Sequence-encoded layered heteroleptic metalla-[2]catenanes for programmable supramolecular function

Abstract

Molecular information encoded within supramolecular frameworks offers a powerful paradigm for directing emergent function beyond the genetic code, but systematic investigations into alternative spatial configurations and their consequences remain scarce. Here we use metalla-[2]catenanes to probe sequence–function relationships in layered architectures. By combining two or three size-matched N-heterocyclic carbene ligands with Ag(I) nodes, we selectively construct heteroleptic metalla-[2]catenanes through both direct assembly and supramolecular fusion pathways. X-ray crystallographic analysis unambiguously confirms the targeted sequences, while semiempirical and density functional theory calculations reveal their thermodynamic preference over alternative isomers. Photothermal conversion studies further demonstrate that sequence-specific charge-transfer interactions yield distinct macroscopic responses. Collectively, these results identify heteroleptic metalla-[2]catenanes as a robust model for elucidating how spatial arrangement governs system-level behavior and for advancing molecular coding principles in functional supramolecular design.

Molecular information encoded within supramolecular frameworks offers a powerful paradigm for directing emergent function beyond the genetic code, but systematic investigations into alternative spatial configurations and their consequences remain scarce. Here the authors use metalla-[2]catenanes to probe sequence–function relationships in layered architectures.

Introduction

Molecular information is emerging as a transformative paradigm that extends far beyond the genetic code of nucleic acids¹. Information can be encoded within molecular frameworks that have distinct structural and chemical features, redefining how it is stored, processed, and transmitted at the molecular level^2,3. In self-assembly, the spatial arrangement of functional motifs acts as a “molecular code” that dictates intermolecular interactions, symmetry, and dynamic response pathways, ultimately directing emergent function^4–8. These principles provide a powerful foundation for the bottom-up design of advanced functional materials, with numerous implications for biomimetic manufacturing^9,10, energy conversion¹¹, and chiral technologies^12,13.

Despite these advancements, establishing precise structure–function relationships between the sequence of functional motifs at the molecular level and the macroscopic properties of assembled systems remains a central challenge^14–16. Current research on the controllable design of assemblies with specific motif sequences remains limited. To address this issue, recent efforts have focused on the development of multicomponent self-assembly systems incorporating diverse motifs^17–19. In metallosupramolecular assemblies, synthetic chemists have achieved nonstatistical precision^8,20–22 in multicomponent architectures—overcoming entropy barriers, avoiding self-sorting, and suppressing statistical mixtures—through strategies such as guest templation^23,24, geometric complementarity^25,26, coordination-site engineering^27–31, hierarchical assembly^32,33, and kinetic control^34–36. Nevertheless, most studies have focused on cavity-centered functions, arising from the anisotropic radial (Fig. 1a, for metallocages) or circular arrangement (Fig. 1b, for metallacycles) of motifs, particularly in recognition^37,38 and confined catalysis^33,39–42. In contrast, systematic investigations into alternative spatial organizations—such as layered or planar grid arrangements—and their effect on properties (Fig. 1c), including electron transport, optical behavior, and mechanical performance, remain scarce. There is therefore an urgent need to construct self-assembly systems with spatial configurations beyond radial architectures to determine how specific motif arrangements translate into system-level functions.

Fig. 1. Schematic presentation for three arrangements of different motifs and relative sequences in supramolecular architectures.

a Radial arrangement of different motifs in metallocages. b Circular arrangement of different functional motifs in metallacycles. c Stereoscopic layered arrangement of different motifs in tetranuclear NHC MCAT. d Two-component MCATs based on four NHC ligands with 8 component sequences. e Three-component MCATs based on four NHC ligands with 18 component sequences. x, y, and z were used to denote the numbers of the three NHC ligands. B, G, R, and Y were the capital letters of blue, green, red, and yellow, which were used to denote different building blocks. x, y, and z were used to denote the numbers of the three NHC ligands.

Building on the previous work with metal N-heterocyclic carbene (NHC) assemblies^43–46, we identified tetranuclear metalla-[2]catenane (MCAT) as an ideal platform for this purpose^47,48. Unlike conventional metallocages and metallacycles featuring radial arrangements, the four NHC ligands in MCAT use a distinctive stereoscopic layered configuration. Heteroleptic assembly of two or three size-matched NHC ligands with four Ag(I) nodes can theoretically yield MCATs with 8 and 18 possible outcome sequences (Figs. 1d, e), respectively, providing a versatile platform to probe stereoscopic, layered sequence–function relationships.

In this study, we report the selective construction of three heteroleptic MCATs with integrative self-sorting behavior, including two distinct two-component sequences and one three-component sequence. These assemblies were accessed either via direct assembly of NHC ligands with Ag(I) nodes or structural transformation from homoleptic precursors trough a supramolecular fusion strategy. All the structures were unambiguously confirmed by single-crystal X-ray diffraction and high-resolution mass spectrometry. Complementary theoretical calculation studies that combine semiempirical methods and density functional theory (DFT) calculations revealed that the binding energies of the observed MCATs are significantly lower than those of their respective isomers, supporting the thermodynamic preference for the obtained structures. Furthermore, photothermal conversion experiments demonstrated sequence-dependent differences in performance, which can be attributed to charge-transfer interactions between the distinct functional motifs. Collectively, these findings establish heteroleptic MCATs as a versatile platform for probing the relationship between spatial motif arrangement and macroscopic function in multicomponent assemblies. This approach provides new insights into how molecular coding principles can be applied to the design of functional supramolecular systems.

Results

Synthesis and characterization of homoleptic MCAT_AAAA and MC_BB

Imidazo[1,5-a]pyridinium salts H₂A and H₂B were prepared in four steps from 2,7-dibromophenazine and 2,6-dibromoanthracene by adjusting the reported procedure⁴⁹. The complete synthetic route is outlined in the Supplementary Information (Supplementary Figs. 1–26). The imidazo[1,5-a]pyridine-based NHC ligands were strategically chosen to fulfill the structural requirement for optimal interplanar surface-to-surface separation ( ~ 6.6 Å), facilitating the directional assembly of mechanically interlocked supramolecular architectures through precise π-system alignment^47–49. The homoleptic MCAT_AAAA was obtained in 93% yield via the reaction of H₂A with excess Ag₂O in acetonitrile under the exclusion of light at 65 °C for 12 h (Fig. 2, method a). The formation was verified by two-dimensional (2D) NMR spectroscopy and ESI mass spectrometry (Supplementary Figs. 27–32) as follows: Peaks in the ¹H NMR spectrum of MCAT_AAAA (Supplementary Fig. 27) showed distinct shifts and were split into two sets of signals, revealing the presence of two chemically inequivalent di-NHC ligands within the assembly. As expected for Ag^I–NHC [2]catenane, two di-NHC ligands are located inside the cavity, and the remaining ligands constitute the outer framework. The formation of MCAT_AAAA was also evident from ¹³C{¹H} NMR spectroscopy showing the two well-resolved sets of carbon signals (Supplementary Fig. 28). The ¹H diffusion-ordered spectroscopy (DOSY) NMR spectrum of MCAT_AAAA showed a single diffusion coefficient (D = 5.01 × 10^-10 m² s^-1) for all of the aromatic proton signals, suggesting that all the resonances belong to a single assembly (Supplementary Fig. 31). ESI mass measurements also indicated the [2]catenane structure of MCAT_AAAA, with prominent peaks at m/z = 891.5664 (calcd for [MCAT_AAAA ‒ 3(PF₆^‒)]³⁺ 891.5541) and 1409.8347 (calcd for [MCAT_AAAA ‒ 2(PF₆^‒)]²⁺ 1409.8134) (Supplementary Fig. 32), which was consistent with the theoretical distribution.

Fig. 2. Self-assembly of MCATs from di-NHC ligands H₂A, H₂B and H₂C and structural transformations of the assemblies.

a–c Synthesis of MCAT_AAAA, MR_BB and MCAT_CCCC through the reaction of H₂A, H₂B and H₂C, respectively, with Ag₂O in MeCN at 65 °C. d, e Synthesis of MCAT_CAAC and MCAT_CBBC through the reaction of H₂C and H₂A or H₂B, respectively, in a 1:1 ratio with Ag₂O in acetonitrile at 65 °C. f, g Supramolecular fusion-based transformation of MCAT_CCCC and MCAT_AAAA or MR_BB at a 1:1 ratio or 1:2 ratio to MCAT_CAAC and MCAT_CBBC, respectively, in MeCN at r.t. Preparation of MCAT_CABC through three alternative routes: h Reaction of MCAT_CAAC and MCAT_CBBC in a 1:1 ratio in MeCN at r.t. i Reaction of MCAT_AAAA, MR_BB and MCAT_CCCC in a 1:2:2 ratio in MeCN at r.t. j Reaction of H₂A, H₂B and H₂C in a 1:1:2 ratio with Ag₂O in MeCN at 65 °C.

The single-crystal X-ray diffraction (SCXRD) structure of MCAT_AAAA is shown in Fig. 3a. The two identical binuclear MR_AAs are interlocked, constituting a Hopf link topology. SCXRD analysis revealed efficient π···π stacking interactions between the four phenazine rings, as inferred from the centroid−centroid separations of 3.4, 3.3, and 3.4 Å between adjacent phenazine planes. These interactions are the main driving forces for the formation of the interlocked structure. In addition, there is a crossing angle of ca. 85° between the two “inner” phenazine rings, whereas the “outer” adjacent pairwise phenazine rings are nearly parallel.

Fig. 3. X-ray crystal structures of homoleptic and heteroleptic multicomponent MCATs with one, two or three different ligands.

Molecular structure of MCAT_AAAA (a), MCAT_CAAC (b), MCAT_CBBC (c) and MCAT_CABC (d). From left to right: stick representation, space-filling representation, stacking of ligand backbones, and components of the monomeric binuclear macrocycle. Hydrogen atoms, anions and solvents are omitted for clarity. Three letters (A, B, and C) in the space-filling representation are used to denote phenazine-, anthracene-, and pyrene-based building blocks, respectively.

However, when H₂B (with anthracene replaced by phenazine in H₂A) was treated with Ag₂O under the same conditions as described for the assembly of MCAT_AAAA, only the simple binuclear MR_BB was isolated rather than an interlocking topology (Fig. 2, method b). The MR_BB was identified by NMR spectroscopy and ESI mass (Supplementary Figs. 33–37). The single resonance pattern (rather than the common doublet signals of interlocked topologies^47,49) of the ¹H (Supplementary Fig. 33) and ¹³C{¹H} (Supplementary Fig. 34) NMR spectra of the resulting assembly suggested a simpler topology. Moreover, the ¹H DOSY spectrum (Supplementary Fig. 36) as well as the ESI mass spectrum (Supplementary Fig. 37) also confirm the formation of MR_BB (m/z = 630.1871, calcd. for [MR_BB ‒ 2(PF₆^‒)]²⁺ 630.1832). These results indicate that compared with purely hydrocarbon-based aromatic groups, heteroaromatic groups have greater advantages in the construction of interlocked topologies⁵⁰.

Synthesis and characterization of heteroleptic MCATs with two different di-NHC ligands

Inspired by previous work⁵¹, we realized that the charge-transfer interaction between two NHC ligands with similar sizes and opposite electronic centers facilitates the formation of heteroleptic Ag^I-NHC assemblies while inhibiting the formation of homoleptic Ag^I–NHC assemblies. Thus, we employed a similar strategy to construct two-component MCATs.

First, we attempted to assemble the system by reacting imidazo[1,5-a]pyridinium salts H₂A and H₂B in a 1:1 molar ratio with a slight excess of Ag₂O. Subsequent ¹H NMR spectroscopy (Supplementary Fig. 38) revealed that assembly in the presence of two imidazo[1,5-a]pyridinium salts and Ag^I ions failed to yield a single product. Mass spectrometry (Supplementary Fig. 39) indicated the presence of peaks corresponding not only to MR_BB but also to a heteroleptic MCAT, where the ratio of NHC ligands A to B was 3:1. Although we obtained single crystals of MCAT via vapor diffusion, X-ray crystallography confirmed that the structure was only an MCAT formed by the catenation of two MRs (Supplementary Fig. 40). Owing to the similar scattering factors of carbon (C) and nitrogen (N) atoms, the specific arrangement of the ligands could not be definitively assigned. Consequently, we speculated that the products included MR_BB with a mixture of possibly MCAT_ABAA and MCAT_BAAA. To avoid this issue, we considered introducing the NHC ligand C, which has a similar size but an opposite charge center. This ligand was previously shown to enable the high-yield construction of homoleptic MCAT_CCCC with Ag^I (Fig. 2, method c). Considering the electron-donating central core of the proligand H₂C, when it is strategically assembled with a di-NHC ligand containing an electron-deficient aromatic scaffold in the presence of Ag₂O as a metallic precursor, it may facilitate the creation of a new heteroleptic MACT system. This architecture is anticipated to exhibit increased thermodynamic stability and structural complexity, guided by complementary electronic interactions between the donor–acceptor pair and the directional coordination preferences of the Ag^I nodes.

To test this hypothesis, heating a mixture of H₂A and H₂C in a ratio of 1:1 with a slight excess of Ag₂O in acetonitrile indeed resulted in the exclusive formation of heteroleptic [2]catenane with a yield of 92% (Fig. 2, method d). With the slow diffusion of diethyl ether into the acetonitrile solution at ambient temperature, yellow block crystals suitable for SCXRD were obtained. Structural analysis of single crystals unambiguously established the composition and molecular structure of this heteroleptic MCAT with a sequence of CAAC (Fig. 3b). Inspection of the structure reveals that every silver(I) ion simultaneously coordinates NHC ligands A and C linearly to create two heteroleptic MR_ACs, which are further interlocked via mechanical bonds to form heteroleptic MCAT_CAAC. This arrangement results in extensive π···π stacking between the ligand backbones, which are arranged into pyrene–phenazine–phenazine–pyrene stacks with distances of ca. 3.3–3.4 Å between the mean planes of the stacked rings. Interestingly, the selective formation of this heteroleptic MCAT structure with a donor–acceptor–acceptor–donor stack provides a compelling depiction of intramolecular self-sorting that differs from the commonly accepted self-sorting, in which donor and acceptor aromatic moieties are arranged into complementary alternating stacks that are generally regarded as the most favorable arrangement for providing optimum electronic overlap.

After the solid-state molecular structure of MCAT_CAAC was determined, relevant NMR spectroscopic experiments performed to further explore its behavior in solution (Supplementary Figs. 41–45). With the ¹H NMR spectra of ligand precursors H₂A (Fig. 4a) and H₂C (Fig. 4c) and the 2D NMR spectra of MCAT_CAAC, the peaks in the ¹H NMR spectrum of MCAT_CAAC can be clearly identified (Fig. 4d and Supplementary Fig. 41) and are consistent with the characterization of the heteroleptic metalla[2]catenane structure. The ¹H NMR spectrum of MCAT_CAAC (Fig. 4d) shows one set of signals that are equally integrated for each ligand. The total amount of 15 aromatic signals indicated that each ligand maintained its twofold symmetry in the interlocked assembly. Strongly upfield-shifted resonances were observed for the aromatic protons of the phenazine units H9’, H13’ and H10’ (δ = 7.33, 6.74 and 6.66 ppm, respectively), whereas the proton resonances for the pyrene groups H13, H14 and H11 were recorded at δ = 7.81, 7.70 and 7.36 ppm (Fig. 4d), which were attributed to the shielded surroundings inside the cavity of the interlocked architectures. In addition, the ¹H DOSY NMR spectrum of MCAT_CAAC revealed that only one diffusion coefficient D (5.50 × 10^-10 m² s^-1) (Supplementary Fig. 45), suggesting that only one stoichiometric group of the assembly formed.

Fig. 4. NMR spectroscopy and ESI mass analysis.

Partial ¹H NMR spectra (CD₃CN, 400 MHz, 298 K) of (a) H₂A, (b) H₂B, (c) H₂C, (d) MCAT_CAAC, (e) MCAT_CBBC and (f) MCAT_CABC. Experimental (top, blue) and theoretical (bottom, red) ESI mass spectra of [MCAT_CAAC – 3(PF₆)]³⁺ and [MCAT_CAAC – 2(PF₆)]²⁺ (g), [MCAT_CBBC – 3(PF₆)]³⁺ and [MCAT_CBBC – 2(PF₆)]²⁺ (h), and [MCAT_CABC – 3(PF₆)]³⁺ and [MCAT_CABC – 2(PF₆)]²⁺ (i).

Additionally, the assembly of MCAT_CAAC in solution was further confirmed by ESI mass data (Fig. 4g and Supplementary Fig. 46). Prominent peaks at m/z = 906.2563 (calcd for [MCAT_CAAC ‒ 3(PF₆^‒)]³⁺ 906.2271) and 1431.8179 (calcd for [MCAT_CAAC ‒ 2(PF₆^‒)]²⁺ 1431.8231) were observed, which closely matched the anticipated theoretical distributions.

We subsequently examined the combination of H₂B and H₂C at a ratio of 1:1 with Ag₂O under the same conditions as described for the assembly of MCAT_CAAC, which resulted in the formation of heteroleptic MCAT_CBBC with a yield of 90% (Fig. 2, method e). Single crystals of MCAT_CBBC were obtained by slow diffusion of diethyl ether into an acetonitrile solution of the assembly. SCXRD analysis (Fig. 3c) confirmed that the molecular structure of MCAT_CBBC closely resembled that of MCAT_CAAC, where two identical heteroleptic MR interpenetrate to form the heteroleptic MCAT. Moreover, each independent MR is composed of di-NHC ligands B and C, as well as two silver ions. Like the MCAT_CAAC assembly, the pyrene-bridged ligand C was located at the top and bottom, whereas the anthracene-bridged ligand B was located in the middle of the MCAT_CBBC assembly. Triple typical π···π stacking interactions were also observed between the ligand backbones, which are arranged into pyrene–anthracene–anthracene–pyrene stacks with distances of approximately 3.5 Å between the mean planes of the stacked rings.

Like assembly MCAT_CAAC, the structure of the heteroleptic MCAT_CBBC in solution was also investigated (Supplementary Figs. 51–56). Consistent with expectations, the ¹H NMR analysis of MCAT_CBBC (Fig. 4e and Supplementary Fig. 51) confirmed the equimolar presence of ligand precursors H₂B and H₂C (1:1 ratio). In addition, strongly upfield-shifted resonances were also observed for the aromatic protons of the anthracene units H9’, H13’, H10’ and H14’ (δ = 6.91, 6.59, 6.52 and 4.95 ppm, respectively), whereas the proton resonances for the pyrene groups H13, H14, H11 and H10 were recorded at δ = 7.77, 7.69, 7.27 and 6.96 ppm, respectively, consistent with the pyrenyl−anthracene−anthracene−pyrenyl stacking mode in MCAT_CBBC. The ¹H DOSY NMR spectrum (Supplementary Fig. 55) of MCAT_CBBC revealed that all of the aromatic proton signals had a single diffusion constant (D = 5.11 × 10^-10 m² s^-1), indicating the presence of only a unique assembly stoichiometry. Moreover, the formation of heteroleptic MCAT_CBBC in solution was further confirmed by the ESI mass spectrum, which showed a series of peaks at m/z = 904.9167 (calcd for [MCAT_CBBC ‒ 3(PF₆^‒)]³⁺ 904.9002) and 1429.8345 (calcd for [MCAT_CBBC ‒ 2(PF₆^‒)]²⁺ 1429.8326), which is consistent with the theoretical isotopic distributions (Fig. 4h and Supplementary Fig. 56).

Considering the labile nature of the Ag–C_NHC bond and the efficient assembly of two-component MCATs from NHC ligands and Ag^I ions, we subsequently performed experiments to investigate whether these heteroleptic systems can be formed by the rearrangement of homoleptic assemblies. Indeed, mixing MCAT_AAAA and MCAT_CCCC at a ratio of 1:1 resulted in rapid (few seconds) conversion to heteroleptic MCAT_CAAC (Fig. 2, method f), as suggested by the ¹H NMR spectrum (Supplementary Fig. 47). Interestingly, although the reaction of the proligand H₂B with Ag₂O produced only the binulear MR_BB rather than MCAT, MCAT_CCCC and MR_BB could still be rapidly converted to MCAT_CBBC when it was mixed in CD₃CN at a ratio of 1:2 (Fig. 2, method g), as revealed by ¹H NMR spectroscopy (Supplementary Fig. 57). These fusion reactions indicate that heteroleptic MCAT_CAAC and MCAT_CBBC, which are selectively formed through integrative self-sorting, represent thermodynamically favorable reaction products.

To investigate whether the structures of MCAT_CAAC and MCAT_CBBC varied with respect to the concentration and temperature of the solution, concentration dilution and variable-temperature experiments were subsequently performed. Variable-concentration ¹H NMR spectra of MCAT_CAAC (Supplementary Fig. 48) and MCAT_CBBC (Supplementary Fig. 58) in CD₃CN revealed that only the peak intensities gradually decreased as the concentration was diluted from 6.0 to 0.1 mM, while both the peak count and integral ratios remained largely unchanged. Variable-temperature ¹H NMR spectra of MCAT_CAAC (Supplementary Fig. 49) and MCAT_CBBC (Supplementary Fig. 59) in CD₃CN revealed that the number, shape and chemical shifts of signals changed negligibly over the range of −20 ~ 65 °C. These observations indicate that both MCAT_CAAC and MCAT_CBBC can be maintained without structural changes in the acetonitrile solution. Furthermore, NMR titration studies of MCAT_CAAC (Supplementary Fig. 50) and MCAT_CBBC (Supplementary Fig. 60) revealed that gradually changing the solvent from CD₃CN to DMF-d₇ resulted exclusively continuous shifts of their characteristic proton resonances. These shifts are consistent with expected solvent effects, indicating that both MCAT_CAAC and MCAT_CBBC maintain their structural integrity in the presence of DMF.

Synthesis and characterization of heteroleptic MCAT_CABC with three different di-NHC ligands

We have shown the construction of heteroleptic MCATs (with two different di-NHC ligands) either by a one-pot strategy with Ag₂O and the corresponding individual di-NHC precursors or via catenane–to–catenane transformations from homoleptic precursors. Notably, in contrast to the assembly from the individual ligands, the conversion of homoleptic assemblies to heteroleptic MCATs can be accomplished in a few seconds at room temperature, presumably because no deprotonation step is required in supramolecular fusion.

On the basis of these results, we were interested in whether the structural complexity of the system could be further increased by further fusing two heteroleptic MCATs. We carefully inspected heteroleptic MCAT_CAAC and MCAT_CBBC and determined that not only do they have the same pyrene-bridged external ligand, but their internal ligands differ by only two atoms. Considering the similarities in the structure and composition of MCAT_CAAC and MCAT_CBBC, we speculated that their fusion can result in heteroleptic MCATs with more sequences. When two heteroleptic MCAT_CAAC and MCAT_CBBC were mixed at room temperature at a ratio of 1:1 in CD₃CN (Fig. 2, method h), a complex new set of signals could be observed in the NMR spectra (Fig. 4f and Supplementary Figs. 61–64), indicating the formation of a new assembly through self-sorting, denoted MCAT_CABC. In addition, the ¹H DOSY NMR spectrum of the reaction mixture revealed that there was a single diffusion coefficient (D = 5.25 × 10^-10 m² s^-1), suggesting that only one stoichiometric group of the assembly formed (Supplementary Fig. 65). The ESI mass data clearly revealed that the product with the formula [Ag₄(A)(B)(C)₂] formed, and the peaks at m/z = 905.5603 and 1430.8234 could be attributed to the {MCAT_CABC ‒ 3(PF₆^‒)}³⁺ and {MCAT_CABC ‒ 2(PF₆^‒)}²⁺ species (Fig. 4i and Supplementary Fig. 66).

Single crystals of the MCAT_CABC assembly suitable for SCXRD analysis were obtained by slow diffusion of diethyl ether into the solution (V_MeCN/V_DMSO = 15:1) of the compound. Notably, SCXRD analysis (Fig. 3d) unambiguously confirmed the MCAT structure of MCAT_CABC, comprising three distinct di-NHC ligands. We note that this is probably the architecture of MCAT with the highest degree of ligand differentiation and precisely defined structure achieved to date. Moreover, the solid-state structure of the MCAT_CABC assembly revealed that one MR_AC consists of phenazine-bridged di-NHC ligand A and pyrene-bridged di-NHC ligand C connected by two silver(I) ions, whereas the other MR_BC is supported by ligand C and anthracene-bridged ligand B connected by two silver ions. Notably, two pyrene units are located outside the cavity of the two interlocked metallacycles, while anthracene and phenazine units are threaded through the second interlocked metallacycle, resulting in a pyrene–anthracene–phenazine–pyrene parallel arrangement. In the solid structure of MCAT_CABC, π···π stacking interactions are observed between the pyrene and anthracene groups with an interplanar distance of 3.4 Å, in addition to donor‒acceptor interactions between the phenazine unit and the adjacent pyrene or anthracene group, as inferred from the interring distances of 3.3–3.4 Å. These interactions are the main driving forces for the formation of the interlocked structure.

Further experiments were performed to investigate the stability of heteroleptic MCAT_CABC relative to those of the homoleptic MCAT_AAAA, MCAT_CCCC, and MR_BB. A mixture of MCAT_AAAA (1 equiv), MR_BB (2 equiv) and MCAT_CCCC (2 equiv) was allowed to equilibrate in CD₃CN at room temperature (Fig. 2, method i). Analyses of the reaction mixture by ¹H NMR were consistent with the presence of MCAT_CABC as the unique product in solution (Supplementary Fig. 67b). Alternatively, treating the H₂A, H₂B and H₂C ligands at a ratio of 1:1:2 with Ag₂O in acetonitrile at 65 °C for 12 h (Fig. 2, method j) resulted in the formation of MCAT_CABC, as revealed by ¹H NMR spectroscopy (Supplementary Fig. 67c). These observations indicate that heteroleptic assembly MCAT_CABC is the thermodynamically favored reaction product. The stability of the formed heteroleptic MCAT_CABC was further investigated. Notably, the number, shapes and chemical shifts of the signals changed negligibly in the ¹H NMR spectra of MCAT_CABC either at different concentrations (6.0 − 0.1 mM) (Supplementary Fig. 68) or at a range of temperatures ( − 20 − 65 °C) (Supplementary Fig. 69), revealing the stability of the heteroleptic MCAT structure in acetonitrile solution. Moerever, NMR titration studies (Supplementary Fig. 70) indicate that MCAT_CABC can also maintained without structural changes in the presence of DMF.

Furthermore, with the visual molecular dynamics (VMD) program⁵² and wavefunction software Multifwn 3.8⁵³, an independent gradient model based on Hirshfeld partition (IGMH) analysis^54,55 revealed three IGMH δg^inter isosurfaces corresponding to van der Waals interactions within three heteroleptic MCATs (Supplementary Fig. 71). Both pyrene and phenazine moieties contain π electrons (prerequisite for π···π stacking), the interaction itself stems from dispersion forces between parallel π electron clouds, and the nearly flat shape of the δg^inter isosurfaces between π-conjugated fragments is a hallmark feature of π···π stacking^56,57.

Theoretical study

Unlike MCATs that bear a single type of NHC ligand, those incorporating two or more distinct NHC ligands have different component sequences defined by the arrangement of their ligands. Notably, dimerization of either two identical or two different heteroleptic MRs should yield MCATs with numerous possible sequences. However, the abovementioned experimental results revealed only three specific sequences. To elucidate this selective self-assembly, we conducted theoretical investigations employing semiempirical and DFT calculations.

Beginning with MCAT_CAAC, we focused on its heteroleptic metalla-rectangle building unit, MR_AC. As shown in Fig. 5a, statistical analysis indicated that dimerization of two MR_AC units could yield not only MCAT_CAAC but also other sequence isomers (MCAT_AACC and MCAT_ACCA). Structures of MR_AC and three MCATs were optimized via the semiempirical GFN2-xTB method⁵⁸, followed by DFT calculations for single-point energy calculations via the ωB97X-D hybrid functional⁵⁹ (Fig. 5c and Supplementary Table 1). The calculated binding energies (E_BEs) for the dimerization of the two MR_ACs to the MCATs were –387.5 kJ mol⁻¹ (for MCAT_AACC), –306.0 kJ mol⁻¹ (for MCAT_ACCA), and –417.5 kJ mol⁻¹ (for MCAT_CAAC). Comparisons reveal that MCAT_CAAC is significantly more stable than MCAT_AACC and MCAT_ACCA by approximately 30.0 and 111.5 kJ mol⁻¹, respectively. This stability of MCAT_CAAC is consistent with experimental observations, indicating that the CAAC component sequence is highly favored during the assembly of MCATs. Similarly, the binding energy E_BE (–380.0 kJ mol⁻¹) of MCAT_CBBC was substantially greater than that of MCAT_BBCC (E_BE = –358.0 kJ mol⁻¹) and MCAT_BCCB (E_BE = –290.1 kJ mol⁻¹), confirming the increased stability of the CBBC component sequence isomer that was observed experimentally (Supplementary Fig. 72 and Table 2).

Fig. 5. Schematic presentation for the formation of MCATs from two MRs, and semiempirical and DFT calculations analysis.

a Statistical outcome of three MCATs with different component sequences (AACC, ACCA, and CAAC) via the dimerization of two identical heteroleptic MRs. b Statistical outcome (excluding the narcissistic self-sorting) of four MCATs with different component sequences (ABCC, BACC, ACCB, and CABC) via the assembly of two distinct heteroleptic MRs. c Computational study of the relative stabilities of seven feasible MCATs via the GFN2-xTB method (for optimized structures of MCAT_AACC, MCAT_ACCA, MCAT_CAAC, MCAT_ABCC, MCAT_BACC, MCAT_ACCB, and MCAT_CABC) with DFT calculations for single-point calculations of seven optimized structures using ωB97X-D hybrid functional with the 6-31 G(d,p) basis set for non-metal atoms (C, H, N) atoms and the LANL2DZ basis set for metal atoms (Ag).

Next, we extended the theoretical analysis to explain the exclusive experimental observation of MCAT_CABC among possible MCATs incorporating three distinct NHC ligands. Unlike the abovementioned cases, MCAT_CABC arises from the dimerization of two different heteroleptic metalla-rectangles (MR_AC and MR_BC). Statistical analysis revealed three other possible sequence isomers, excluding homodimers: MCAT_ABCC, MCAT_BACC, and MCAT_ACCB (Fig. 5b). Semiempirical optimization and DFT single-point energy calculations (Fig. 5c and Supplementary Table 3) revealed that compared with MCAT_ABCC (E_BE = –372.6 kJ mol⁻¹), MCAT_BACC (E_BE = –376.1 kJ mol⁻¹), and MCAT_ACCB (E_BE = –298.3 kJ mol⁻¹), MCAT_CABC has a significantly higher binding energy E_BE (–400.3 kJ mol⁻¹). This finding indicates the excellent stability of the CABC component sequence, which is consistent with its exclusive experimental formation.

Near-infrared photothermal conversion study

Although the prevalence of π···π stacking interactions in supramolecular assemblies constitutes a key stabilization mechanism, it further facilitates nonradiative energy migration and photothermal conversion processes. We have demonstrated that different conjugate planes can be integrated into a single discrete supramolecular assembly by a metal‒carbon bond-mediated supramolecular fusion strategy. To reliably determine the effect of different conjugate planes and stacking modes on the photothermal conversion efficiency, the near-infrared photothermal conversion properties of MR_BB and five MCATs with different component sequences (AAAA, CCCC, CAAC, CBBC, and CABC) were investigated.

As shown in Figs. 6a, b, the three heteroleptic MCATs experienced different and obvious warming processes. Specifically, a solution of MCAT_CBBC (1.0 mM in CH₃CN, 1.0 mL) experienced a notable increase in temperature from 22.8 to 37.3 °C at 808 nm at 1.5 W cm⁻². Intriguingly, a greater temperature increase was detected for MCAT_CABC (ΔT = 30.1 °C, from 22.4 to 52.5 °C) than for MCAT_CBBC (ΔT = 14.5 °C, from 22.8 to 37.3 °C). A 39.7% photothermal conversion efficiency was calculated for MCAT_CBBC (Supplementary Fig. 78), and a 46.5% efficiency was calculated for MCAT_CABC (Supplementary Fig. 79). Compared with MCAT_CBBC and MCAT_CABC, MCAT_CAAC resulted in greater temperature differences (ΔT = 49.3 °C, from 23.4 to 72.7 °C). According to the equations shown in the ESI, a 79.6% efficiency was calculated for MCAT_CAAC (Supplementary Fig. 77). Analysis of the solid-state structures reveals that although all three heteroleptic MCATs have the same number of π···π stacking interactions, compared with MCAT_CBBC, MCAT_CAAC and MCAT_CABC exhibit stronger stacking strengths. In addition, in the structure of MCAT_CAAC, two electron-deficient phenazine skeletons are adjacent to the electron-rich pyrene conjugated groups and arranged in parallel, which is more conducive to electron transport, resulting in a higher photothermal conversion efficiency for MCAT_CAAC.

Fig. 6. NIR photothermal conversion analysis.

a NIR thermal images of five MCATs with different component sequences (AAAA, CCCC, CAAC, CBBC, and CABC) and MR_BB under 808 nm laser irradiation. b Photothermal conversion curves of the six topologies. c Absorption in the near-infrared region of the six topologies (λ = 700 to 900 nm), the UV−vis absorption spectra (λ = 300 to 900 nm) was shown in Supplementary Fig. 73.

In contrast, the solution temperatures of the homoleptic assemblies MCAT_AAAA, MR_BB and MCAT_CCCC increased by 28.1 °C, 8.6 °C, and 13.9 °C, respectively (Figs. 6a, b and Supplementary Figs. 74–76), which were significantly lower than the temperature increases of the corresponding heteroleptic MCATs. This result further confirms that the near-infrared photothermal conversion performance of the materials can be effectively regulated by changing the ratio and arrangement of different π-conjugated skeletons in the metalla[2]catenane structure. Furthermore, we recorded the electron paramagnetic resonance (EPR) spectra of MR_BB and MCATs. The strong signals indicate the presence of unpaired electrons, which is consistent with the charge-transfer interaction in the ground state. After irradiation, the EPR signals of MCAT_AAAA, MR_BB, MCAT_CCCC, MCAT_CAAC, MCAT_CBBC and MCAT_CABC increased by 12.9, 1.2, 2.4, 16.3, 2.4 and 13.9 times, respectively (Supplementary Figs. 80–85). This observation is consistent with the established conclusion that under 808 nm laser irradiation, the difference in the strong EPR signal with 808 nm laser irradiation of MCAT_CAAC indicates its excellent NIR photothermal conversion efficiency.

Discussion

We have demonstrated that heteroleptic metalla-[2]catenanes with stereoscopic layered arrangements provide a versatile platform for exploring how spatial molecular coding guides emergent functions. By selectively constructing sequence-defined assemblies and correlating their structures with thermodynamic stability and photothermal performance, this work establishes a direct link between motif arrangement and system-level properties. In addition to revealing sequence–function relationships in multicomponent assemblies, our findings highlight molecular coding as a guiding principle for the bottom-up design of supramolecular materials.

Extending this strategy to other spatial organizations and functional motifs could enable systematic decoding of molecular information in increasingly complex assemblies. These efforts may uncover new modes of emergent behavior, from electron transport and photonic activity to adaptive mechanical responses. Moreover, embedding coding principles into supramolecular design could reshape how we conceive and engineer functional matter, bridging molecular information with next-generation materials, devices, and energy technologies.

Methods

Synthesis

The ligands H₂A, H₂B and H₂C were synthesized from 2,7-dibromophenazine, 2,6-dibromoanthracene, and 1,6-dibromopyrene, respectively; detailed methods and characterizations of H₂A, H₂B and H₂C are provided in the Supplementary Information. The homoleptic assemblies MCAT_AAAA, MR_BB and MCAT_CCCC were synthesized using Ag₂O and ligands H₂A, H₂B and H₂C, respectively; the heteroleptic assemblies MCAT_CAAC, MCAT_CBBC and MCAT_CABC were synthesized either via direct assembly of NHC ligands with Ag(I) nodes or structural transformation from homoleptic precursors trough a supramolecular fusion strategy; detailed methods and characterizations of MCAT_AAAA, MR_BB, MCAT_CCCC, MCAT_CAAC, MCAT_CBBC and MCAT_CABC are provided in the Supplementary Information.

X-ray crystallography

Single crystals of MCAT_AAAA, MCAT_CAAC, MCAT_CBBC and MCAT_CABC suitable for X-ray diffraction were obtained at room temperature. X-ray intensity data for MCAT_AAAA, MCAT_CBBC, and MCAT_CABC were collected at T = 240 K and those for MCAT_CAAC at T = 180 K on a Bruker D8 Venture system and used Mo K_α radiation (λ = 0.71073 Å). Using Olex2⁶⁰, the structures were solved with the SHELXT-1997^61,62 structure solution program using direct methods and refined with the SHELXL⁶³ refinement package using least-squares minimization. All non-hydrogen atoms were refined with anisotropic displacement parameters, and hydrogen atoms were placed in geometrically calculated positions. With these data, the disordered solvent molecules that could not be restrained properly were removed using the SQUEEZE method. The ORTEP-style illustration of four structures can be seen in Supplementary Figs. 86 (for MCAT_AAAA), 87 (for MCAT_CAAC), 88 (for MCAT_CBBC) and 89 (for MCAT_CABC), with probability ellipsoids. Details of the X-ray crystallographic refinement of MCAT_AAAA, MCAT_CAAC, MCAT_CBBC and MCAT_CABC can be found in Supplementary Tables 4, 5, 6 and 7, respectively.

Computational details

The computational study using the combination of self-consistent tight-binding (GFN2-xTB)⁵⁸, and density functional theory (DFT) calculations were performed with the Guassian09 quantum chemistry package⁶⁴. The geometries of MRs and MCATs were optimized using semi-empirical GFN2-xTB method implemented with D4 model⁶⁵. The relative stability of feasible supramolecular structures was investigated through single-point calculations using ωB97X-D hybrid functional⁵⁹ with the 6-31 G(d,p)⁶⁶ basis set for non-metal atoms (C, H, N) atoms and the LANL2DZ⁶⁷ basis set for metal atoms (Ag). Acetonitrile was used as a solvent for calculations (SMD model)⁶⁸.

Near-infrared photothermal conversion research

To guarantee same amount of conjugated-π area, the applied molar ratio of the six topologies MCAT_AAAA / MR_BB / MCAT_CCCC / MCAT_CAAC / MCAT_CBBC / MCAT_CABC was 1:2:1:1:1:1. Compound MCAT_AAAA (9.33 mg, 0.003 mmol) was added into a solvent of CH₃CN (3 mL). After the solid dissolved absolutely, 1.0 mL of this solution was taken into a 2.0 mL sample vessel and put into the bright spot of a laser with 808 nm wavelength at 1.5 W/cm². Temperature variation of the solution was detected by an infrared camera. Compound MR_BB (9.29 mg, 0.006 mmol), compound MCAT_CCCC (9.59 mg, 0.003 mmol), compound MCAT_CAAC (9.46 mg, 0.003 mmol), compound MCAT_CBBC (9.45 mg, 0.003 mmol) and compound MCAT_CABC (9.45 mg, 0.003 mmol) were detected with the same procedure as compound MCAT_AAAA.

Author contributions

Y.-F.H. conceived and supervised the project. Y.-W.Z. and H.-N.Z. performed the synthesis of ligands and silver(I)-N-heterocyclic carbene assemblies. Y.-W.Z., H.-N.Z., M.-X.W., and X.L. performed NMR analyses, X-ray crystallographic analysis, electrospray ionization mass spectrometry, theoretical study and near-infrared photothermal conversion study. Y.-W.Z., H.-N.Z. and Y.-F.H. wrote the paper. All authors contributed to the data analysis and discussion.

Peer review

Peer review information

Nature Communications thanks Mandeep Chahal, Jun-Hua Wan and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The authors declare that all data supporting the findings of this study are available within the article and Supplementary Information files, and are also available from the corresponding author upon request. The X-ray crystallographic coordinates for structures have been deposited at the Cambridge Crystallographic Data Center (CCDC) under deposition numbers CCDC-2483191 (MCAT_AAAA), CCDC-2483192 (MCAT_CAAC), CCDC-2483193 (MCAT_CABC), CCDC-2483194 (MCAT_CBBC), respectively. These data can be obtained free of charge via http://www.ccdc.cam.ac.uk/data_request/cif. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-68348-w.