Introduction

Interpenetration is a natural phenomenon that occurs on a wide range of biomacromolecules, such as DNA, RNA and proteins1,2,3,4. Such a phenomenon is also frequently encountered in various artificial assemblies, e.g., some interpenetrating molecules/frameworks, including knots, catenanes, rotaxanes, cages, metal-organic frameworks and so on5,6,7,8,9,10, which have been synthesized and extensively studied. Among these, interlocked cages have received increasing attention in recent years due to their unique topological structures and the synthetic challenge they pose11,12,13,14. Generally, interlocked cages are constructed through the dimerization of monomeric cages, in which an enthalpy gain is required, and the dimerization process can be driven by many factors15,16,17. Interlocked cages have often been found to be thermodynamically more stable than their constituent monomeric cages13,18,19,20,21,22,23. Indeed, the self-assembly of interlocked cages usually involves the initial formation of a semi-stable intermediate monomeric cage which is then transformed into the thermodynamically preferable interlocked dimer upon some external stimuli, such as heat15,24,25,26, solvents13,14,27,28, guests15,24,29, and so on. Therefore, the successful isolation and full characterization of both the semi-stable monomeric cage and its interlocked dimer is very challenging. In addition, a comprehensive understanding of the factors and mechanisms underlying the interconversion is of great importance but lacking15,24,28.

Coordination cages with a Pd2L4 composition consisting of four concave pyridine-based bis-monodentate ligands and two square-planar-coordinated Pd(II) ions have been intensively studied30,31,32,33,34. Interestingly, some of these Pd2L4-type coordination cages tend to dimerize to form interlocked dimers with the general formula Pd4L828,35. As a result of the interlocked topology, such Pd4L8-type cages possess three cavities and are capable of binding small molecules or ions as guests36,37. The tendency to form interlocked dimers was demonstrated to be governed by many factors, such as the length38, the bending and steric hindrance of the ligands24,25, the type of solvent13,14,27,28, and the presence of guest anions15,24,29. Especially solvent effects are found to be key contributors to the interpenetration process, but the underlying mechanism remains elusive. In most related studies, it is widely assumed that small solvents like acetonitrile can facilitate the formation of interlocked cages because the release of these small solvent molecules from the cavity of the monomers induces a significant entropic gain to drive cage dimerization24,27. Moreover, the anion template effect is another main contributor to influence the self-assembly process39,40, i.e., the interconversion between monomeric cages and interlocked dimers. It was shown that this process is extremely sensitive to the type and size of the guest anions, and only small anionic guests can fit inside the small cavities of interlocked cages and, therefore induce the dimerization process by both enthalpic and entropic factors15,18,41. For example, Sekiya and Kuroda reported the formation and degradation of an interlocked cage induced by the anion template effect and eventually succeeded in the separation and single-crystal X-ray diffraction characterization of a monomeric cage and its interlocked dimer with different guest anions, including smaller NO3– and bigger 2-naphthalenesulfonate (ONs–), respectively15. Notably, any attempts to crystallize a monomeric cage and its interlocked dimer with the same guest anion (NO3–) were unsuccessful because the monomeric one was always gradually converted into the thermodynamically favored interlocked dimer during crystallization15,26. In addition, halide anions are widely employed to adjust the self-assembly process and trigger the conversion from monomeric cages to interlocked dimers29,42. Particularly, some Pd2L4-type cages and Pd4L8-type interlocked dimers have been demonstrated to show strong chloride binding affinity, which can even capture chloride ions from very insoluble AgCl and facilitate the cleavage of C–Cl bonds, indicating their great potential in the field of anion-binding catalysis43,44,45,46.

On the basis of the observations discussed above, we herein report the self-assembly, interlocking and interconversion of phenoxazine-based coordination cage (Pd2L4)4+ and its interlocked dimer (Pd4L8)8+ (Fig. 1). Through the rational molecular design, we successfully achieve the sole formation, isolation and single-crystal X-ray structure analysis of both individual monomeric cage 1 and its dimer 2 with the same guest anion by simply changing the solvents utilized during the self-assembly. The solvent effects on the interlocking and interconversion of the two coordination cages are systematically investigated. Our results demonstrate that the formation of monomeric cage 1 and its interlocked dimer 2 is highly dependent on the type of the solvents, and we herein assume a correlation to the coordinating capability of the used solvents. That is, 1 is more thermodynamically favored in a strongly coordinating solvent environment, while the formation of 2 is more thermodynamically favored in a moderately and weakly coordinating solvent environment. Furthermore, the impact of the guest anions on the monomeric-to-dimeric cage transformation is also surveyed. It is found that chloride ion can significantly facilitate the interpenetration process due to the strong affinity of the interlocked cage towards the chloride anion, i.e., the suitable size and charge density of Cl– can promote dimerization. As a consequence, the interlocked cage can readily promote C–Cl bond cleavage, facilitating in situ formation of trityl carbocations, which can be applied to catalyze the Diels–Alder reaction and Meinwald rearrangement of epoxides.

Fig. 1: The main content of this work.
figure 1

Schematic representation of interconversion between phenoxazine-based Pd2L4-type monomeric cage and Pd4L8-type interlocked dimer induced by solvents and chloride anions, respectively. Also shown are carbocation-mediated reactions catalyzed by in situ generation of a trityl cation via chloride abstraction using the interlocked cage.

Full size image

Results

Molecular design, self-assembly, and X-ray crystallographic analysis

Previous study showed that heating a mixture of n-hexyl-phenothiazine-based ditopic ligand and [Pd(CH3CN)4](BF4)2 in acetonitrile afforded a quantitative formation of an interlocked Pd4L8-type double cage38,47, while only a small amount of a semi-stable Pd2L4-type monomeric cage was observed in the self-assembly process at room temperature over a short time (Supplementary Fig. 24). When phenothiazine was replaced by phenoxazine skeleton, the in situ NMR spectra indicated the increase in the generation of semi-stable Pd2L4-type monomeric cage intermediate which was finally converted to the interlocked double cage (Supplementary Fig. 25). Interestingly, when the solvent environment was changed to DMSO, monomeric cage became the main product (a small amount of interlocked cage was also formed) in the case of phenoxazine-based ligand while the assembled product became complicated in the case of phenothiazine-based ligand (Supplementary Fig. 26). These pre-experiments indicated the ditopic ligand based on phenothiazine and phenoxazine skeleton resulted in different self-assembly outcomes, especially in DMSO solvent. This finding inspired us to explore the possibility of the solvent-controlled assembly of Pd2L4- and Pd4L8-type and investigation of their interconversion process. Considering that n-hexyl substituted phenoxazine-based coordination cage exhibited relatively poor solubility (Supplementary Fig. 25), we replaced the n-hexyl with a bulky 3,5-di-tert-butyl-4-methoxyphenyl substituent based on our recent works48,49,50, which can significantly improve the solubility of the coordination cages in both polar and nonpolar solvents and facilitate the structural characterization and investigation of the solvent influence on the interconversion process. The bulky aryl substituted phenoxazine-based ditopic ligand (L) was synthesized in three steps (see Section 2 of Supplementary Information). Then, with L in hand, the self-assembly was carried out firstly in acetonitrile at 70 °C for 8 h with 2:1 molar ratio of L and [Pd(CH3CN)4](BF4)2 (Fig. 2a). The in-situ 1H NMR spectroscopy clearly indicated the formation of a monomeric cage as a dominant product at the initial stage which could be eventually fully converted into interlocked dimer 2, as evidenced by a distinct 1H NMR signal splitting into two sets of equal intensity (Fig. 2a, b). Notably, this spontaneous interlocking observed during self-assembly in acetonitrile made it impossible to isolate the pure monomeric cage 1, indicating it was likely to be a semi-stable intermediate in this solvent24. To our delight, when acetonitrile was replaced with dimethylsulfoxide (DMSO), monomeric cage 1 rather than interlocked dimer 2 was formed as the final dominant assembled product as clearly indicated by 1H NMR spectroscopy (Fig. 2c and Supplementary Figs. 8−11)51,52. The resultant assembly 1 could be isolated cleanly by precipitation using diethyl ether in a nearly quantitative yield. Notably, in most studies, one can only isolate Pd2L4-type monomeric cage or Pd4L8-type interlocked cage, while the realization of the sole formation, isolation, and full characterization of both monomeric cage and its dimer, particularly with the same guest anion, is very rare in this field.

Fig. 2: Synthesis and characterization of Pd2L4- and Pd4L8-type cages.
figure 2

a Schematic representation of the self-assembly of phenoxazine-based monomeric cage 1 and interlocked cage 2 in DMSO and acetonitrile, respectively. b Partial 1H NMR spectra (400 MHz, acetonitrile-d3, 298 K) of L and self-assembly of L and [Pd(CH3CN)4](BF4)2 in acetonitrile at 70 °C for different time. c Partial 1H NMR spectra (400 MHz, acetonitrile-d3, 298 K) of ligand L, cage 1 and cage 2, and their 2D DOSY NMR spectra (500 MHz, acetonitrile-d3, 298 K). Color codes of protons ortho to the nitrogen atom of pyridine in 1H NMR spectra: L, green; cage 1, red; cage 2, blue. d Side view and top view of the X-ray crystal structures of cage 1. e X-ray crystal structure of interlocked cage 2 and the side and top views of its constituent monomeric sub-units. Hydrogen atoms, counter anions out of the cage, and 3,5-di-tert-butyl-4-methoxyphenyl substituents are omitted for clarity.

Full size image

Single crystals of monomeric cage 1 and interlocked cage 2 suitable for X-ray structure determination were grown by slow diffusion of diethyl ether into acetonitrile solution (Fig. 2d, e). The structure revealed that monomeric cage 1 displayed a D4-symmetric lantern-like conformation with two BF4– locating in its cavity. The interlocked topology of 2 was clearly indicated in its X-ray crystal structure, and 2 features three internal cavities occupied by three BF4– anions. Interestingly, the two monomeric cages in 2, especially their conformations, differ significantly wherein one cage resembles the structure of 1 but the other one is seriously distorted. The dynamically deformed structures in 2 may give an implication for the reversible assembly/disassembly process and interconversion mechanism (vide infra). Besides, another noticeable distinction between 1 and 2 was the different anion template effect displayed by guest anions. BF4– anions in cages 1 and 2 were surrounded by inner protons and Pd2+ ions, wherein the average C–H⋯F distances in three pockets of 2 were 2.40, 2.52 and 2.43 Å, respectively (Table 1), which were shorter than the corresponding distance in cage 1 (2.56 Å). Similarly, the Pd⋯F distances in three pockets of 2 were also shorter than in 1 (Table 1). The shorter distances between the anions and surrounding atoms also led to the extremely high cavity occupancy in cage 2 (Table 1). These differences seem to indicate that 2 features the stronger C–H⋯F anion hydrogen binding interactions and Pd⋯F electrostatic interactions compared to 1, i.e., BF4– in cage 2 showed a more potent anion template effect than in cage 1, making 2 thermodynamically favored in weakly and moderately coordinating solvents. The above crystallographic analysis revealed that there might be a “trade-off” between the ligand and anions, that only in specific cases both monomeric and interlocked cage can be stabilized by appropriate anion template effect.

Table 1 Summary of CH⋯F, Pd⋯F distances, the volume and occupancy of cavities in cages 1 and 2
Full size table

Monomeric-to-dimeric cage conversion

Since the self-assembly carried out in acetonitrile resulted in the interlocked cage via a semi-stable monomeric intermediate, one can expect that the pure monomeric cage formed in DMSO should be able to convert into the dimeric cage under specific conditions. As expected, 1 could be converted into 2 when it was dissolved in acetonitrile and heated, and the kinetics of the conversion were investigated. First, the effect of different concentrations on monomeric-to-dimeric cage conversion was surveyed by in-situ 1H NMR spectroscopy in acetonitrile-d3 at 70 °C (Supplementary Figs. 42−45). 1H NMR experiments revealed that 1 could be fully converted into 2 after 9 hours at a higher concentration of 4.20 × 10-3 M (Fig. 3a). The structural conversion slowed down at decreased temperatures (Fig. 3b, c, Supplementary Figs. 43 and 47−49). Notably, 1 remained intact in acetonitrile or during crystallization at room temperature, indicating a high temperature is needed to overcome the activation energy for this conversion, which is different to many studies that semi-stable monomeric cages inevitably convert to their dimers during crystallization15,26. The plots of [1]-1 versus time at four different temperatures exhibited a good linear relationship and the half-life of cage 1 was inversely proportional to its initial concentration (C0), indicating a second-order reaction (Supplementary Fig. 50). This finding is consistent with many monomer-to-dimer conversions, i.e., the conversion proceeds via the dimerization of two monomeric cages15,26,28. The activation energy for the dimerization process was then estimated to be 134.71 kJ mol-1 by the Arrhenius plot at 298 K (Supplementary Fig. 51). Furthermore, the resultant rate constants derived from the four different temperatures were used to construct an Eyring plot, which gave the following thermodynamic parameters for conversion of 1 to 2: ΔH‡ = 132.23 kJ mol-1, ΔS‡ = 121.30 J mol-1 K-1 and ΔG‡ = 96.08 kJ mol-1 (298 K) (Fig. 3d). The observed monomeric-to-dimeric cage conversion indicates 2 is more thermodynamically favored than 1 in acetonitrile, which is consistent with the prevailing perception that the formation of interlocked cage is usually a thermodynamically favorable process in system while the monomeric cage is a semi-stable intermediate13,18,19,20,21,22,23,25.

Fig. 3: Monomeric-to-dimeric cage conversion in different temperatures.
figure 3

a Partial 1H NMR spectra (300 MHz, acetonitrile-d3, 298 K) of cage 1 (4.20 × 10-3 M) before and after heating in acetonitrile at 70 °C over time. Color codes of protons ortho to the nitrogen atom of pyridine: cage 1, red; cage 2, blue. b Partial 1H NMR spectra (300 MHz, acetonitrile-d3, 298 K) of cage 1 (4.20 × 10-3 M) before and after heating in acetonitrile at 40 °C over time. Color codes of protons ortho to the nitrogen atom of pyridine: cage 1, red; cage 2, blue. c The fraction of cage 2 after cage 1 (4.20 × 10-3M) is heated at different temperatures over time. The fraction of cage 2 can be determined by comparing respective integration of the protons ortho to the nitrogen atom of pyridine to that of the sum of integration values of the pyridine ortho protons of cage 1 and cage 2. d ln(k/T) over 1/T Eyring plot for the interconversion from cage 1 to cage 2 at variable temperatures. The obtained activation parameters are given in the graph.

Full size image

Dimeric-to-monomeric cage conversion

Considering that the formation of dimeric cage 2 in acetonitrile is a thermodynamically favorable process, we wondered if it is possible to achieve the dimeric-to-monomeric cage conversion to generate cage 1. To our delight, it was found that when the DMSO solution of dimeric cage 2 with a concentration ranging from 1.0 mg/1.0 mL to 9.0 mg/1.0 mL was heated, (partial) disassembly of 2 occurred over time, accompanied by the generation of monomeric cage 1 and ligand L (Fig. 4a, Supplementary Figs. 52, 54, 56, 58, and 60). Specifically, in-situ 1H NMR spectroscopy of a dilute solution of cage 2 (1.0 mg in 1.0 mL DMSO-d6) maintained at 50 °C indicated that the dimeric cage 2 underwent disassembly in a manner of the formation of monomeric cage 1 and ligand L, surprisingly in a 1:1 content ratio within 12 hours (Supplementary Fig. 62). Such phenomenon was also observed in the case of concentrations at 2.0 mg/1.0 mL and 3.0 mg/1.0 mL that dimeric cage 2 initially disassembled in a manner of splitting into one monomeric cage 1 and four ligand constituents L (1:1 content ratio) (Supplementary Figs. 63 and 64). These results appear to suggest that the disassembly process of interlocked cage 2 is most likely to involve the partial disengagement of the metal-pyridine bonds to release one monomeric cage 1 and four ligands L. In contrast, dimeric cage 2 in DMSO-d6 solution with a high concentration was found to disassembly into monomeric cage 1 and ligand L in a ratio greater than 1:1, wherein the ratios of the residual dimeric cage 2 and the formed monomeric cage 1 and ligand L were highly dependent on the initial concentration of 2, implying that partial of the dissociated free ligands were further reassembled into 1 (Supplementary Figs. 52, 54, 56, 58, and 60). The dimeric-to-monomeric cage conversion ratio as high as 62% could be achieved when the initial concentration of 2 in DMSO was 6 mg/mL (Fig. 4b, Supplementary Fig. 60 and Supplementary Table 4). The successful conversion might be related to the different stability of cages 1 and 2 in DMSO. Then, the stability of 1 and 2 in DMSO was thus surveyed and compared, and the results indicated that dimeric cage 2 was more likely to disassemble than 1 at the same mass concentration in DMSO solution (Fig. 4 and Supplementary Figs. 52−61). The above observations imply that dimeric cage 2 seems to be more vulnerable than monomeric cage 1 in DMSO environment, which may be associated with its highly dynamically deformed structure that is vulnerable to strongly coordinating solvent like DMSO.

Fig. 4: Dimeric-to-monomeric cage conversion in DMSO.
figure 4

a Time-course 1H NMR spectra (300 MHz, DMSO-d6, 298 K) of cage 1 (6.0 mg/mL) in DMSO-d6 at 50 °C. b Time-course 1H NMR spectra (300 MHz, DMSO-d6, 298 K) of cage 2 (6.0 mg/mL) in DMSO-d6 at 50 °C. Color codes of protons ortho to the nitrogen atom of pyridine: cage 1, red; cage 2, blue; L, green. c The remaining cage percentages of 1 and 2 in DMSO-d6 over time compared to their respective initial concentrations.

Full size image

In addition, the dimeric-to-monomeric cage conversion could also be achieved through the addition/removal of suitable competing ligand. N,N’-dimethylaminopyridine (DMAP) as a competitive ligand was utilized to compete for coordination with Pd(II) and disassemble the interlocked cage53,54,55. 1H NMR titration experiments in DMSO-d6 revealed that with the addition of DMAP (from 0–8.0 equiv.), the resonances belonging to cage 2 disappeared, while two new sets of peaks appeared, which were belonging to L and [Pd(DMAP)4](BF4)2, respectively (Fig. 5b-e, Supplementary Fig. 66). After the disassembly finished, p-toluenesulfonic acid (TsOH, 8.0 equiv.) was added to protonate DMAP and release the Pd(II) from [Pd(DMAP)4]2+ (Fig. 5f). Evidently, the 1H NMR signal peaks belonging to cage 1 appeared and enhanced upon the addition of TsOH, signifying the protonation of DMAP within [Pd(DMAP)4]2+, simultaneously triggering the reassembly process (Fig. 5f, g). Notably, TsO- anion could be observed in the single crystal structure of the reassembled cage 1 (Fig. 5a, Supplementary Fig. 41). Therefore, we successfully achieve a full dimeric-to-monomeric cage conversion by disassembly-reassembly involving the addition/removal of competing DMAP ligand.

Fig. 5: Dimeric-to-monomeric cage conversion induced by DMAP and TsOH.
figure 5

a Schematic diagram of the disassembly-reassembly process through the addition/removal of DMAP. Partial 1H NMR spectra (300 MHz, DMSO-d6, 298 K) of b [Pd(DMAP)4](BF4)2, c cage 2, d cage 2 after adding 8.0 eq. DMAP, e L, f cage 1 formed after adding 8.0 eq. TsOH to the solution of d, g cage 1 after adding 8.0 eq. TsOH.

Full size image

Solvent effect on self-assembly and interconversion

Both 1 and 2 have good solubility in various polar and non-polar solvents, e.g., acetonitrile (MeCN), DMSO, N,N’-dimethylformamide (DMF), N,N’-dimethylacetamide (DMAC), methanol (MeOH), acetone, tetrahydrofuran (THF), chloroform (CHCl3), dichloromethane (DCM), tetrachloroethane (TCE) and so on, mainly due to the introduction of the bulky aryl substituents into these cationic cages. It should be noted that the good solubility of the two assemblies makes it possible to investigate the effect of different solvents on self-assembly and interconversion comprehensively. It was found that the self-assembly of L and [Pd(CH3CN)4](BF4)2 could also occur in other nitrile solvents with bigger size than acetonitrile, such as isobutyronitrile (IBN), butyronitrile (BuCN) and benzonitrile (PhCN) (Supplementary Figs. 69−76). Specifically, the ligand and the metal source were dissolved in these bulky nitriles, heated and the assembly products were precipitated by the addition of diethyl ether (as in situ NMR monitoring was not feasible in these non-deuterated solvents). Subsequently, the products were solubilized in cold acetonitrile-d3 and immediately subjected to NMR analysis. One may argue that the corresponding NMR results may not directly reflect the assembly outcome in the bulky nitriles. In order to rule out the interference of acetonitrile, the 1H NMR spectra of the precipitates were also recorded in non-coordinating CD2Cl2 in which assembly processes would be extremely slowed down. Indeed, the NMR spectra recorded in both CD3CN and CD2Cl2 furnished the same results, showing that self-assembly in the bulky nitrile solvents proceeded through the initial formation of a semi-stable intermediate monomeric cage 1 and final formation of interlocked cage 2 (Supplementary Figs. 69−74), similar to the result of the self-assembly in acetonitrile. Interestingly, the interpenetration reaction was generally unavoidable regarding the self-assembly in acetonitrile, IBN and BuCN (Supplementary Figs. 67−72), while interpenetration can be well controlled in PhCN, resulting in the formation of pure monomeric cage 1 as the final product under low concentration conditions and short reaction times (Fig. 6a, Supplementary Figs. 75 and 76). In addition, monomeric cage 1 can be completely transformed into 2 upon heating in these bulky nitrile solvents (Fig. 6a, Supplementary Figs. 82−84). Unlike in nitrile solvents, when L and [Pd(CH3CN)4](BF4)2 were mixed and heated in deuterated THF, CHCl3, DCM or TCE, self-assembly was not observed (Supplementary Fig. 79), while the assembled product was complicated and not clear in CD3OD, containing cage 2 and other by-products (Supplementary Fig. 80). However, monomeric-to-dimeric cage conversion can proceed by properly heating solutions of 1 in these solvents (Supplementary Figs. 86−90). Notably, the self-assembly carried out in acetone solution also affords 2, but the reaction rate is slower than in nitrile solvents (Supplementary Fig. 85)56. To our surprise, besides DMSO, the self-assembly of monomeric cage 1 could also be achieved when utilizing DMF or DMAC as a solvent (Supplementary Figs. 77 and 78), meanwhile cage 1 could not be transformed into 2 in these solvents upon heating, either (Supplementary Figs. 92 and 93). Like in DMSO, the stability test of 1 and 2 in DMF confirmed that cage 2 was labile in DMF and partially disassembled into cage 1 and ligand L gradually at 323 K, after one month (Supplementary Fig. 95). In contrast, the in situ 1H NMR spectra of cage 1 showed no change over the same period (Supplementary Fig. 94).

Fig. 6: Solvent effect on self-assembly and interconversion.
figure 6

a Partial 1H NMR spectra (400 MHz, acetonitrile-d3, 298 K) of (i) pure cage 1, (ii) pure dimer 2, (iii) L (10 mg/mL) mixed with [Pd(CH3CN)4](BF4)2 in benzonitrile at 70 °C for 8 h, (iv) L (30 mg/mL) mixed with [Pd(CH3CN)4](BF4)2 in benzonitrile at 70 °C for 12 h, v) cage 1 (10 mg/mL) after heated in benzonitrile at 70 °C for 9 h. b Schematic representation of the results and explanation of the solvent effect on the self-assembly and interconversion of cage 1 and cage 2 from the perspective of coordination ability of solvents.

Full size image

These results imply that the solvent effect on self-assembly and interconversion seems to be very complicated and elusive. The sizes of the solvents and the induced entropic gain upon solvent release may not be the only decisive factor for the formation of monomeric or interlocked structures. Considering that solvent environment, especially their coordinating ability, is generally necessary for the success of the self-assembly, mainly because they can promote solvent and ligand exchange through associative mechanisms by their transient coordination to the Pd(II) centers which allows access to higher-order species through error checking57,58. On this basis, we infer that the observed different outcomes of the self-assembly and interconversion may be related to the coordinating capabilities of solvents and the distinct thermodynamic stability/favourability of two assemblies in different solvents. Specifically, the assembly reaction cannot occur in THF, CHCl3, DCM or TCE, because these solvents generally have very weak coordination ability (Fig. 6b). Besides, these solvents cannot well solubilize the used Pd(II) salt or may even reduce the Pd(II) cations. In contrast, the assembly carried out in solvents such as nitriles, DMSO, DMF, and DMAC with moderately to strongly coordinating abilities can occur successfully. Interlocked cage 2 is likely to be more thermodynamically favored in weak and moderate coordination solvents, and thus the monomeric-to-dimeric cage conversion could proceed in these solvents. Meanwhile, interlocked cage 2 is increasingly vulnerable to be attacked by strongly coordinating solvents, resulting in the disassembly of the interlocked architecture to release monomeric cage 1. This may explain why the dominant assembled product in DMSO, DMF, and DMAC is monomeric cage 1. Therefore, the coordinating ability of solvents is crucial in determining whether the Pd2L4- or Pd4L8-type cage represents the thermodynamic minimum. However, the factors governing the self-assembly and interconversion of these cages are complex and multifaceted, involving both entropic and enthalpic contributions that depend on the behavior of confined versus free solvent, which can vary with the solvent type59. In addition, the size and conformation of solvents should not be overlooked. As a result, fully understanding the mechanism behind this process remains challenging.

Anion template effect on interconversion

Because guest anions usually play an important role in determining the outcome of the coordination cages regarding the sizes, shapes, and host-guest properties, anion template effect on interconversion of three anions (PF6–, BF4– and Cl–) of different sizes was then surveyed. It was found that the monomeric-to-dimeric cage conversion was significantly suppressed when the bigger sized PF6– was added into monomeric cage 1, resulting in a 70% conversion ratio after heating at 70 °C for 8.0 hours while 1 could be fully converted into interlocked cage 2 under the same conditions without adding PF6– (Supplementary Fig. 96). 1H NMR titration experiments revealed the presence of host-guest interaction between monomeric cage 1 and PF6–, i.e., the two encapsulated BF4– anions in 1 could be replaced by up to two PF6– anions to form 2PF6@1 (Fig. 7a, b and Supplementary Fig. 97). As expected, single crystal of 2PF6@1 showed that two PF6– ions were located inside the cavity of cage 1. The Pd⋯F distance and CH⋯F distance in 2PF6@1 were determined to be 3.19 Å and 2.53 Å (Supplementary Table 1), respectively, both slightly shorter compared to those in cage 1. Besides, the occupancy of the cavity in 2PF6@1 was 49.34%, which was significantly higher than in cage 1 (Supplementary Table 1)60. Thus, the bigger size of two PF6– ions together with its more potent anion template effect may account for the inhibited interpenetration process. In contrast to the situation of adding PF6–, it was found that the addition of Cl– anion could significantly facilitate the interpenetration of monomeric cage 1, leading to very fast and efficient monomeric-to-dimeric cage conversion even under ambient temperature without heating (Fig. 7a, c). These results imply that the interconversion between monomeric cage and interlocked dimer is extremely sensitive to the size of the guest anions, and only small anionic guests can fit inside the cavity of interlocked cages and therefore facilitate the dimerization process. The tendency to form interlocked dimer in the presence of Cl– anion indicated that interlocked cage 2 itself may have stronger binding affinity to Cl– than to BF4–24,29,36,42. 1H NMR titration experiments clearly showed the signals for 2 gradually diminished during the titration and were replaced by new signals associated with the formation of the host-guest complex (Supplementary Fig. 98), with BF4– and Cl– guest anions being in slow exchange on the NMR timescale, indicating very strong chloride binding affinity of the interlocked cage. After careful analysis of the 1H NMR titration results (Supplementary Fig. 98), we found the slow-exchange conversion to the new species was already completed after adding 2 equivalent Cl–, and after that only a fast-exchanging process by gradual signal shifting was observed. This result suggests that in solution state the two outer BF4– anions are replaced by Cl– firstly (slow exchange), but the host-guest interaction between the middle pocket and the third Cl– is in a fast exchange process, resulting in an equilibrium between 2Cl@2 and 3Cl@229. Above conclusion was also supported by 19F NMR (Supplementary Fig. 99) and ESI-TOF-MS spectra (Supplementary Fig. 35) that 2Cl@2 and 3Cl@2 species may co-exist in the solutions in the presence of 4 equivalent Cl–. Interestingly, the X-ray crystallographic analysis explicitly indicated that three BF4– anions in the pockets of interlocked cage 2 were fully displaced by three Cl– anions, resulting in 3Cl@2 (Fig. 7a)29. Compared to 2, 3Cl@2 exhibited a reduction of the two outside pockets to 28.31 Å3 and 28.33 Å3, respectively, while the middle pocket volume was expanded from 47.74 ų to 51.01 ų (Supplementary Table 1). Besides, the occupancy of all cavities in 3Cl@2 decreased significantly as evidenced by the longer CH⋯Cl and Pd⋯Cl distances (Supplementary Table 1). It was worth noting that Cl– in middle pocket of 3Cl@2 was disordered, since Cl– oscillated rapidly between two Pd centers (Supplementary Fig. 40). Moreover, 3Cl@2 featured a more distorted structure than cage 2 (Supplementary Fig. 40), allowing the smaller Cl– to be tightly sandwiched between the four dicationic palladium centers through Coulomb interactions. Consequently, the stability of 3Cl@2 in DMSO decreased significantly compared to cage 2, e.g., 3Cl@2 was found to disassemble into L immediately after being dissolved in DMSO (Supplementary Fig. 100). This finding may further support the inference that the interlocked cage seems to be labile in strong coordination solvent environment like DMSO.

Fig. 7: Anion template effect on interconversion.
figure 7

a Schematic representation of the results and explanation of the anion template effect on interconversion between monomeric cage 1 and interlocked cage 2, and the X-ray crystal structures of host-guest complexes 2PF6@1 and 3Cl@2. Hydrogen atoms, counter anions out of the cage and 3,5-di-tert-butyl-4-methoxyphenyl substituents are omitted for clarity. b Partial 1H NMR titration spectra (300 MHz, acetonitrile-d3, 298 K) of cage 1 with KPF6. c Partial 1H NMR titration spectra (300 MHz, acetonitrile-d3, 298 K) of cage 1 with tetrabutylammonium chloride (TBAC), and cage 2 with TBAC.

Full size image

Anion-binding catalysis

The strong chloride-binding affinity of interlocked cage 2 inspired us to further explore its potential use for anion-binding catalysis and catalytic chemical activation. It was found that interlocked cage 2 can catalytically break the carbon-chloride bond of diphenylmethyl chloride and triphenylmethyl chloride in acetonitrile in the presence of trace amounts of H2O (as source for the OH group), generating 3Cl@2 and corresponding diphenylmethanol and triphenylmethanol, respectively (Fig. 8a, b, Supplementary Figs. 101 and 104, Supplementary Tables 5 and 6). Interestingly, it was found that the hydrolysis of diphenylmethyl chloride and triphenylmethyl chloride could also be catalyzed by cage 1, while cage 1 was converted into 3Cl@2 (Supplementary Figs. 105 and 106) in this process. The formation of alcohols indicates that this interlocked cage-promoted C–Cl bond cleavage reaction proceeds through an SN1 mechanism, in which the carbocation as an intermediate is formed and gets attacked by water in solvent43,61,62. As a result, the pH of the reaction medium decreased because the reaction produced hydrogen ions (H+) (Supplementary Fig. 102). To further understand the reaction mechanism, dehalogenation reaction was also carried out in dry CD3CN with other conditions remained unchanged. The reaction produced corresponding acetamide, which confirmed the formation of diphenylmethyl carbocation in catalytic process (Supplementary Fig. 103)61,62. Notably, the C–Cl bond cleavage of triphenylmethyl chloride could occur readily at room temperature in the presence of 2 (Fig. 8 and Supplementary Fig. 104, Supplementary Table 6). Thus, this interlocked cage-promoted C–Cl bond cleavage reaction provides a facile route to generate trityl carbocation under mild conditions in-situ which could be further utilized to promote other reactions. For example, Diels–Alder reaction and Meinwald rearrangement of the epoxide are two classic organic reactions catalyzed by carbocation species, but the inherent instability of carbocations has restricted now their use in catalysis with decent performance61,63,64,65. To our delight, the in-situ generated trityl carbocation formed through the interlocked cage-promoted C–Cl bond cleavage of triphenylmethyl chloride (5.0 mol% 2 and 20 mol% triphenylmethyl chloride loading) was adequate to catalyze the Diels–Alder reaction of acrolein with cyclohexadiene as well as the Meinwald rearrangement of diphenyloxirane in 99% conversion and the reactions can be completed at room temperature within 4 hours (Fig. 8c, Supplementary Figs. 107 and 108, Supplementary Tables 7 and 8). The controlled experiments also demonstrated that such Diels–Alder reaction and Meinwald rearrangement cannot occur in the absence of 2 or triphenylmethyl chloride, or when 2 was replaced by L or [Pd(CH3CN)4](BF4)2, or the mixture of [Pd(CH3CN)4](BF4)2 and monotopic ligand L’, suggesting the in-situ generation of trityl carbocation by interlocked cage was crucial to ensure that reactions can take place (Supplementary Figs. 107 and 108, Supplementary Tables 7 and 8).

Fig. 8: Supramolecular anion-binding catalysis.
figure 8

a Schematic representation of the catalytic hydrolysis of diphenylmethyl chloride. Reaction conditions: diphenylmethyl chloride (0.01 mmol), H2O (10 μL), 2 (0.001 mmol, 10 mol%), 60 °C, 40 h, acetonitrile-d3 (0.4 mL), conversion >99%. b Partial 1H NMR spectra (400 MHz, acetonitrile-d3, 298 K) of (i) diphenylmethyl chloride, (ii) the mixture of diphenylmethyl chloride and 2 heated in acetonitrile-d3 at 60 °C for 40 h and (iii) diphenylmethanol. c Diels–Alder reaction and Meinwald rearrangement catalyzed by in situ generation of a trityl cation via chloride abstraction using 2. Reaction conditions of catalytic hydrolysis of triphenylmethyl chloride: triphenylmethyl chloride (0.02 mmol), 2 (0.001 mmol, 5.0 mol%), H2O (10 μL), r.t., immediately, acetonitrile-d3 (0.4 mL), conversion >99%. Reaction conditions of Diels–Alder reaction: acrolein (0.02 mmol), cyclohexadiene (0.03 mmol), triphenylmethyl chloride (0.004 mmol), 2 (0.001 mmol, 5.0 mol%), r.t., 4.0 h, acetonitrile-d3 (0.4 mL), conversion >99%. Reaction conditions of catalytic hydrolysis of Meinwald rearrangement: diphenyloxirane (0.02 mmol), triphenylmethyl chloride (0.004 mmol), 2 (0.001 mmol, 5.0 mol%), r.t., 4.0 h, acetonitrile-d3 (0.4 mL), r.t., 4.0 h, conversion >99%.

Full size image

Discussion

In summary, we have succeeded in the solvent-controlled assembly of phenoxazine-based monomeric cage 1 and interlocked cage 2 with the same guest anion, whose structures were unambiguously confirmed by single crystal X-ray diffraction. The good solubility of the two coordination cages in various solvents allows us to investigate the solvent effects on their self-assembly, interlocking, and interconversion comprehensively. The monomeric-to-dimeric cage conversion could occur in various solvents like nitriles, acetone, THF, chloroform, DCM and TCE, and the conversion is almost quantitative. By comparison, dimeric-to-monomeric cage conversion could be achieved by dilution in DMSO or the addition/removal of competing ligand through a disassembly-reassembly process. We infer that the coordination ability of solvents is one of the decisive factors that affect the self-assembly and interpenetration process. Specifically, the formation of monomeric cage 1 is favored in strongly coordinating solvents like DMSO, DMF, and DMAc, while interlocked dimer 2 is more thermodynamically favored in solvents with moderately and weakly coordinating ability. Besides, anion template effect on interconversion indicates that the interconversion between monomeric cage and interlocked dimer is extremely sensitive to the size of the guest anions, and the small Cl– could significantly facilitate the dimerization process, mainly as a result of the intrinsically strong chloride binding affinity of interlocked cage 2. On this basis, we have demonstrated the potential use of interlocked cage 2 for anion-binding catalysis and catalytic chemical activation of a variety of carbocation-mediated reactions. Therefore, this work provides a much more comprehensive understanding and specific view of the solvent effect on the interpenetration of Pd2L4-type coordination cages and the underlying mechanism. Besides, the interlocked cage described here represents a supramolecular anion-binding catalyst system to mildly drive carbocation-mediated reactions. We are currently investigating the possibility of the synthesis of poly-interlocked cages by manipulation of the solvent and anion template effect on interpenetration.

Methods

All reagents and starting materials were obtained from commercial suppliers and used without further purification. All air-sensitive reactions were carried out under inert N2 atmosphere. The 1H NMR, 13C NMR spectra were recorded in the solution of CDCl3, TCE-d2, THF-d8, CD3OD, CD2Cl2, acetone-d6, acetonitrile-d3, DMF-d7 and DMSO-d6 on Bruker 300 MHz, Bruker 400 MHz, and Bruker 500 MHz spectrometer. 2D NMR (COSY, DOSY) was measured in solution of DMSO-d6 and acetonitrile-d3 on Bruker 400 MHz and Bruker 500 MHz spectrometer. Coupling constants (J) are denoted in Hz and chemical shifts (δ) are denoted in ppm. Multiplicities are denoted as follows: s = singlet, d = doublet, b = broaden, and m = multiplet. Mass spectra were obtained with a Bruker micro TOF-Q II mass spectrometer (Bruker Daltonics Corp., USA), acetonitrile as solvent, and a Waters Synapt G2 mass spectrometer in the electrospray ionization (ESI) mode and electron impact (EI) mode, dichloromethane as solvent. The single crystals of this work were measured on Rigaku XtaLAB PRO MM003-DS dual system with Cu Kα radiation (λ = 1.54184 Å) and BL17B beamline of National Facility for Protein Science in Shanghai (NFPS) at Shanghai Synchrotron Radiation Facility. The Van der Waals volume of the cage cavity was analyzed using Multiwfn.