Synergistic regulation of metal–organic cage architectures via temperature- and solvent-driven atropisomerism

Significance

This research demonstrates an innovative chemical switching system capable of regulating self-assembly in metal–organic cages with precision. Utilizing atropisomers of a macrocycle ligand (CP2), the system responds dynamically to temperature and solvent changes, enabling the formation and reversible transformation of distinct molecular cages. This multistimulus-responsive approach provides valuable insights into designing smart materials with tunable properties, addressing a critical challenge in advanced material science.

Abstract

Regulating multistimulus responses in artificial systems remains a challenge in smart material development. We present a versatile chemical switching system that precisely controls the self-assembly of metal–organic cages via temperature and solvent changes. The key component, cyclo[2](1,3-(4,6-dimethyl)benzene) (4-pyridine)[6](1,3-(4,6-dimethyl)benzene) (CP2), was generated as three atropisomers (1, 2, and 3) with C_s, C₁, and C_2v symmetries. Thermally, metastable isomers (1 and 2) convert into the stable isomer (3), which reacts with Pd²⁺ to form specific molecular cages. Depending on the solvent, either rectangular M₂L₂ cages (5′ and 5) form in 1,4-dioxane or hexagonal M₃L₃ cages (6) in 1,1′,2,2′-tetrachloroethane. The solvent dictates the cage type and enables reversible transformation between cages 5 and 6. Additionally, cage 5′, formed from metastable isomer 1, can switch to other cage types (i.e., 5 or 6) depending on temperature and solvent conditions. This multipathway system offers a precise strategy for controlling self-assembly in smart materials.

Metal–organic cages (MOCs) (1–7) are pivotal in advanced chemistry and materials science due to their unique architectures, versatile functionalities, and potential applications in addressing critical global challenges, including catalysis, clean energy, healthcare, and sustainability (8). These self-assembled structures consist of multiligands (L) coordinated with transition metal ions (M), forming directional and dynamic metal–ligand bonds that endow them with exceptional flexibility (1, 9). This flexibility enables MOCs to mimic the adaptive behavior of biomolecules, which respond dynamically to environmental signals. For instance, enzymes undergo conformational changes through induced complexation processes to align with specific substrates (1). Similarly, MOCs can undergo isomerization, reconfiguring their structures in response to external stimuli to achieve new thermodynamic states and functions, such as opening and closing their cavities.

The study of MOC isomerization provides critical insights into molecular dynamics and structural adaptability while driving innovations in stimuli-responsive materials, catalysis, sensing, and drug delivery (10, 11). Transformations in MOCs can be triggered by various stimuli, including the introduction of new cage components (12–14), guest molecules (14–16), temperature changes (9), solvent composition (17, 18), pH shifts (19), and concentration variations (20–22). While the literature has extensively explored single-stimulus-induced cage-to-cage transitions, reports on multiple stimuli driving complex structural transitions across different metal–organic architectures remain lacking. Investigating such multistimuli transformations represents an exciting frontier, with the potential to expand the functional versatility of MOCs in diverse applications. Especially, orthogonal regulation of multiple stimulus responses has been developed to achieve independent and precise control of complex systems’ responses to different stimuli, enabling sophisticated and efficient functions in self-assemblies and advanced materials. Up to date, most of the reported multifactor regulation systems, such as molecular switches and logic devices, have been generated involving orthogonal control strategy (23–25). The control factors in these systems are mutually independent, meaning they do not influence each other in determining the ultimate state distribution (26–28). However, synergistic regulation of multiple stimulus responses is still challenging to act as an effective regulation strategy in the artificial self-assembly system (29–31). In the latter cases, the interaction of different stimuli leads to a greater and more effective response than the stimuli would produce individually. The example included hierarchical control over pseudorotaxane-based molecular devices, which utilized protonation/deprotonation and photoirradiation synergistically for the kinetics and thermodynamic control corresponding to pseudorotaxane formation (32). Similarly, the synergistic effects of temperature, solvent, and photoirradiation were employed to achieve an additional level of control over the function of molecular machines (33–35). This principle is essential in designing advanced materials, biological systems, and technologies that require amplified or highly efficient responses to complex environmental conditions.

Herein, our study describes a versatile chemical switching system capable of controlling the architectures of self-assembled MOCS through multiple pathways, driven by variations in temperature and solvent (Fig. 1). The key macrocycle building block, cyclo[2](1,3-(4,6-dimethyl)benzene) (4-pyridine)[6](1,3-(4,6- dimethyl)benzene) (CP2), was synthesized with three distinct, rigid atropisomers: 1 (C_s symmetry), 2 (C₁ symmetry), and 3 (C_2v symmetry). These atropisomers undergo both reversible and irreversible interconversions between the metastable isomers 1 and 2, and the thermally stable isomer 3, depending on temperature. Metastable isomers 1 and 2, or the thermally stable isomer 3, can independently react with Pd²⁺ in the presence of 1,4-dioxane to form M₂L₂ rectangular molecular cages, denoted as 5′ and 5, respectively. When 1,4-dioxane is replaced with 1,1′,2,2′-tetrachloroethane, isomer 3 reacts with Pd²⁺ to form a distinct M₃L₃ hexagonal molecular cage (6). Notably, by altering the solvent, it becomes possible to reversibly transform the M₂L₂ cage into the M₃L₃ cage based on isomer 3. Furthermore, the M₂L₂ cage 5’, containing metastable isomer 1, can be transformed into either cage 5 or cage 6 through the cooperative influence of both temperature and solvent. This work lays a significant foundation for developing intelligent material systems with multipurpose chemical switching mechanisms, offering versatile control over self-assembled structures.

Fig. 1.

The temperature-induced interconversion among the three atropisomers enables the selective formation of distinct molecular cages. The metastable isomers 1 and 2, or the thermodynamically stable isomer 3, independently react with Pd²⁺ in the presence of 1,4-dioxane to yield M₂L₂ rectangular molecular cages, 5′ or 5, respectively. In contrast, replacing 1,4-dioxane with 1,1′,2,2′-tetrachloroethane allows isomer 3 to form a unique M₃L₃ hexagonal molecular cage, 6. Remarkably, solvent exchange facilitates a reversible transformation between the M₂L₂ cage 5 and the M₃L₃ cage 6. Furthermore, the M₂L₂ cage 5′ can transition to either cage 5 or cage 6 under the cooperative effects of temperature variation and solvent changes.

Results

Synthesis and Structural Characterization of the Macrocycle.

The macrocycle CP2, namely cyclo[2](1,3-(4,6-dimethyl)benzene)(4-pyridine)[6](1,3-(4,6-dimethyl)benzene), was synthesized using Suzuki–Miyaura coupling cyclization with 7 and 8 as starting materials (Fig. 2A). Three stable isomers of CP2, labeled 1, 2, and 3, were successfully separated with yields of 5%, 10%, and 8%, respectively. In the high-resolution Matrix-Assisted Laser Desorption/Ionization Time of Flight (MADLI-TOF) mass spectra, these isomers display the same molecular weight values (SI Appendix, Figs. S17–S19).

Fig. 2.

Synthesis and ¹H NMR spectroscopic characterizations of macrocycle CP2 atropisomers. (A) Yields correspond to products purified by column chromatography. (B) Corresponding ¹H NMR spectra of compounds 1 (Top), 3 (Middle), and 2 (Bottom), each measured at a concentration of 1 × 10⁻³ M in PhNO₂-d₅ at 298 K (500 MHz).

The NMR spectra of atropisomers 1, 2, and 3 exhibit distinct signals, reflecting their respective symmetries in solution. Isomer 1 adopts C_s symmetry, as evidenced by two distinct pyridine proton signals, six nonaromatic proton signals in the ¹H NMR spectrum (Fig. 2B and SI Appendix, Fig. S1), and 26 carbon signals in the ¹³C NMR spectrum (SI Appendix, Fig. S2). In contrast, isomer 2, with C₁ symmetry, shows a more complex ¹H NMR spectrum, featuring four distinct pyridine proton signals (Fig. 2B and SI Appendix, Fig. S5) and the highest number of 38 carbon signals observed in the ¹³C NMR spectrum (SI Appendix, Fig. S6). Isomer 3, exhibiting C_2v symmetry, displays a simpler spectral pattern with two pyridine proton signals, four nonaromatic proton signals in its ¹H NMR spectrum (Fig. 2B and SI Appendix, Fig. S9), and 21 carbon signals in the ¹³C NMR spectrum (SI Appendix, Fig. S10).

Three atropisomers of CP2 were further confirmed via single-crystal X-ray diffraction analysis of 1·2.5C₄H₈O (tetrahydrofuran), 2·CH₃OH and 3·C₄H₈O (Fig. 3 A–C and SI Appendix, Figs. S23–S28 and Table S2). Each single-crystal example was generated with slow evaporation from the relevant isomer solution (1 mM). Two 4-(2,6-dimethyl phenyl)pyridinyl (P₁-P₂) and six m-xylene (B₁-B₆) fragments alternately construct each macrocycle. The structure of 1 has C_s symmetry and a circular cavity with a diameter of 8.2 Å (Fig. 3A and SI Appendix, Fig. S23). Its symmetry plane passes through opposite asymmetric B moieties (B₁, B₄). The two P moieties are symmetric, and pointing toward the same direction of the macrocycle. Two symmetrical B group sets point to the opposite direction. The torsion angles between the adjacent aromatics are between −74.6(9)° and 104.4(3)° (SI Appendix, Fig. S24). The angle of 24° between two pyridinyl moieties was observed. With the location of adjacent B₂, P₁, and adjacent B₅, P₂ fragments in a sequence switched respectively, the conformation of 1 will become 2 with C₁ symmetry (Fig. 3B and SI Appendix, Fig. S25). Herein, 2 displayed an elliptic cavity with major and minor distances of 8.8 and 7.6 Å, respectively. There is an angle of 49° between pyridinyl groups (Fig. 3B). The torsion angles between neighboring B and P range from −114.8(3)° to 115.0(5)° (SI Appendix, Fig. S26). The angle of 49° between two pyridinyl moieties was observed. In 3· C₄H₈O, two plumb symmetry planes pass through the macrocycle’s C₂ axis (Fig. 3C and SI Appendix, Fig. S27). A double-bowl-shaped cavity is created The cavity depth of 3 was approximately 5.3 Å, and the upper and lower rims of the structure had approximate diameters of 13.0 Å and 8.0 Å (Fig. 3C). Two P groups point to the same side of the macrocycle with their angle as 110°. The four B units (B₁, B₃, B₄, and B₆) can form a pyramid when they all point to a different side of the macrocycle. In this case, the torsion angle between the adjacent planes is between 58.3(9)° and 120.8(6)° degrees (SI Appendix, Fig. S28). The angle of 110° between two pyridinyl moieties was observed. The solution NMR spectroscopic study of CP2 is carried out in PhNO₂-d₅ at 298 K (Fig. 2 and SI Appendix, Figs. S1–S12). Notably, all three atropisomers retain their symmetry in both solution and crystalline states.

Fig. 3.

Single-crystal X-ray diffraction structures of macrocycles (1, 2, and 3). Single-crystal X-ray diffraction structures of macrocycles (1, 2, and 3), highlighting the angles between the P and B moieties and the cavities of 1 (A), 2 (B), and 3 (C). Hydrogen atoms and solvent molecules have been omitted for clarity. Atoms are color-coded: Carbon in silvery gray, and Nitrogen in bluish violet.

Compared to the reported cyclo[8](1,3-(4,6-dimethyl)benzene) (CDMB-8) (36–38), the core macrocycle skeleton of isomers 1 and 2 –excluding the pyridine group– is structurally similar to that of C_s-CDMB-8. In contrast, the skeleton of isomer 3 corresponds to that of D_4v-CDMB-8. Similarly, when compared with the reported cyclo[4] (1,3-(4,6-dimethyl) benzene)[4] (1,3-(4,6-dimethyl) benzene)(4-pyridine) (CP4) (39), the core macrocycle skeletons of isomers 1 and 2 resemble that of CP4, which also exhibits C_s symmetry. On the other hand, the skeleton of isomer 3 mirrors that of C_4v-CP4. These comparisons suggest that the substitution of the pyridine groups has minimal if any, effect on the atropisomeric core of the molecular architecture.

Theoretical calculations using semiempirical methods (PM7) with MOPAC were employed to gain a comprehensive understanding of the varying cyclization yields of the atropisomers corresponding to CP2 (SI Appendix, Fig. S22). An isomer 4 with C_s symmetry has also been taken into account (see infra). Their respective cyclization intermediates containing Pd²⁺ catalytic linkers were evaluated. It was found that the formation heat of 1, 3, or 4 corresponding intermediates (i.e., 1-im, 3-im, or 4-im) is lower than that of the 2 related intermediates (namely 2-im) (SI Appendix, Table S1). However, subsequent studies demonstrated that with 2 as the intermediate, the transformation of 4 and/or 1 results in the most thermodynamically stable 3. Thus, it is proposed that the distribution of end-product yields is influenced by at least two factors: the stability of intermediates and the conversion of atropisomers.

Heat-Driven Conversion between Various Atropisomers.

Macrocycles play a critical role in self-assembled structures and are fundamental to biomolecular machines (40–44). Among them, biphenyl-based macrocycles exhibit multiple atropisomeric forms, which can undergo temperature-driven interconversion. However, precise control over the transformation between three or more stiff atropisomers remains a challenge. In this study, we investigate the temperature-induced interconversion of CP2 atropisomers, focusing on the influence of substituent number on their structures, assembly formation, and thermodynamic properties.

To examine atropisomer transformations, compounds 1 and 2 were heated at 473 K in an argon atmosphere for 1 h, leading to a quantitative conversion to 3 (Fig. 4A and SI Appendix, Fig. S29). Further insights into the interconversion process were obtained via temperature-dependent NMR spectroscopy, which revealed the formation of an intermediate atropisomer 4, exhibiting distinct pyridine (i.e., P unit) and aromatic (namely B unit) proton signals (Fig. 4B and SI Appendix, Fig. S30). Theoretical calculations suggested that 4 arises from the positional exchange of P1/P2 and B1/B4 groups within the macrocycle framework (SI Appendix, Fig. S31). Within 60 s at 473 K, 84% of 1 underwent rapid conversion, yielding 2 (64%), 3 (13%), and 4 (6.6%). The relative concentrations of these atropisomers evolved over time, ultimately reaching equilibrium at a ratio of 0.5:2.3:97:0.2 after 10³ s (Fig. 4C). Notably, kinetic analysis revealed that conversion proceeds via two fast transitions (1 → 4 → 2) followed by a slower transformation (2 → 3) (Fig. 4D). The metastable atropisomer 1 ultimately converted to the thermodynamically stable 3, with 2 and 4 acting as intermediates.

Fig. 4.

The transformation between three atropisomers of CP2 at 473 K. (A) Schematic illustration of the interconversion among 1, 2, and 3 in PhNO₂-d₅ at 473 K. (B) Time-dependent ¹H NMR spectra of 1 (1.00 × 10⁻³ M) in PhNO₂-d₅ at 473 K, showing the progression of the transformation. (C) Time-dependent concentration profiles of 1 (blue dots), 2 (green dots), 3 (red dots), and 4 (light blue dots), fitted with nonlinear curves (solid lines) at 473 K. (D) Potential energy diagram depicting the conversion pathways among 1, 2, 3, and 4 in PhNO₂-d₅ at 473 K, with transition states identified based on theoretical calculations.

The thermodynamic and kinetic parameters of the interconversion were determined using time-dependent ¹H NMR data (SI Appendix, Table S3). Equilibrium constants were calculated as follows:

$$\mathbf{1}\begin{array}{c}{k}_1\\ \rightleftharpoons \\ k_{-1}\end{array}\mathbf{4}\begin{array}{c}{k}_2\\ \rightleftharpoons \\ k_{-2}\end{array}\mathbf{2}\begin{array}{c}{k}_3\\ \rightleftharpoons \\ k_{-3}\end{array}\mathbf{3},$$

$${\mathrm{K}}_1=k_1/k_{-1},{\mathrm{K}}_2=k_2/k_{-2},{\mathrm{K}}_3=k_3/k_{-3},$$ [1]

A potential energy diagram at 473 K was generated using the PM7 method, considering 1 as the reference state (0 kJ mol⁻¹) (Fig. 4D).

At 393 K, the four atropisomers coexisted in equilibrium over extended periods, with interconversion rates slower than at 473 K (SI Appendix, Figs. S35–S38). The investigation of the temperature-induced isomerization of 2 at 393 K showed that its conversion curve closely followed that of isomer 1, suggesting similar behavior under thermal conditions (SI Appendix, Fig. S38). The calculated energy trends were consistent with experimental results, confirming the interconversion pathways (SI Appendix, Fig. S39 and Tables S5).

Compared to CP4, which exhibits only three atropisomers, CP2’s additional atropisomer arises from the altered symmetry induced by pyridine substituents. Interestingly, CP2 exhibits reversible interconversion at 393 K, whereas complete conversion to the thermodynamically stable state occurs at 473 K. The atropisomerism of CP2 was further compared to CDMB-8, a methylated macrocycle undergoing slow conversion at 393 K. Unlike CP2, which undergoes thermal transformation without chemical modifications, other macrocycles (e.g., [n]cyclo-4,10-pyrene, porphyrins, and cuproaromatics) typically require external stimuli such as guest interactions, pH modulation, or chemical modifications (45–53). The ability of CP2 and related macrocycles to regulate atropisomer distributions solely through temperature variation offers a promising avenue for designing adaptive supramolecular systems.

Investigations were conducted on the transformation of 2 in solvents with varying polarities, including o-dichlorobenzene (o-PhCl₂), 1,1,2,2-tetrachloroethane (TCE), N,N-dimethylformamide (DMF), and dimethyl sulfoxide (DMSO). Due to poor solubility of 2 in DMF and DMSO, the studies were limited to o-PhCl₂ and TCE (SI Appendix, Figs. S40–S44). At 393 K, the transformation rate of 2 in o-PhCl₂ was comparable to that in PhNO₂ [k_o-PhCl2 = (1.7 ± 0.1) × 10⁻⁵ s⁻¹, k_PhNO2 = (1.9 ± 0.3) × 10⁻⁵ s⁻¹] (Table 1). However, in TCE solution, the rate was significantly lower [k_TCE = (6.3 ± 0.1) × 10⁻⁶ s⁻¹], suggesting that the transformation rate is influenced by solvent selectivity (Table 1).

Table 1.

Solvent-dependent conversion rate of compound 2 in various solvents

Solvent	Dipole moment (54)	k
Ph*	0 D	—
TCE	1.32 D	(6.3 ± 0.1) × 10−6 s−1
THF*	1.75 D	—
o-PhCl2	2.50 D	(1.7 ± 0.1) × 10−5 s−1
DMF†	3.82 D	—
DMSO†	3.96 D	—
PhNO2	4.22 D	(1.9 ± 0.3) × 10−5 s−1

^*Indicates that the solvent’s boiling point is below 393 K, preventing the reaction from proceeding under the required thermal conditions.

^†Refers to the relatively low solubility of compound 2 in the corresponding solvent.

Construction and Single-Crystal X-ray Diffraction Analysis of MOCs.

The macrocycles were extensively utilized as ligands in metal–organic self-assembly (55–57). Due to the inherent aperture size, multiple binding sites, modifiability, and the modularity of the partial skeleton of the macrocycle building blocks, these self-assemblies have shown great potential for creating novel intelligent architectures. Here, the formation of self-assembled “2+2” metal–organic molecular cages 5 is observed. 102 μL of PdCl₂ (20 mM, acetonitrile solution), 500 μL of 3 (4 mM, 1,4-dioxane solution) were mixed with 1.5 mL of 1,4-dioxane/acetone (2:1, v/v) solution. A large amount of bright yellow prismatic crystals (3@PdCl₂·5.38 C₄H₈O₂·0.5CH₃OH; 5) were obtained after heating for 8 h at 338 K (Fig. 5A). The single-crystal X-ray diffraction analysis showed the crystal has a large unit cell size [a = 12.6445(5) Å, b = 19.8470(6) Å, c = 20.2220(5) Å, α = 80.652(2)^o, β = 82.977(3)^o, γ = 81.399(3)^o, V = 4926.1(3) ų] (SI Appendix, Table S6). Each individual cage exhibited an orthogonal structure (Fig. 5 A and B). It consists of two molecules of 3 and two molecules of PdCl₂. In this case, each Pd²⁺ ion forms a 176° coordination with two pyridine groups, which subsequently serve as a linker connecting the two 3. (Fig. 5B). The angle between two pyridine groups on the same macrocycle is 18°. The significant characteristics of 5 are its enormous orthogon cavity and high porosity (Fig. 5B). The maximum and minimum diameters are 19 and 10 Å, respectively. The cavity of 5 encapsulated three 1,4-dioxane molecules, while each of the two macrocycles contains one 1,4-dioxane molecule outside of 5. The binding between 1,4-dioxane and 5 is stabilized with possible π–O interactions (SI Appendix, Fig. S45A). The stability of 5 in solution was examined using ¹H NMR and ¹H DOSY spectra in CDCl₃. The ¹H NMR spectra indicate that compared to macrocycle 3, the signal peaks corresponding to H_9″, H_10″, and H_11″ of 5 experience a downfield shift, while the signal peaks of H_1″-7″ show an upfield shift. The results confirmed the complexation between 3 and Pd²⁺ (Fig. 5C). The ¹H DOSY NMR analysis provided additional evidence of the production of a single product, which exhibited a single diffusion coefficient (D) value as 3.96 × 10⁻¹⁰ m² s⁻¹ (Fig. 5D and SI Appendix, Fig. S46). According to the cylindrical model (58), the diffusion coefficient (D) can be expressed as follows:

$$D=\frac{k_{\text{BT}}}{3\pi \eta \mathrm{L}}\left\{\text{ln}\left(\frac{\mathrm{L}}{\mathrm{d}}\right)+0.312+0.565{\left(\frac{\mathrm{L}}{\mathrm{d}}\right)}^{-1}-0.100{\left(\frac{\mathrm{L}}{\mathrm{d}}\right)}^{-2}\right\},$$ [2]

Fig. 5.

The assembly process and single-crystal X-ray structural analysis of 5. (A) Schematic representation of the complexation between 3 and Pd²⁺ ions in a 1,4-dioxane/acetone/acetonitrile mixture (15:5:1, v/v/v) at 338 K for 8 h. (B) Single-crystal X-ray diffraction structures of 5 shown from front, side, and top views. Color code: Carbon (silvery gray), Nitrogen (bluish violet), Palladium (orange), and Chlorine (green). Hydrogen atoms and solvent molecules have been omitted for clarity. Dashed line L₃ is parallel to dashed line L_3’. (C) ¹H NMR spectra comparing 5 (Top) and 3 (Bottom). (D) DOSY spectrum of 5 (600 MHz, CDCl₃, 298 K), illustrating its diffusion properties.

where k_B, T, η, L, and d denote the Boltzmann constant, temperature, solvent viscosity, length of the cylinder, and diameter of the cylinder, respectively. Using Eq. 2 and the d value (1 nm) acquired from single-crystal data, we calculated L to be 2 nm. The successful procurement of assembler cage 5 was further proved through the analysis of mass spectrometry data ([M-Cl⁻]⁺ (m/z): calcd. for C₁₄₈H₁₄₀N₄Cl₃Pd₂, 2,293.8252; found, 2,293.8008. SI Appendix, Fig. S20).

When the solvent component 1,4-dioxane is replaced with 1,1′,2,2′-tetrachloroethane, the complexation between 3 and PdCl₂ leads to the formation of distinct metal–organic molecular cages featuring expansive triangular cavities (Fig. 6A). The single-crystal X-ray diffraction investigation revealed that each cage 6 contained a total of three macrocycles 3 and three PdCl₂ units. In this case, each Pd²⁺ ion establishes a 168° coordination with two pyridine groups (Fig. 6B). These Pd²⁺ serve as a connector between two 3 to form a hexagon. The hexagon has a longer side length of 21 Å, a shorter side length of 9.0 Å, and the distance between neighboring Pd²⁺ ions is 18 Å. The angle between two pyridine groups on the same macrocycle is 60° (Fig. 6B).

Fig. 6.

The assembly process and single-crystal X-ray structural analysis of 6. (A) Schematic representation of the complexation between 3 and Pd²⁺ ions in a 1,1′,2,2′-tetrachloroethane/acetone/acetonitrile mixture (3:1:1, v/v/v) at 338 K for 2 h. (B) Single-crystal X-ray diffraction structures of 6 shown from front, side, and top views. Color code: Carbon (silvery gray), Nitrogen (bluish violet), Palladium (orange), and Chlorine (green). Hydrogen atoms and solvent molecules have been omitted for clarity. (C) ¹H NMR spectra comparing 6 (Top) and 3 (Bottom). (D) DOSY spectrum of 6 (600 MHz, CDCl₃, 298 K), illustrating its diffusion properties.

The stability of 6 in solution was also examined using ¹H NMR and ¹H DOSY spectra in CDCl₃. The ¹H NMR spectra indicate that compared to macrocycle 3, the formation of 6 caused the signals corresponding to H_9" and H_10" to change from individual peaks to multiple peaks and shift toward downfield (Fig. 6C). Additionally, other signal peaks on pyridine units also exhibited varying degrees of downfield shift. These findings prove the complexation between 3 and Pd²⁺. The ¹H DOSY NMR analysis gave out additional evidence of the production of a single product, which exhibited a single diffusion coefficient (D) as 3.20 × 10⁻¹⁰ m² s⁻¹ (Fig. 6D and SI Appendix, Fig. S48). According to the Spherical model, the diffusion coefficient (D) can be expressed as follows:

Stokes–Einstein equation (39):

$$D=\frac{k_{B}T}{6\pi \eta {\mathrm{r}}_{\mathrm{H}}},$$ [3]

$$L=\sqrt{3}{\mathrm{r}}_{\mathrm{H}},$$ [4]

D is the molecular diffusion coefficient; r_H is the van der Waals radius of the molecule; k_B is Boltzmann’s constant; T is absolute temperature; and η is the viscosity of the medium. L is the van der Waals length of the long side of a hexagon. Furthermore, the radius of the complex L as 2.1 nm can be calculated. The successful acquisition of assembler cage 6 was further confirmed by mass spectrometry data ([M-Cl⁻]⁺ (m/z): calcd. for C₂₂₂H₂₁₀N₆Cl₅Pd₃, 3458.2293; found, 3458.2983. SI Appendix, Fig. S21).

Upon heating at 338 K for 12 h, isomers 1 or 2, in the presence of Pd²⁺ and a 1,4-dioxane (DOE)/acetonitrile/acetone (3:1:1, v/v/v) solvent mixture, undergo self-assembly to form the corresponding complexes, 1@PdCl₂·DOE or 2@PdCl₂·DOE. These complexes contain a small amount of crystalline material, which was analyzed using single-crystal X-ray diffraction (SI Appendix, Table S8). The analysis confirmed that organic molecular cage 5′, involving isomer 1, is present within the crystalline structure (Fig. 7A and SI Appendix, Fig. S49). The single-crystal X-ray diffraction study reveals that each Pd²⁺ ion forms a 180° coordination with two pyridine groups, which serve as a linker between the two macrocycles (Fig. 7B). The angle between two pyridine groups on the same macrocycle is 5°. The cavity size of 5′ is slightly smaller than 5, ranging from a maximum of 18 Å to a minimum of 7 Å. Additionally, the shape of the cavity tends to be a rectangle (Fig. 7B). After heating isomer 1 at 338 K for 12 h, ¹H NMR spectra reveal its partial conversion into the major product 2, a minor product 4, and unreacted residual compound 1 (SI Appendix, Figs. S50 and S51). To gain insight into the assembly process of 1 or 2 with Pd²⁺, theoretical calculations were conducted using the MM+ method. The results indicate that the formation energy of the M₂L₂ assembly with 2 or 4 is higher than that of the 5′ cage involving 1 (SI Appendix, Fig. S54 and Table S9). This energy difference likely accounts for the predominant assembly of isomer 1 into the metal–organic molecular cage 5′ with Pd²⁺ at 338 K.

Fig. 7.

The assembly process and single-crystal X-ray structural analysis of 5′. (A) Schematic representation of the complexation between 1 or 2 and Pd²⁺ ions in a 1,4-dioxane/acetone/acetonitrile mixture (3:1:1, v/v/v) at 338 K for 12 h, with the single-crystal structure of the complex with 1 highlighted in a red box. (B) Single-crystal X-ray diffraction structures of 5′ shown from front, side, and top views. Color code: Carbon (silvery gray), Nitrogen (bluish violet), Palladium (orange), and Chlorine (green). Hydrogen atoms and solvent molecules have been omitted for clarity.

To simulate and better understand the assembly processes, we investigated the formation mechanisms of metal–organic cages 5 and 6. ¹H NMR analysis revealed the coexistence of both cages in solvents like 1,4-dioxane/acetonitrile/acetone (3:1:1, v/v/v) (SI Appendix, Fig. S53). Time-dependent studies showed a gradual increase in the concentrations of both cages over 1.08 × 10⁴ s (SI Appendix, Fig. S54). After this period, the concentration of cage 5 decreased abruptly, while cage 6 reached equilibrium, likely due to precipitation of the less-soluble cage 5. Interestingly, at 1.8 × 10⁴ s, cage 5 concentration increased again, followed by a rise in cage 6 at 2.52 × 10⁴ s, indicating a dynamic equilibrium influenced by solubility. After 24 h of continuous heating at 338 K and solvent removal, the ¹H NMR spectrum in CDCl₃ confirmed the exclusive coexistence of cages 5 and 6 in a 1:1 molar ratio (SI Appendix, Fig. S55). Notably, dividing the heating process into three 8 h intervals, with solvent removal between each cycle, resulted in the exclusive formation of cage 5 after 24 h, suggesting that solvent removal disrupts the equilibrium, favoring cage 5 generation (SI Appendix, Fig. S56). ¹H NMR spectroscopic monitoring of cage 6 formation upon mixing precursor 3 with PdCl₂ in a 1,1′,2,2′-tetrachloroethane/acetonitrile/acetone (3:1:1, v/v/v) solvent system revealed the immediate assembly of cage 6, indicating a kinetically facile process with a low formation energy barrier (SI Appendix, Figs. S57 and S58).

To investigate solvent-dependent effects on structural assembly, the reaction solvent system comprising tetrachloroethane and 1,4-dioxane was systematically substituted with alternative solvents, including THF, benzene, o-PhCl₂, and PhNO₂. Equimolar mixtures of compound 3 and PdCl₂ were heated in each solvent at 338 K for 24 h. Following solvent removal, the resulting reaction mixtures were analyzed by ¹H NMR spectroscopy in CDCl₃ (SI Appendix, Fig. S59). The ¹H NMR spectra revealed a strong dependence of product distribution on solvent choice. In THF, the major product was cage 6 (82%), accompanied by a minor formation of cage 5 (9%). In contrast, benzene as the reaction medium led to the formation of cage 6 in ≥95% yield, with cage 5 accounting for ≤5% of the products. Remarkably, o-PhCl₂ and PhNO₂ systems exhibited ≥98% selectivity for cage 6 (SI Appendix, Fig. S59). Acetonitrile was essential for PdCl₂ dissolution across all systems. Removing acetone from the 1,4-dioxane solvent system reduced cage 5 yield from 50% to 32%, highlighting acetone’s stabilizing role (SI Appendix, Fig. S60). Excluding THF or benzene maintained ≥98% cage 6 formation (SI Appendix, Figs. S61 and S62). These results indicate weak solvent dependence for cage 6 assembly, while acetone uniquely promotes cage 5 formation in specific solvent matrices, emphasizing its targeted structural influence under controlled conditions.

Topology of Solvent-Modulated Assemblies.

The aforementioned study found that CP2 can undergo temperature-dependent conversion between different atropisomers. Furthermore, the atropisomers of CP2 may self-assemble various complexation architectures with Pd²⁺ in different solvent systems. These results led us to explore the possibility of achieving synergistic control of the assembly involving CP2 using temperature and solvent. The study highlights the self-assembly process of macrocycles, demonstrating the coregulation of rigid ligand assembly by temperature and solvent effects. To date, examples of such coregulation remain limited. Notable cases include studies by Nitschke (20) and Ward (59), which detail how solvent and temperature variations can control the interconversion of metal–organic molecular cages constructed from identical metal/small molecule ligand systems.

Solvent plays a crucial role in chemical reactions. It is frequently utilized to modulate the interactions of macrocycles in supramolecular assembly through the affinity solvent effect (60–65). However, there are limited examples reported where the adjustment of the angle of molecular coordination groups has been employed to enhance the transformation of molecular self-assembly systems. The transformation between 5 and 6 can be modulated by switching the solvent system (Fig. 8). At 338 K, 5 was heated for 9 h in a mixture of 1,1′,2,2′-tetrachloroethane /acetonitrile /acetone (3:1:1, v/v/v). A significant quantity of crystals was generated in the solution. Upon removal of the solvent, the resultant solid exhibited ¹H NMR spectra identical to that of 6 in CDCl₃ (SI Appendix, Fig. S63). The crystal was analyzed using single-crystal X-ray diffraction, which confirmed that its crystal parameters matched those of cage 6. This provides further evidence that cage 5 had successfully converted into cage 6.

Fig. 8.

A schematic diagram illustrating the synergistic control of M₂L₂ and M₃L₃ metal–organic cages containing different CP2 atropisomers, regulated by temperature and solvent. Metastable isomers (1 and 2) or the thermodynamically stable isomer (3) independently react with Pd²⁺ in the presence of 1,4-dioxane to form M₂L₂ rectangular molecular cages, specifically 5′ or 5. Alternatively, isomer 3 binds with Pd²⁺ in 1,1′,2,2′-tetrachloroethane, producing a distinct M₃L₃ hexagonal molecular cage (6). Remarkably, solvent conditions enable reversible transformation between the M₂L₂ cage (5) and the M₃L₃ cage (6), controlled by isomer 3. Furthermore, the M₂L₂ cage (5′) derived from metastable isomer 1 can transform into either cage 5 or cage 6, driven by the synergistic effects of temperature and solvent changes.

In contrast, when cage 6 was heated for 24 h in a mixture of 1,4-dioxane, acetonitrile, and acetone (3:1:1, v/v/v) at 338 K, a different outcome was observed. A large amount of crystals were created in the solution. After removing the solvent, the resulting solid exhibited ¹H NMR spectrum in CDCl₃ that was consistent with cage 5 (SI Appendix, Fig. S64). Single-crystal X-ray diffraction study further confirmed the conversion of 6 into 5. Combined with the analysis of the single-crystal structure of 5 and 6, it is proposed that in polar solvents, 6 initially undergoes ring-opening to form a chain structure. The polar solvents then engage in π–O interactions with macrocycle 3, which decreases the angle between the pyridine groups. This, in turn, promotes the assembly of 5. In a nonpolar solvent environment, the interaction between macrocycle 3 and the solvent is weakened, leading to an increase in the angle between the pyridine groups. This structural change facilitates the formation of 6. Furthermore, solvent switching was employed as a regulatory mechanism to achieve reversible transitions between assemblies 5 and 6.

Theoretical calculations were used to deeper insight into the mechanism of switching between solvent-regulated assemblies (SI Appendix, Figs. S65–S68 and Table S10). In this case, the polar solvent 1,4-dioxane was substituted with H₂O as the solvent environment system for the assemblies. The purpose is to accurately simulate the details of the conversion process from 6 to 5. Theoretical calculations indicate that the energy of the complexation between 3 and Pd²⁺ to produce either 6 or 5 assemblies is lower than that of the corresponding linear assembly structures (SI Appendix, Table S10). Furthermore, the formation energy of 5 is lower than that of 6 in polar solvents. Therefore, it can be further concluded that the formation of 5 is more favorable than 6 in polar solvents. Taking the formation energy of 6 as the standard as 0 kJ mol⁻¹, a potential energy diagram was generated to illustrate the conversion process between assemblies 5 and 6 (SI Appendix, Fig. S69 and Table S10). The potential energy diagram also implies that the energy required to create the transition intermediate dimer is lower than that of the trimer. Therefore, it can be inferred that 5 is classified as a kinetic product, while 6 is considered a thermodynamically stable product. The conversion of 6 to 5 occurred through the intermediate dimer in the presence of a polar solvent.

The prior findings about the heat-induced macrocycle atropisomer transformation led us to further investigate the regulation of the related assemblies. It was found that both aforementioned cages (namely 5 and 6) exhibit not just a sensitivity to solvent, but also a strong dependence temperature (Fig. 8). Thermal treating the solution of 5′ in 1,4-dioxane/acetonitrile/acetone at 393 K for 5 d with N₂ protection, the ¹H NMR spectra of the resulting product exhibited identical signals to those of 5 in CDCl₃ (SI Appendix, Figs. S70 and S73). After heating the solution of 5′ in 1,1′,2,2′-tetrachloroethane/acetonitrile/acetone under N₂ atmosphere for 5 d, the ¹H NMR spectra in CDCl₃ indicate that 5′ has been entirely transformed into 6 (SI Appendix, Figs. S71 and S74).

At 338 K, attempts to generate single-crystal samples of the self-assembled complexes containing 1 or 2 and PdCl₂‚ in a solution of 1,1′,2,2′-tetrachloroethane, acetonitrile, and acetone were unsuccessful. As a result, the mixture was heated further at 393 K for 5 d to encourage crystallization. The ¹H NMR spectra analysis revealed the formation of M₃L₃ cage 6 (SI Appendix, Figs. S72 and S75). Thus, we create a multiway switching system where the combination of solvent and temperature induces atropisomerism, leading to the conversion of MOC architectures (Fig. 8).

To explore the interconversion process between metallorganic cages 5 and 6, we monitored their transformation in solution using time-dependent ¹H NMR spectroscopy (Fig. 9B and SI Appendix, Fig. S76). For cage 5, in a 1,1′,2,2′-tetrachloroethane/ acetonitrile/acetone (3:1:1, v/v/v) solvent at 338 K, 91% of cage 5 underwent rapid conversion within 1,200 s, with the characteristic peaks of cages 5 (H_t) and 6 (H_s) tracked to measure the transformation. After 1,600 s, the system reached equilibrium, with a final molar ratio of 5:6 = 5:95 (Fig. 9C). The thermodynamic and kinetic parameters were calculated from these data, with equilibrium constants determined as shown in Eq. 5.

$$\mathbf{3}\times 5\begin{array}{c}{k}_4\\ \rightleftharpoons \\ k_{-4}\end{array}\mathbf{2}\times 6,$$

$${\mathrm{K}}_4=k_4/k_{-4},$$ [5]

Fig. 9.

Thermal interconversion between Compounds 5 and 6 at 338 K. (A) Schematic illustration of the reversible transformation between 5 and 6 at 338 K. (B) Time-dependent ¹H NMR spectra of 5 in 1,1′,2,2′-tetrachloroethane/acetonitrile-d₃/acetone-d₆ (3:1:1, v/v/v) at 338 K, showing the conversion process. (C) Time-dependent concentration profiles of 5 (black dots) and 6 (red dots) in 1,1′,2,2′-tetrachloroethane/acetonitrile-d₃/acetone-d₆ (3:1:1, v/v/v) at 338 K. (D) Time-dependent ¹H NMR spectra of 6 in 1,4-dioxane/acetonitrile-d₃/acetone-d₆ (3:1:1, v/v/v) at 338 K, showing the reverse transformation. (E) Time-dependent concentration profiles of 6 (black dots) and 5 (red dots) in 1,4-dioxane/acetonitrile-d₃/acetone-d₆ (3:1:1, v/v/v) at 338 K.

A potential energy diagram at 338 K was constructed using the formation energy of 5 as a reference (SI Appendix, Fig. S77 and Table S11), revealing that the energy barrier for 6 is significantly lower than for 5, indicating a thermodynamically favorable pathway for 6.

For cage 6, in a 1,4-dioxane/acetonitrile/acetone (3:1:1, v/v/v) solvent system at 338 K, 18% of cage 6 underwent rapid conversion within 7.2 × 10³ s. However, prolonged heating led to the precipitation of cage 5, which caused an initial increase in the molar ratio of cage 6 before equilibrium was reached at 4.68 × 10⁴ s (Fig. 9E). Due to precipitation, the thermodynamic and kinetic analysis was conducted using the data from the initial 7.2 × 10³ s period. Using the formation energy of 6 as the reference (0 kJ mol⁻¹), a potential energy diagram for the interconversion at 338 K was constructed (SI Appendix, Fig. S79 and Table S11). The diagram confirms that the pathway to form cage 6 is more thermodynamically favorable under these conditions, with a lower energy barrier.

Discussion

In summary, we demonstrated a synergistic control of the conversion between M₂L₂ and M₃L₃ metal–organic cage systems. The external stimuli include temperature and solvent. Macrocycle CP2, which exhibits three distinct stable stiff atropisomers (i.e., 1, 2, and 3) as prepared, was used as the building blocks of these molecular cages. 1, 2, and 3 adopted different symmetries, namely C_s, C₁, and C_2v, respectively. In contrast to the previously reported CP4 case, there are variations in the atropisomer type numbers, the symmetry of the astropisomers, the process of temperature-induced astropisomer transformation, and its properties. When Pd²⁺ is mixed with 1, 2, or 3 in a solvent system containing 1,4-dioxane, M₂L₂ rectangular organic molecular cages 5′ or 5 can be created. On the other hand, in the solution system involving 1,1′,2,2′-tetrachloroethane, triangular M₃L₃ organic molecular cages can be generated with 3 as building blocks. Interestingly, the two assemblies with distinct architectures can be transformed into one another with altering the solvent species. The M₂L₂ assembly cage 5′ consisting of metastable 1 can undergo temperature-induced direct transformation to create cage 5 containing 3. Especially, under the influence of both temperature and solvent conditions, the conversion to a different structure of assembly molecule M₂L₂ or M₃L₃ cage (i.e., 5 or 6) can be achieved. Here, we present a multiple-switching system that allows for the modification of metal–organic cage structure through the use of solvents and temperature changes. This effort aims to develop a highly efficient approach for building smart material systems that incorporate multiple responses and multiple pathway switches.

Materials and Methods

Materials.

Deuterated solvents were purchased from Cambridge Isotope Laboratory (Andover, MA). Other reagents were purchased commercially (Aldrich, Acros, Adamas, Greagent, or Energy Chemical) and used without further purification.

Methods.

A comprehensive suite of analytical techniques was employed to characterize the synthesized compounds and assembled structures. One-dimensional and temperature-dependent ¹H and ¹³C NMR spectra were recorded alongside two-dimensional (COSY and NOESY) and diffusion-ordered spectroscopy (DOSY) experiments, with all chemical shifts referenced to residual solvent peaks. High-resolution mass spectrometry (HRMS) was used to confirm molecular compositions, while single-crystal X-ray diffraction (SCXRD) data were collected to determine the solid-state structures. Full crystallographic parameters are available in the corresponding CIF files. Compounds 7 and 8 were synthesized according to established protocols. Macrocycles 1 to 3 were constructed via palladium-catalyzed cross-coupling reactions performed in toluene under reflux conditions. The supramolecular cage structures 5′/5 and 6 were subsequently assembled through solution-phase coordination with PdCl₂. Their formation and integrity were confirmed by NMR spectroscopy and MALDI-TOF HRMS. Structural optimizations were performed using computational approaches adapted to each molecular environment. As detailed in SI Appendix, structures shown in SI Appendix, Figs. S22 and S39 were optimized in vacuum using the semiempirical PM7 method implemented in MOPAC. SI Appendix, Fig. S52 was refined using the MM⁺ force field in HyperChem 8.0, also under vacuum conditions. In contrast, SI Appendix, Figs. S65–S68 were optimized using the MM⁺ force field within a 3.3 nm³ periodic water box containing 1110 explicit water molecules to simulate solvated conditions. All associated experimental data and comprehensive computational protocols are provided in SI Appendix.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

The X-ray crystallographic data of corresponding structures reported in this study have been deposited to the Cambridge Crystallographic Data Centre (CCDC), with deposition numbers as 2382350 (1) (66), 2382351 (2) (67), 2382352 (3) (68), 2382353 (5) (69), 2382354 (5′) (70), or 2382355 (6) (71). These data can be obtained free of charge from CCDC via www.ccdc.cam.ac.uk/data_request/cif. The crystallographic structures of target molecules are also available in this paper. The optimized molecular geometries obtained via semiempirical PM7 and MM⁺ methods, along with crystallographic data files in CIF format, have been deposited in the Figshare repository (https://doi.org/10.6084/m9.figshare.28666316) (72). The crystallographic datasets were validated using the International Union of Crystallography (IUCr) CheckCIF service. All other data are included in the manuscript and/or SI Appendix.