiral with a tilted orientation of the macrocyclic planes. These findings not only highlight the advantages of using co-conformationally flexible catenane hosts for facile adaption and rearrangement of binding motifs to accommodate guests of various structural and electronic features, but also demonstrate a unique example in which the realms of dynamic mechanostereochemistry and host–guest chemistry characteristic to MIMs intersect, where new and unexplored opportunities may arise.

Methods

Full experimental details and characterization of compounds can be found in the Supplementary Information.

Materials

All reagents (purity ≥98% unless otherwise noted) were purchased from commercial suppliers (J&K, Sigma-Aldrich, TCI, Merck, Aladdin, Energy, Macklin, and Dkmchem) and used without further purification unless otherwise noted. All the solvents were of analytical grade (ACI Labscan and DUKSAN Pure Chemicals).

General procedures for [2]catenane synthesis

Under an argon atmosphere, a mixture of bis-t-butylamine 1 (1 eq.) and [Cu(MeCN)4](PF6) (0.5 eq.) or bis-t-butylamine 1 (1 eq.), macrocycle 3 (1.1 eq.) and [Cu(MeCN)4](PF6) (1.1 eq.) in MeCN/CHCl3 (v/v 1:1) was heated at 50 °C for 1 h to 3 h. The red reaction mixture was cooled to room temperature and filtered to remove insoluble materials. Solvents were removed by a rotary evaporator to produce a red solid. The red solid was re-dissolved in dry 1,2-dichloroethane under an argon atmosphere and the corresponding diisocyanate (1.1–1.3 eq.) was added. The reaction mixture was heated at 50 °C for overnight and cooled to room temperature. The product mixture was filtered and quickly added trifluoroacetic acid (2 mL), and the mixture was stirred for 6 h at room temperature. Solvents were removed by a rotary evaporator to afford the Cu(I)-catenane, which was re-dissolved in MeOH/CHCl3 (v/v 1:1) for recrystallization or direct demetallation.

To obtain the metal-free [2]catenane, a solution of the Cu(I)-catenane in 10 mL MeOH/CHCl3 (v/v 1:1) or CH2Cl2 was added 2 mL of neat ethylenediamine (30 mmol), and the mixture was stirred for 5 min, followed by addition of CHCl3 or CH2Cl2 (3 mL) and water (3 mL). The yellow organic phase was separated, and another aliquot of ethylenediamine (1 mL, 15 mmol) was added and stirred for another 5 min to give a colorless solution. Additional CHCl3 or CH2Cl2 (3 mL) and water (3 mL) were added, and the organic phases were combined, washed by water (2 × 5 mL) and dried over Na2SO4. The solvent was removed using a rotary evaporator and the residue was purified by solvent washing or chromatography.

Computational details

Empirical structures of the 1:1 and 1:2 catenane complexes with anions that showed significant binding (i.e. SO42–, H2PO4, CH3COO and Cl) were generated using the Avogadro software63, and refined using density functional theory (DFT) calculations performed with Gaussian 16 (revision A.03)64. Geometry optimizations and single-point energy calculations were carried out using the spin-restricted hybrid density functional B3LYP65,66,67, and the 6–31 G (d,p) basis set was employed for all atoms. The solvation effects of DMSO (ε = 46.826) were taken into account using the conductor-like polarizable continuum model (CPCM)68,69. During the structural optimization step, a convergence criterion of 10–8 for the density matrix was applied. Vibrational analyses were conducted to validate the optimized structures.