Introduction

Coordination-driven self-assembly has been proven to be an effective approach for constructing supramolecular structures, due to its high predictability1,2,3,4,5,6. As such, a variety of discrete metallo-supramolecular architectures have been fabricated through different types of self-assembly strategies, including the coordinations of pyridyl7, bipyridyl8, terpyridyl9, or heterotopic ligands10,11,12 with naked metal ions or metal-organic components13,14. Given the increasing demands for diverse and predesigned molecular structures and functions, scientists are encouraged to further develop novel building blocks to achieve high coordination selectivity in precise and controllable supermolecular assembly.

2,2′:6′,2″-tpy and its derivatives have been widely employed in coordination-driven self-assembly on account of their variable coordination ability with transition metal ions, as well as the unique properties and wide applications in optical devices15, catalysis16, self-healing17, and drug delivery18 after complexation. Among the design of tpy-based metallo-supramolecules, the most common strategy is connecting tpy units and directing units at the central pyridine for subsequent coordination, which usually caused the poor selectivity of coordination units4,19,20,21,22,23. To further enhance geometrical diversity and complexity of assemblies, a hierarchical stepwise assembly strategy has been developed to promote the controllability of the self-assembly process by using metals that can form strong coordinative bonds with tpy, such as <tpy-Ru(II)-tpy> or <tpy-Fe(II)-tpy>9,24,25. However, the tedious synthesis and column separation procedures limit its extensive application in supermolecular constructions25,26,27. Currently, it still remains a challenge to precisely construct complex supramolecular structures through the one-step method.

Nature has given us the enlightenment in the exploration of controllable self-assembly of supramolecules with the self-sorting process, such as the complementary pairing of DNA and the folding of polypeptides of proteins28. In recent years, much more attention has been paid to the design of different ligands for the diversity and controllability of supramolecular assembly through social self-discrimination or narcissistic self-recognition manner29,30,31,32,33. Particularly, social self-discrimination is relatively common for tpy moieties, and multiple different building blocks can be engineered to form impressively complex structures34,35. For example, phenanthroline with modification at 2,9-positions and tpy with modification at 6,6″-position could show the complete social self-sorting behaviors when mixed with tpy (Fig. 1a)36,37. In a recent review, Schmittel summarized a dynamic library formed by hetero-ligand motifs and introduced a lot of elegant structures constructed by the social self-sorting process38. In contrast, the narcissistic self-recognition manner for tpy moieties is rarely reported39,40,41. It has been more than 20 years since Lehn reported the first narcissistic self-sorting of a double helix, and there have also been many reports about narcissistic self-sorting in metal coordination42,43,44,45,46. However, most of these studies considering the narcissistic self-sorting of the structure as a whole, while the report focusing on narcissistic self-sorting used as self-recognition sites remains limited45,47,48,49,50,51. Furthermore, the homogeneous interactions in biological systems also prompted scientists to urgently develop ligand motifs for self-recognition52,53.

Fig. 1: The self-assembly of metallo-macrocycles by the narcissistic self-recognition tpy moieties.
figure 1

Two different self-sorting processes in coordination-driven self-assembly; a social self-discrimination motifs in literatures; b narcissistic self-recognition motifs in the current work. We combined the narcissistic self-recognition tpy moieties with tpy to form multitopic ligands for precise self-assembly of metallo-macrocycles. c Schematics for the self-assembly of hexagonal macrocycles Zn9(LA)6 and Zn12(LB)6.

In this paper, to achieve coordination with high selectivity and specificity, we design two ortho-modified terpyridine ligands, i.e., MA and MB as model systems to form head-to-tail coordination complexes Zn2(MA)2 and Zn3(MB)2, respectively (Fig. 1b). The single-crystal structures of Zn2(MA)2 and Zn3(MB)2 show the non-coaxial feature of ortho-modified tpy after coordination, which is apparently different from the rotation of conventional tpy around its axis19,54. Remarkably, MA, MB, and tpy mixture with Zn(II) could assemble into three distinct complexes Zn2(MA)2, Zn3(MB)2, and Zn(tpy)2, suggesting the narcissistic self-sorting behaviors. We then combined these moieties with tpy to form sterically congested multitopic ligands (LA and LB) for precise self-assembly of hexagonal macrocycles Zn9(LA)6 and Zn12(LB)6 (Fig. 1c). It’s a common strategy to promote the hierarchical assembly of metallacycles by introducing orthogonal interactions55,56,57. The tpy-based metallacycles could hierarchically self-assembly into hollow tubular or berry-type nanostructure through the face-to-face stacking according to the previous reports15,21,58,59. Interestingly, it is found that the significantly steric congestion and non-coaxial structure of Zn9(LA)6 caused the restriction in the edging rotation of the hexagons, which further causes the hexagons to hierarchically assemble into giant cyclic nanostructures and metallogels via edge-to-edge stacking rather than face-to-face stacking.

Results and discussion

Synthesis and characterization of model system MA and complex Zn2(MA)2

In our design, the modification at terpyridyl 6-position played a central role in the self-soring assembly. The simple and efficient synthesis process of MA only included a three-step reaction of the starting material, followed by the one-step purification of the product using column chromatography. Especially, the key compound 1 was synthesized by a typical Kröhnke reaction with pyridinium salt 5 (Supplementary Fig. 1)60. The final motif MA was obtained via Suzuki coupling reaction in a good yield (77%) and characterized by NMR and mass spectrometry (Supplementary Figs. 28). After that, MA and Zn(NO3)2•6H2O (with a precise stoichiometric ratio of 1:1, Fig. 2a) were mixed in CHCl3 and MeOH (1:3, v/v) at 50 °C for 12 h, followed by the addition of excessive NH4PF6 (to exchange NO3 with PF6) in methanol to give a white precipitate Zn2(MA)2 in a yield of 91%.

Fig. 2: The self-assembly and characterization of head-to-tail complex Zn2(MA)2.
figure 2

a The self-assembly of model system MA with Zn(II). b The 1H NMR spectra (600 MHz, 300 K) of MA in CDCl3 and complex Zn2(MA)2 in CD3CN (3 mg/mL). c X-ray crystal structure in a side view of complex Zn2(MA)2. d X-ray crystal structure in a front view of complex Zn2(MA)2. Non-coordinated anions and solvent are omitted for clarity (C, gray or yellow; H, white; N, blue; Zn, orange).

1H NMR spectra of MA and complex Zn2(MA)2 were shown in Fig. 2b. Given the dissymmetrical nature of MA, five kinds of expected pyridines signals attributed to the terpyridine moieties were confirmed by 2D-COSY results (Supplementary Fig. 6). In the spectrum of complex Zn2(MA)2, the peaks were sharp and well-resolved. The 2D-COSY and NOESY for Zn2(MA)2 (Supplementary Figs. 914) also showed five sets of pyridines signals as MA, suggesting a highly symmetrical structure of complex Zn2(MA)2. Compared with MA, 6-position (6 and a6) signals of tpy were shifted upfield due to the electron shielding effect61. At the same time, the proton signals of e-tpy, f-Ph, and g-Ph were also shifted to upfield, which should be attributed to the existence of π–π interaction between tpy moiety and phenyl in the other ligand. In ESI-MS (Supplementary Fig. 15A), one prominent set of peaks with charge states from 2+ to 4+ was observed (due to the loss of different numbers of PF6). After deconvolution, the molecular weight of complex Zn2(MA)2 was 1940 Da, matching well with its expected chemical composition of two MA moieties, two Zn(II) ions, and four PF6. All the experimental isotope patterns agree excellently with the corresponding simulated isotope patterns (Supplementary Fig. 16). In traveling wave ion mobility-mass spectrometry (TWIM-MS)23, complex Zn2(MA)2 showed a series of charge states with narrow drift time distribution ranging from 2+ to 4+, indicating the formation of a discrete product but without any isomers and conformers (Supplementary Fig. 15B).

To further confirm the structure, the single crystal of Zn2(MA)2 was obtained by slowly diffusing the vapor of ethyl acetate into Zn2(MA)2 in acetonitrile for over 2 weeks. As expected, X-ray crystallographic analysis (Fig. 2c, d and Supplementary Table 1) showed that two Zn(II) are sandwiched between two MA to form a dimer with the head-to-tail structure (Supplementary Data 1 and Supplementary Movie 1). The phenyl groups in the middle are parallel to each other (yellow part), and the distance between the two interlayers is 3.57 Å, indicating the existence of π–π interactions41,62.

Synthesis and characterization of model system MB and complex Zn3(MB)2

By referring to the synthetic and characterization procedures of MA, we further designed and obtain another model system MB (Fig. 3a and Supplementary Fig. 17), which includes one additional tpy unit compared to MA and thereby exhibits an enhanced sterically congested effect. An extra tert-butyl was introduced to improve the solubility. Tpy-based molecules have been extensively studied since the 1930s, however, it is worth noting that the unique structure of MB, has never been reported40,63,64. The synthesis of MB (yield 75%) and self-assembly of complex Zn3(MB)2 (yield 87%) (MB:Zn(II) = 2:3) followed the same procedure as that of MA and Zn2(MA)2.

Fig. 3: The self-assembly and characterization of head-to-tail complex Zn3(MB)2.
figure 3

a The self-assembly of model system MB with Zn(II). b The 1H NMR spectra (600 MHz, 300 K) of MB in CDCl3 and Zn3(MB)2 in CD3CN (3 mg/mL). c X-ray crystal structure in a front view of Zn3(MB)2. Non-coordinated anions and solvent are omitted for clarity (C, gray or yellow; H, white; N, blue; Zn, orange).

The 1H NMR data of MB and Zn3(MB)2 were shown in Fig. 3b. In the spectrum of MB with three tpy units, eight kinds of pyridines signals could be observed (Supplementary Figs. 18–24). Moreover, the complex Zn3(MB)2 exhibited a more complicated structure and spectrum than Zn2(MA)2 (Supplementary Figs. 2530), due to the enhanced steric hindrance effect caused by the introduction of the additional tpy unit. For each tpy, the differences in chemical shift (Δδ) were 1.48 ppm and 0.8 ppm for 6 position proton and a6 proton, respectively, suggesting that the 6-position of a-tpy shows the strong shielding effect. The signals of tpy-3 and tpy-5 displayed upfield or downfield shifts after coordination, which were consistent with the previous studies65. In addition, the restrictive structure induced the split of NMR signals. Both the signals of the middle phenyls and tpy-b6 split into two sets of peaks, which indicate their rotation restriction caused by the altered chemical environment. ESI-MS and TWIM-MS (Supplementary Figs. 31, 32) spectra of Zn3(MB)2 also exhibited a similar prominent set of peaks with different charge states and narrow drift time distribution ranging from 2+ to 4+, suggesting the formation of a single product without any overlapping isomers or conformers. The molecular weight of complex Zn3(MB)2 (3020 Da) agrees well with its expected chemical composition of two MB ligands, three Zn(II) ions, and six PF6.

The single-crystal data of Zn3(MB)2 was obtained by slowly diffusing the vapor of carbon tetrachloride into acetonitrile solution for over 3 weeks. Complex Zn3(MB)2 also showed a sandwich-like structure with a certain degree of distortion, suggesting higher steric congestion caused the helical shape (Fig. 3c). The different distances between the two middle phenyls (yellow part) were 3.6 and 3.7 Å, respectively, which confirmed the helical structure (Supplementary Data 2, Supplementary Table 2, and Supplementary Movie 2). More importantly, under the sterically congested environment, the different tpy units coordinated with the same Zn(II) cannot maintain the vertical alignments and are forced to exhibit twisted structures. Furthermore, Zn3(MB)2 has two chiral conformations due to the unique head-to-tail coordination mode, and both of the conformations could be found in a single crystal (Supplementary Fig. 33).

Self-sorting behavior of MA, MB, and tpy

To evaluate the selectivity of these model systems, we investigated the self-sorting behavior of MA, MB, and conventional tpy66. In the self-sorting study, MA, MB, and conventional tpy were mixed together in an equimolar ratio with the corresponding amount of Zn(II) for the overnight self-assembly at 50 °C. ESI-MS clearly illustrated three sets of signals for the corresponding complexes but without statistical complexes (Fig. 4a, b), indicating that these model complexes exhibit ideal self-sorting properties in a pluralistic system. Both ESI-MS and NMR of any binary mixture (Supplementary Figs. 3436) showed independent signals for three complexes Zn2(MA)2, Zn3(MB)2, and Zn(tpy)2, suggesting a characteristic narcissistic self-sorting29,67,68. Moreover, we also monitored the kinetic process of narcissistic self-sorting, and the system basically reached equilibrium after 24 h (Supplementary Fig. 37). As expectation, in accord to the maximum occupancy of coordination sites proposed by Lehn46, MA, MB, and tpy tend to narcissistically coordinate to form the most energy favorable structure. In this context, we are inspired to use MA and MB as self-recognition sites to further construct complex supramolecular structures.

Fig. 4: The narcissistic self-sorting behavior of MA, MB, and tpy.
figure 4

a A schematic representation for the narcissistic self-sorting process. b ESI-MS of the assembly of mixed MA, MB, and tpy in an equimolar ratio with the corresponding amount of Zn(II).

Synthesis and self-assembly of ligands LA and LB with multiple tpy moieties

It is reported that the ditopic tpy ligands with 120° angle tend to self-assemble into a mixture of macrocycles with uncontrollable size and structure20. However, we expect that the dissymmetrical tritopic (LA) and tetratopic (LB) (Fig. 1c) ligands with 120° angle that contain narcissistic self-sorting moieties could be used to precisely construct complex supramolecular macrocycles Zn9(LA)6 and Zn12(LB)6 (Supplementary Movie 3 and Supplementary Movie 4). In order to improve the solubility of ligands, alkoxy chains (R = -OC6H13) were introduced. Compared with LA (yield, 76%), the synthesis of LB with one more tpy units turns to be significantly challenging (Supplementary Fig. 38 and Supplementary Fig. 39). Fortunately, we successfully obtained LB (yield, 37%) by adding CH3COOH/DMF as an auxiliary agent. The obtained ligands LA and LB were assembled into complexes Zn9(LA)6 (yield, 86%) and Zn12(LB)6 (yield, 84%) and were characterized by NMR, MALDI-TOF, ESI-MS, and TWIM-MS (Fig. 5 and Supplementary Figs. 4070).

Fig. 5: Characterization of the hexagonal macrocycles Zn9(LA)6 and Zn12(LB)6.
figure 5

a 1H NMR spectra (600 MHz, 300 K) of ligand LA in CDCl3 and hexagon Zn9(LA)6 in CD3CN (4 mg/mL). b ESI-MS, and c TWIM-MS plots (m/z vs drift time) of Zn9(LA)6. d 1H NMR spectra (600 MHz, 300 K) of ligand LB in CDCl3 and hexagon Zn12(LB)6 in CD3CN (4 mg/mL). e ESI-MS and f TWIM-MS plots (m/z vs drift time) of Zn12(LB)6.

Compared with LA and LB, 1H NMR spectra of the complexes Zn9(LA)6 and Zn12(LB)6 were significantly broadened (Fig. 5a, d), suggesting the assembly of large complexes with a slower tumbling on the NMR time scale24. Importantly, the peaks of pyridines on the b-tpy and c-tpy turn broader, due to the sterical congestion of the coordination units and the restriction in free rotation of the phenyl caused by the high steric congestion of alkoxy chains. In the spectrum of LA, seven kinds of expected pyridines signals were observed with respect to the tpy moieties. Owing to the continuous dissymmetrical modification of a- and b-tpy, LB displayed more complicated proton signals (i.e., ten distinguished kinds of pyridine peaks) than that of LA. All six positions of pyridines were shifted towards upfield, because of the electron shielding effects after complexation. However, the chemical shifts of different protons differ significantly owing to the variable shielding strength. For instance, the proton at a6 position showed a maximum chemical shift of Δδ = 1.41 ppm; while a slightly upfield-shift Δδ = 0.55 ppm was observed for b6 position. 2D-COSY and NOESY for two complexes Zn9(LA)6 and Zn12(LB)6 (Supplementary Figs. 53–56 and Supplementary Figs. 5962) also showed the same sets of tpy signals as ligands, suggesting the symmetrical structures of these two complexes. The diffusion-ordered NMR spectroscopy (DOSY) provided dimensional information for Zn9(LA)6 and Zn12(LB)6. As shown in Supplementary Fig. 66, all proton signals appeared at the same band, indicating the formation of discrete assemblies. The diffusion coefficients in CD3CN were log D = −9.61 for Zn9(LA)6 and log D = −9.64 for Zn12(LB)6, respectively. The experimental hydrodynamic radii (rH) for Zn9(LA)6 (2.4 nm) and Zn12(LB)6 (2.6 nm) agreed well with the modeling structures.

In ESI-MS (Fig. 5b, e), one prominent set of peaks with charge states (from 7+ to 14+ for Zn9(LA)6 and from 8+ to 14+ for Zn12(LB)6) was observed on account of the loss of different numbers of PF6. After deconvolution, the molecular weights of these two complexes were 8869 and 12974 Da, respectively, matching well with the expected chemical composition of Zn9(LA)6 (6 LA ligands, 9 Zn(II) ions, and 18 PF6) and Zn12(LB)6 (6 LB ligands, 12 Zn(II) ions, and 24 PF6). All the experimental isotope patterns agreed excellently with the corresponding simulated isotope patterns (Supplementary Figs. 6970). Moreover, the TWIM-MS spectra of complexes Zn9(LA)6 and Zn12(LB)6 also showed narrow drift time distribution, supporting the formation of single species with rigid structures (Fig. 5c, f). The full characterization of NMR, ESI-MS, and TWIM-MS confirmed that the two types of tpy moieties in ligands LA and LB could precisely self-assemble into hexagonal macrocycles through the narcissistic self-recognition mechanism. Such moieties with high coordination selectivity and specificity may find their potential in the constructions of more complex metallo-supramolecules with precisely controlled shapes and sizes.

Interestingly, when complexes Zn9(LA)6 and Zn12(LB)6 were mixed in an equimolar ratio at 50 °C for overnight, ESI-MS and TWIM-MS clearly illustrated two additional sets of signals besides Zn9(LA)6 and Zn12(LB)6 (Supplementary Fig. 71). The compositions of Zn10(LA)4(LB)2 and Zn11(LA)2(LB)4 were confirmed by analyzing the ESI-MS and TWIM-MS data (Supplementary Figs. 72,  73). No exchange for odd number ligands was observed in this mixture (Supplementary Figs. 7477), suggesting that it could be a subcomponent exchange based on the dominant coordination dimer rather than the usual ligand exchange69,70. It indicated that the dimer structures based on ditopic and tritopic tpy moieties not only exhibit higher selectivity but also show higher stability than single-tpy unit.

Hierarchical self-assembly

Our study shows that, with increasing the concentration of Zn9(LA)6 and Zn12(LB)6 in acetonitrile at room temperature, both of the solutions could form metallogels at a concentration of 45 mg/mL for Zn9(LA)6 and 25 mg/mL for Zn12(LB)6 (Fig. 6l). The gels were also temperature-responsive and could be converted into solutions after heating at 50 °C for 1 h. To explore the mechanism of gelation, we then used TEM to investigate the morphology of Zn9(LA)6 and Zn12(LB)6 at different concentrations (Supplementary Figs. 78,  79). At low concentrations (10−6 M), the assemblies were almost distributed in a monodispersed manner. Both complexes exhibited unique hollow structures, suggesting high rigidity, stability, and large inner space. The measured sizes of the two complexes were 5.5 nm for Zn9(LA)6 and 6.5 nm for Zn12(LB)6, respectively (Fig. 6b, c, f, g). However, at a high concentration (10−5 M), some giant cyclic nanostructures in larger size were observed (Fig. 6d). The diameters of these structures were much larger than that of individual Zn9(LA)6, implying the clustering of Zn9(LA)6. The height of the hierarchically assembled nanostructures measured by the AFM image (Fig. 6j, k and Supplementary Figs. 8083) was about 1.5 nm, which was very close to the height of the monolayer. When the concentration further increased, these giant macrocycles formed a three-dimensional network via intermolecular interaction (Fig. 6e). For Zn12(LB)6 with a larger size, although similar monolayer nanostructures were observed (Fig. 6h), the degree of aggregation was significantly increased (Supplementary Figs. 8486). Further, at a much higher concentration of 10−4 M, the formed giant macrocycles showed obvious vertical stacking in space (Fig. 6i).

Fig. 6: Characterization of the hierarchical self-assembly of Zn9(LA)6 and Zn12(LB)6.
figure 6

a Schematic diagram of edge-to-edge stacking by Zn9(LA)6. b The energy-minimized structure of Zn9(LA)6 from molecular modeling. c TEM images of individual Zn9(LA)6 at a concentration of 10−6 M in CH3CN (scale bar, 50 and 20 nm for zoom-in image). d TEM images of the assembled cyclic nanostructures of Zn9(LA)6 at a concentration of 10−5 M in CH3CN (scale bar, 50 and 20 nm for zoom-in image). e TEM images of the 3D network structure formed by assembled cyclic nanostructures of Zn9(LA)6 at a concentration of 10−4 M in CH3CN (scale bar, 100 nm). f The energy-minimized structure of Zn12(LB)6 from molecular modeling. g TEM images of individual Zn12(LB)6 at a concentration of 10−6 M in CH3CN (scale bar, 50 and 20 nm for zoom-in image). h TEM images of the assembled cyclic nanostructures of Zn12(LB)6 at a concentration of 10−5 M in CH3CN (scale bar, 100 nm). i TEM images of the 3D network structure formed by assembled cyclic nanostructures of Zn12(LB)6 at a concentration of 10−4 M in CH3CN (scale bar, 100 nm). j AFM images of the hierarchically assembled cyclic nanostructures by Zn9(LA)6 (scale bar, 50 nm). k Cross-section of the cyclic nanostructure shown in image j. l Photograph of Zn12(LB)6 gel formation at a concentration of 25 mg/mL in CH3CN.

As well-known, the <tpy-M(II)-tpy> unit could rotate freely around the axis without space obstruction. However, due to the special spatially complementary coordination of LA and LB, these two axial dissymmetrical ligands exhibited entirely steric congestions. Therefore, it can be inferred that the hexagons are basically composed of vertical edges, since the rotation of the dimer can result in the disassembly (Fig. 6a). In order to study the edge-to-edge stacking of Zn9(LA)6, we tried to grow single crystal of Zn9(LA)6. However, all efforts to grow a single crystal of Zn9(LA)6 has proven to be unsuccessful. Instead, we synthesized a model ligand MA-OC6H13 (Supplementary Figs. 87–94) and further got the complex Zn2(MA-OC6H13)2 which has exactly the same structure as the ditopic part of Zn9(LA)6 (Supplementary Figs. 95102). The single-crystal packing of Zn2(MA-OC6H13)2 provided strong evidence for the edge-to-edge stacking of Zn9(LA)6 (Supplementary Data 3 and Supplementary Table 3). As shown in Fig. 7, the intermolecular packing of Zn2(MA-OC6H13)2 is more compact than that of Zn2(MA)2 (Supplementary Fig. 103), which indicated that the introduction of alkyl chains can enhance the intermolecular interactions. The distance between pyridine and central phenyl of another molecule is 3.8 Å, which indicated the existence of π–π interaction. The distance between the alkyl chain and central phenyl of another molecule is 2.6 Å, which indicated the existence of CH–π interactions. Since Zn12(LB)6 could provide additional sites in the vertical direction than complex Zn9(LA)6, Zn12(LB)6 is hard to find the discrete cyclic nanostructures as that of Zn9(LA)6.

Fig. 7: Crystallographic structure of complex Zn2(MA-OC6H13)2.
figure 7

a Crystal packing of two complexes Zn2(MA-OC6H13)2. b Crystal packing structure in a side view of complex Zn2(MA-OC6H13)2. c Crystal packing structure in a front view of complex Zn2(MA-OC6H13)2. H atoms, non-coordinated anions, and solvent are omitted for clarity (C, gray or yellow; N, blue; Zn, orange).

Photophysical properties

Considering the enhanced conjugation and restricted rotation of the structures, we further investigated UV/Vis absorption and photoluminescence of all ligands and complexes in the solution state (10−6 M). As shown in Fig. 8a, compared with the absorption spectra of ligands (Supplementary Figs. 104105), all complexes displayed a significant redshift originating from intra-ligand charge transfer (ILCT)71. In sharp contrast with Zn(tpy)2, the maximum emission wavelength of Zn2(MA)2 displayed a ca. 75 nm redshift, ascribed to the enhanced conjugation (Fig. 8b). However, complex Zn3(MB)2 with a more complicated structure also showed a similar maximum emission wavelength to Zn2(MA)2, the reason should be the twisted structures induced the conjugation. For hexagonal macrocycles Zn9(LA)6 and Zn12(LB)6, the restrictive intramolecular rotation in the congested and non-coaxial structures caused further enhance of their fluorescence intensity. Moreover, due to the introduction of alkoxy chains as electron donors, the maximum emission peaks of Zn9(LA)6 and Zn12(LB)6 exhibited 20 nm red-shifted than that of Zn2(MA)2 and Zn3(MB)2.

Fig. 8: Photophysical data of the complexes.
figure 8

a Normalized UV/Vis absorption of all complexes in CH3CN. b PL spectra of all complexes in CH3CN (10−6 M, λe = 325 nm).

Conclusion

In summary, we have designed and synthesized a series of multitopic tpy ligands with designed coordination selectivity by ortho-modification. As shown by the single-crystal detections, the model systems MA and MB can assemble into head-to-tail coordination complexes Zn2(MA)2 and Zn3(MB)2, respectively. Moreover, the assembly of mixtures of MA, MB, and tpy exhibit excellent selectivity due to the narcissistic self-sorting mechanism. Further, we introduced these moieties with high selectivity into multitopic ligands LA and LB, which were used to further precisely construct the hexagonal metallo-supramolecules. Such a design caused the restriction in rotations of the hexagons, which further leads to their hierarchical assembly into giant cyclic nanostructures and metallogels. Moreover, these complexes showed significantly enhanced fluorescence intensity in the solution state than the complexes based on conventional tpy due to the additional ligand conjugation and the restricted chemical environment. Our design and fabrication of dissymmetrical coordination moieties could pave a new avenue for the development of a set of congested coordination pairs with high selectivity and specificity for the assembly of sequence-specific metallo-supramolecular architectures.

Methods

General procedures

All reagents were purchased from Sigma-Aldrich, Matrix Scientific, Alfa Aesar, Jilin Chinese Academy of Sciences—Yanshen Technology Co. Ltd., and used without further purification. Column chromatography was conducted using SiO2 (VWR, 40–60 µm, 60 Å) and the separated products were visualized by UV light.

Synthesis

All the new compounds were fully characterized and spectra are given in Supplementary Methods.

Nuclear magnetic resonance (NMR)

NMR spectra data were recorded on a 400, 500, and 600 MHz Bruker Avance NMR spectrometer in CDCl3 or CD3CN with TMS as reference. For NMR spectra see Supplementary Information.

ESI-MS and TWIM-MS

Electrospray ionization (ESI) mass spectra were recorded with a Waters Synapt G2 tandem mass spectrometer, using solutions of 0.5 mg sample in 1 mL of MeCN/MeOH (3:1, v/v) for complexes. The TWIM-MS experiments were performed under the following conditions: ESI capillary voltage, 3 kV; sample cone voltage, 30 V; extraction cone voltage, 3.5 V; source temperature 100 °C; desolvation temperature, 100 °C; cone gas flow, 10 L/h; desolvation gas flow, 700 L/h (N2); source gas control, 0 mL/min; trap gas control, 2 mL/min; helium cell gas control, 100 mL/min; ion mobility (IM) cell gas control, 30 mL/min; sample flow rate, 5 μL/min; IM traveling wave height, 25 V; and IM traveling wave velocity, 1000 m/s.

Single-crystal X-ray diffractions

X-ray diffraction data for Zn2(MA)2 and Zn2(MA-OC6H13)2 were measured by a Bruker D8 Venture X-ray single-crystal diffractometer using a Cu·Kα radiation (λ = 1.54178 Å) at 100 K. X-ray diffraction data for Zn3(MB)2 was collected using synchrotron radiation and MAR325 CCD detector at Shanghai Synchrotron Radiation BL17B Beamline. Crystallographic data and structural characteristics of Zn2(MA)2, Zn2(MA-OC6H13)2, and Zn3(MB)2 are summarized in Supplementary Tables 13, respectively. Crystallographic information files for the complexes are provided in Supplementary Data 13, respectively.

Transmission electron microscopy (TEM) analysis

The sample of Zn9(LA)6 and Zn12(LB)6 were dissolved in CH3CN at concentrations of 10−6, 10−5, and 10−4 M, respectively. The solutions were left overnight at 298 K and then dropped cast onto copper grids (ultrathin carbon supported by a lacey carbon film on a 400 Mesh copper grid) and the extra solution was absorbed by filter paper to avoid aggregation. The TEM images of the drop cast samples were taken with a JEM-2100F transmission electron microscope.

AFM imaging

AFM imaging was performed on a Bruker Dimension Icon AFM system with ScanAsyst and the data was processed by NanoScope Analysis version 2.0 (Bruker Software, Inc.). The sample of Zn9(LA)6 and Zn12(LB)6 were dissolved in CH3CN at a concentration of 10−5 M, The solution was left overnight and then dropped cast onto a silicon wafer after surface cleaning.

Photophysical measurements

UV-vis spectra of solutions were recorded on a PerkinElmer LAMBDA-365 spectrophotometer. Fluorescence emission spectra were measured by a Shimadzu spectrofluorimeter RF-5301PC. Solutions were placed in 1 cm path length quartz cells.

Molecular modeling

Energy-minimized structures were obtained following the settings in the literature. Calculations were proceeded with Geometry Optimization and followed by Anneal in Forcite module of Materials Studio version 8.0 program (Accelrys Software, Inc.).