Introduction

The construction of large self-assembly structures, such as metallacycles and metallacages, leveraging the reversibility and directionality of coordination bonds, has attracted sustained and extensive research interest1,2,3,4,5,6. Recent advances in the non-statistical self-assembly7,8,9 of low-symmetry supramolecular coordination complexes (SCCs)10 have enabled precise topological engineering using strategies such as shape-complementary assembly (SCA)11,12,13,14,15, coordination sphere engineering (CSE)16,17,18,19, adjacent backbone interactions20 and hierarchical assembly21. These approaches collectively enable the rational design of sophisticated low-symmetry [MnL2n]-type metallacages, prompting our exploration of cis–trans stereochemistry in [M2L2] metallacycles.

In molecular systems, the geometrical configuration of the C═C bonds often play a critical role in determining the properties of olefins22,23,24,25,26,27,28,29, while the analogous control of coordination-driven assemblies is only beginning to emerge as a transformative paradigm. For instance, square-planar metal centers (e.g., cis-Pt(OTf)2(PEt3)2) can self-assemble with bis-pyridyl ligands to form [M2L2] metallacycles30 with D2 symmetry, topologically akin to ethylene (D2h symmetry), where the bis-pyridyl ligands emulate C = C double bonds and the phosphine substituents mimic hydrogen atoms. This structural analogy raises intriguing questions regarding stereochemistry in such systems. Heteroleptic phosphine coordination at the metal center induces cis–trans isomerization in [M2L2] metallacycles; however, rational stereo control over supramolecular cis–trans isomers remain scarce in the literature, and their dynamic nature poses unique challenges. The combination of new assembly frameworks and material structures, harnessing the dual isomerism of both the metallacycle core and the coordinating ligands, offers a promising platform for creating unique chiral materials and symmetry-breaking functional assembled systems.

In this study, we design and synthesize asymmetric 90o Pt(II) acceptors bearing two distinct phosphine ligands: flexible hydrogen-bond donors (PR3) and rigid, π-rich phosphangulene (Phang)31. These acceptors assemble with bis(4-pyridyl) acetal, which was demonstrated by Stang et al. to be effective for constructing rhomboidal [2 + 2] metallacycles30, to form [Pt2L2] metallacycles with a 1:1 statistical ratio of cis- and trans- isomers, as confirmed by 1H/31P{1H} NMR spectroscopy and density functional theory (DFT) calculations. By exploiting the synergistic supramolecular interactions between modified pyridyl/carboxylate and phosphine ligands, we achieved the stereoselective assembly of cis- or trans- isomers. Comprehensive characterization, including 1H/31P{1H} NMR spectroscopy, investigation of the electrospray ionization time-of-flight mass spectrometry (ESI-TOF-MS), and single-crystal X-ray diffraction (SCXRD), was employed to elucidate the spatial stereochemistry and confirm the formation. Moreover, SCXRD provided unambiguous evidence of their distinct hierarchical structures: trans-metallacycles self-assembled into one-dimensional trans-polyacetylene-like supramolecular polymers through directional Phang interactions in the crystalline form, whereas cis-isomers adopted irregular packing modes due to their steric constraints.

Results

Synthesis of asymmetric 90o Pt(II) acceptors

Initially, we introduced phosphangulene (Phang)32,33 ligands into the Pt(II) acceptor, whose awkward curved shape distinguishes it from the high degree of flexibility of the alkylphosphines ligands, and in addition, it features a distinctive conical shape and an π-rich surface for hierarchical self-assembly34. Phang combined with the conventional phosphine ligand PR3 (R = Me or Et) and PtCl2 according to the ratio of 1 : 1 : 1 reaction and respectively synthesized the cis-PtCl2(PMe3)(Phang) (Pt-A1) and cis-PtCl2(PEt3)(Phang) (Pt-A2), which formed a stable primary assembly structure due to the strong-field phosphine ligand-directed coordination. The 31P{1H} and 1H NMR spectra of the reaction products suggested the formation of a single product with low symmetry. As can be seen in Supplementary Fig. 8 & 11, the 31P{1H} spectra of Pt-A1 show two doublets of approximately equal intensity at δ = -21.54 and -94.58 ppm with accompanying 195Pt satellites (JPt–P = 1693 Hz and JPt–P = 2106 Hz, respectively), and corresponding to two different chemical phosphorus environments, while Pt-A2 demonstrates analogous characteristic peaks. The magnitude of |2J(P-P)| indicates the cis-configuration35. Peaks corresponding to the two phosphine ligands were also observed in the 1H NMR spectra (Supplementary Fig. 7 & 10).

Investigation of the ESI-TOF-MS provided support for the stoichiometry of Pt-A1 and Pt-A2. As shown in Supplementary Fig. 9 & 12, the mass spectra of Pt-A1 exhibited a peak at m/z = 685.9449, corresponding to the positive charge state resulting from the addition of a potassium ion, [M + K+]+, where M represents the intact Pt(II) acceptor Pt-A1. And the mass spectra of Pt-A2, peaks at m/z = 653.0529 and 711.0106, which correspond to the [M − Cl-]+ and [M + Na+]+ species. The experimental isotope pattern matched the theoretical distribution, confirming the molecular composition of Pt-A1 and Pt-A2.

Colorless crystals suitable for X-ray diffraction analysis were obtained through the slow diffusion of diethyl ether into a chloroform solution of Pt-A1 or Pt-A2 (Supplementary Note 4 & 5), which was consistent with the structural characterization in the solution. The crystal structures of Pt-A1 with Cs symmetry and Pt-A2 with C1 symmetry indicate that two phosphine ligands, PR3 (R = Me or Et) and Phang, occupy the two ligand sites of Pt(II), and the two phosphine ligands show a cis-arrangement, with an angle of 98.05o between P-Pt-P and 88.0o between Cl-Pt-Cl in Pt-A1 (Fig. 1a) and an angle of 102.7o between P-Pt-P and 88.4o between Cl-Pt-Cl in Pt-A2 (Fig. 1c). In addition, hydrogen-bonding interactions between the two phosphine ligands were observed in the crystal structure (Supplementary Fig. 39 & 40), further enhance the structural stability.

Fig. 1: ORTEP-style crystal structures of Pt-A1 and Pt-A2.
figure 1

The front view a and side view b of the crystal structure of Pt-A1. The front view c and side view d of the crystal structure of Pt-A2.

Full size image

Self-Assembly of statistical cis–trans isomers M1 & M1’

The self-assembly of the two-component system composed of pyridyl donor LA and Pt-A1 was investigated by adding bis-pyridyl donor LA to the DMSO solution of acceptor Pt-A1 with AgNO3 at a ratio of 1:1, followed by a further 12 h reaction at room temperature (Fig. 2a). 31P{1H} and 1H NMR multinuclear analysis of the reaction mixture indicated the formation of a mixed system of cis-trans isomerism (M1, trans-Pt2LA2 & M1’, cis-Pt2LA2). The 31P{1H} spectra (Fig. 2d, e and Supplementary Fig. 17 & 59) exhibit two sets of doublet of doublets of approximately equal intensity at δ = -25.58 to -25.86 ppm (PMe3) and δ = -96.80 to -96.98 ppm (Phang) with accompanying 195Pt satellites (JPt–P = 1508 Hz and JPt–P = 1891 Hz, respectively), upfield-shifted about 4.81 ppm (PMe3) and 2.31 ppm (Phang) compared to the signal of the staring acceptor Pt-A1 due to coordination with the pyridine rings. The first-order splitting pattern observed in the 31P{1H} NMR spectrum (doublet of doublets) indicates that the acceptor Pt-A1 coordinates with the bis-pyridyl donor in different spatial orientations, with two separate chemical environments for each phosphine ligand in the product. These results can be explained only by the generation of heterogeneous mixed products. In the corresponding 1H NMR spectra of products M1 & M1’ (Fig. 2b), the signals for the protons of the pyridine rings in the donor ligand LA (Fig. 2c) exhibited downfield shifts resulting from the loss of electron density upon coordination of the pyridine nitrogen atoms with the platinum metal center. The characteristic peaks of acetal methylene in the cis-trans isomers can be observed within δ = 4.02 ~ 4.30 ppm (Fig. 2g), respectively. Among them, at δ = 4.21 ~ 4.14 ppm, the splitting into multiplets because of the different chemical environments in which the hydrogen atoms c & c’ in the trans-configuration are located, while at δ = 4.02 and 4.3 ppm, two singlets are observed because of the symmetry of the hydrogen atoms f & f’ in the cis- configuration. The peak areas after integration of f & f’ and c & c’ are equal (Supplementary Fig. 16). Variable-Temperature 1H NMR (VT-NMR) analysis (25 to 50 °C) further confirmed retained structural integrity throughout the thermal cycle (Supplementary Fig. 19). DFT calculations showed that there was no significant energy difference between the two isomers, which exist as a mixture in solution (Fig. 2a, Supplementary Fig. 36, Supplementary Data 1 and Supplementary Data 2 & 3). And the dynamic reversibility of self-assembly renders standard chromatographic separation ineffective.

Fig. 2
figure 2

Synthesis and structural characterize of M1 & M1’. a Synthetic route and energy with DFT calculated of M1 & M1’. Partial 1H NMR spectrum (400 MHz, 298 K) of b M1 & M1’(CD3OD), c LA (CDCl3). 31P{1H} NMR spectra (162 MHz, 298 K) of d M1 & M1’ (CD3OD), e Pt-A1 (CDCl3). f Experimental (red) and theoretical (blue) ESI-TOF-MS spectra of M1 & M1’ [M – 4NO3-]4+. g Characteristic ¹H NMR singals of acetal methylene in the M1 & M1’.

Full size image

The observed isotopic distributions in the ESI-TOF-MS spectra (Fig. 2f and Supplementary Fig. 18) unambiguously support their assignment as [2 + 2] metallacycles. In the mass spectra of mixtures, peaks at m/z = 401.5633 corresponding to the [M  − 4NO3-]4+ were observed, and their isotopic distributions were in good agreement with the theoretical expectation, indicating the formation of [Pt2L2] metallacycles.

Self-Assembly of trans-metallacycle M2

Building upon statistical self-assembly principles and inspired by social-sorting assembly strategies (e.g., SCA11,12,13,14,15, CSE16,17,18,19) and ligand-ligand interactions36,37,38. In this study, we designed strategies for the self-assembly of non-statistical trans-metallacycle. Firstly, methyl modification of the pyridine group on one side of the ligand LA to obtain the ligand LB. A strategy wherein methylated pyridines have been demonstrated to enable selective construction of low-symmetry Pd(II) complexes through steric hindrance between methyl groups38,39,48. This modified ligand LB was subsequently assembled with Pt-A1, and the C–H···π interactions between the methyl groups of LB and the π-electron-rich surface of the Phang unit induced a trans-arrangement at both termini of the assembly, leading to the selective construction of the trans-metallacycle and avoiding statistical isomeric mixtures. (Fig. 3a and Supplementary Fig. 23).

Fig. 3
figure 3

Synthesis and structural characterize of M2. a Synthetic route of M2. Partial 1H NMR spectrum (400 MHz, 298 K) of b M2 (CD3OD), c LB (CDCl3). 31P{1H} NMR spectra (162 MHz, 298 K) of d. M2 (CD3OD) e. Pt-A1 (CDCl3) f proton signal of acetal methylene in M1 & M1’ (top) and M2 (bottom). g The crystal structure of M2 (Hydrogen atoms, counterions, and solvent molecules are omitted for clarity). Theoretical (blue) and experimental (red) ESI-TOF-MS spectra of M2 h. [M – 4NO3-]4+, i [M – 3NO3-]3+.

Full size image

The trans-metallacycle was fully characterized using 1D multinuclear (31P{1H} and 1H) NMR, ESI-TOF-MS, and SCXRD. In the 1H NMR spectra (Fig. 3b), it can be observed that the proton signals of the LB (Fig. 3c) and Pt-A1 are shifted to the downfield after self-assembly, especially the methyl signal of LB, which changes from δ = 2.48 ppm to δ = 2.79 ppm after reaction, which may be attributed to the effect of hydrogen bonding. Meanwhile, at δ = 4.30 ~ 4.02 ppm, compared with the cis-trans isomeric system (M1 & M1’), only the characteristic peak signals c & c’ of acetal methylene belonging to the trans-structure can be peculiarly observed (Fig. 3f). Two broad peaks observed at 7.1 and 7.6 ppm can be attributed to the Phang ligand in the reacting species cis-Pt(NO3)2(PMe3)(Phang) and a trace amount of oxidized Phang ligand formed during the assembly process (Supplementary Fig. 20b, c and d). The 31P{1H} spectrum (Fig. 3d, e and Supplementary Fig. 21) exhibits two sets of doublets of approximately equal intensity at δ = -23.71 ppm (PMe3) and δ = -99.99 ppm (Phang) with accompanying 195Pt satellites (JPt–P = 1501 Hz and JPt–P = 1851 Hz, respectively), indicating that the different phosphine ligands have only one chemical phosphorus environment, respectively, and further confirming that the nonstatistical metallacycle was formed.

The ESI-TOF-MS spectra of M2 was measured to provide information about the self-assemble stoichiometry. In the mass spectrum of M2, peaks at m/z = 415.5790 and 574.7681, which correspond to the [M  − 4NO3-]4+ and [M   − 3NO3-]3+ species, and their isotopic distributions were in good agreement with the theoretical expectation (Fig. 3h, i & Supplementary Fig. 22), indicating the M2 was formed.

Colorless crystals of M2 suitable for X-ray diffraction analysis were grown by the vapor diffusion of diethyl ether into a 1:1 (v/v) DMSO/ethyl acetate solution (Supplementary Note 6), and the solid-state structure corroborated the solution-phase structural analysis. The crystal structure of M2 is shown in Fig. 3g, where two phosphine ligands at the M2 ends show a trans-arrangement, while at the same time, the two ligands LB are coordinated to Pt-A1 in the opposite direction because of the steric crowding caused by the methyl group, which will avoid the formation of unfavorable steric products. Beyond the conventional strategy of employing methyl steric hindrance to build low-symmetry Pd2L4 cages48, the present design leverages an additional supramolecular positioning effect between the methyl groups and the Phang unit. In this process, the methyl group in LB interacts C-H···π with the Phang (d[C-H···π] = 3.0 - 3.5 Å) (Supplementary Fig. 41), which is a key factor for selective self-assembly to produce a single trans-configuration metallacycle, while lowering the energy of the system and facilitating the generation of stable thermodynamic products. 1H NMR (VT-NMR) analysis (25 to 50 °C) further confirmed retained structural integrity throughout the thermodynamic stable trans-metallacycle (Supplementary Fig. 24). Additional DFT calculations on the trans-isomeric pair M2 and M2′ demonstrated a relative energy difference of 21.35 kJ/mol in favor of M2 (Supplementary Fig. 37, Supplementary Data 1 and Supplementary Data 4 & 5). This notable stability indicates that the assembly of metallacycle M2 is governed by thermodynamic control, rather than being driven by kinetic factors such as the trans effect.

Self-Assembly of cis-metallacycle M5

Inspired by the successful assembly of trans-metallacycle, a similar assembly strategy was employed to construct a cis-metallacycle. The other side of the ligand LB was modified to obtain the LC, combined the LA with the Pt-A1, and utilized the hydrogen bond formed by the methyl group and the Phang to induce the generation of the cis-configuration M3 (cis-Pt2LALC) (Supplementary Note 1). Unfortunately, it did not obtain social-sorting configuration M3, and the 31P{1H} spectrum was disordered and could not recognize the characteristic peak signals (Supplementary Fig. 25). The 1H NMR spectra demonstrated characteristic peaks suggesting a tendency to generate self-sorting products trans- & cis-Pt2L2C and trans- & cis-Pt2L2A (Supplementary Fig. 26).

Building on these findings, we employed the charge-separation assembly strategy40 to construct cis-metallacycle by introducing a dicarboxylate ligand (LD). Preliminary experiments revealed severely limited solubility of Pt-A1 in this assembly system, promoting the substitution with Pt-A2 to enhance solvation. This modification enabled the design of cis-metallacycle M4 through the coordination of Pt-A2 with ligands LA and LD. Crucially, the charge separation effect drives electrostatically directed assembly: the positively charged pyridine moiety of LA in M4 engages with π-electron-rich Phang units, thereby enforcing a cis-arrangement of Phang ligands at the assembly termini (Supplementary Note 2).

DFT calculations for the envisaged target configuration M4 and the possible configurations M4’ and M4” during assembly revealed M4 was the most thermodynamically stable configuration (Supplementary Fig. 38, Supplementary Data 1 and Supplementary Data 6, 7 & 8). To validate this prediction, Pt-A2-NO3- was mixed with LA and LD in a 2:1:1 ratio, followed by the addition of water and acetone. After 12 h of stirring at room temperature, all solvents were removed from the clear solution, and acetone was added to the mixture. A clear solution was obtained after 12 h of stirring at room temperature. The 31P{1H} and 1H NMR spectra reflect no formation of the pure cis-configuration M4 we envisioned (Supplementary Fig. 27 & 28), and the acetal methylene at δ = 4.15 - 4.30 in the 1H NMR spectrum, as well as the appearance of characteristic peak signals in the 31P{1H} spectrum, may form mixtures. The system suggests that charge-separation factors alone cannot fully induce stereo-control construction of the cis-metallacycle.

Therefore, LA was exchanged for LC via the synthetic route of M4 to induce the precise construction of the M5 (cis-Pt2LCLD) through the dual effects of hydrogen binding and charge separation. These strategies avoid statistical mixtures (Supplementary Fig. 32). M5 was fully characterized using 31P{1H} and 1H NMR, ESI-TOF-MS, and SCXRD. The 31P{1H} spectrum exhibits two sets of doublets of approximately equal intensity at δ = 11.73 ppm (PEt3) and δ = -102.85 ppm (Phang) with accompanying 195Pt satellites (JPt–P = 1591 Hz and JPt–P = 2027 Hz, respectively), indicating that the different phosphine ligands have only one chemical phosphorus environment, respectively, and confirming that the self-assemble systems has a nonstatistical structure (Fig. 4d). In the 1H NMR spectra (Fig. 4b), like M2, the proton signal shifted to the downfield after self-assembly and the characteristic peak of the acetal methylene group showed a singlet at δ = 4.22 ppm, referring to the previous analysis, which further demonstrated that our expected cis-configuration was successfully constructed. The ESI-TOF-MS spectrum of M5 was measured to provide information about the stoichiometry (Fig. 4f). In the mass spectrum of M5, peaks at m/z = 814.1646 corresponding to the [M  − 2NO3-]2+ were observed, and their isotopic distributions were in good agreement with the theoretical predictions.

Fig. 4
figure 4

Synthesis and structural characterize of M5. a Synthetic route of M5.Partial 1H NMR spectrum (400 MHz, 298 K) of b M5 (CD3OD), c LC (CDCl3). 31P{1H} NMR spectra (162 MHz, 298 K) of d M5 (CD3OD), e Pt-A2 (CD2Cl2). f Theoretical (blue) and experimental (red) ESI-TOF-MS spectra of M5 [M – 2NO3-]2+. g The crystal structure of M5 (Hydrogen atoms, counterions, and solvent molecules are omitted for clarity.

Full size image

Colorless crystals of M5 suitable for X-ray diffraction analysis were grown by vapor diffusion of diethyl ether into a methanol solution (Supplementary Note 7), and the solid-state structure corroborated the solution-phase structural analysis. The crystal structure of M5 is shown in Fig. 4g, which consists of one pyridyl ligand, one dicarboxylate ligand, and two cis-[(PEt3)(Phang)]Pt(II) centers. This indicates the formation of a heteroligated N, O-coordination motif with one pyridyl group and one carboxylate moiety per Pt(II) center. The two phosphine ligands at the M5 end exhibit a cis-arrangement. The charge separation factor and the C–H···π weak interactions formed by the methyl group and the π electrons in the Phang together induce the Phang ligand to show a cis-arrangement at the M5 ends. In the crystal structure, we can also observe multiple C–H···π and C–H···O hydrogen bonding of the two phosphine ligands with the pyridine and dicarboxylate ligands, which is the reason for the stability of the M5 (Supplementary Fig. 42), and VT-NMR analysis (25 to 50 °C) further confirms the configuration stability (Supplementary Fig. 33).

Hierarchical Self-Assembly

During the process of M2 and M5 recrystallization, the stacking mode showed a representative offset clamshell pair33 between Phang to induce a hierarchical self-assembly during structure of the metallacycles. In hierarchical self-assemble structures of M2 and M5, the former formed supramolecular polymers through this distinctive stacking mode (Fig. 5a) and then expanded to supramolecular frameworks (Supplementary Fig. 46); the latter formed supramolecular meshwork through stacking between Phang and C–H···π interactions between triethylphosphine and Phang (Fig. 5b and Supplementary Fig. 45), but the M5 adopted irregular packing modes due to their steric constraints. (Supplementary Fig. 47). This reflected the differences in the hierarchical assembly stacking patterns of the cis-trans isomeric system, offers fundamental insights into the stereo-controlled supramolecular design.

Fig. 5: Hierarchical self-assembly structure.
figure 5

a M2. b M5. Counterions and solvent molecules are omitted for clarity.

Full size image

Discussion

At the platinum center, two kind of phosphine ligands with distinct properties were used: trialkylphosphines (e.g., PEt₃ or PMe₃) and phosphangulene (Phang). The trialkylphosphines, featuring saturated alkyl chains, are structurally flexible and provide multiple C–H groups as hydrogen-bond donors. In contrast, the Phang ligand is relatively rigid and capable of engaging in specific non-covalent interactions, such as π–π stacking and C–H⋯π contacts. Through rational design and synthesis of pyridine-based donors (LB and LC), combined with their heterotypic coordination preference (pyridine and carboxylate) toward the unsymmetrical platinum acceptor, the self-assembly process was selectively directed to form either cis- or trans-configured metallacycles.

The trans-configuration was precisely directed by the synergistic cooperation of two mechanisms: supramolecular interactions(C-H···π) between the methyl group in the LB and Phang, as well as spatial complementarity with ligand LB. It is noteworthy that employing the all-triethylphosphine platinum acceptor, cis-Pt(NO3)2(PEt3)2, in reaction with ligand LB led to a complex reaction mixture rather than the exclusive formation of metallacycle 1 (Supplementary Note 3). The product distribution potentially includes species 1, 2 and other possibilities, as depicted. Definitive spectral assignment proved challenging in both the 1H and 31P{1H} NMR spectra (Supplementary Fig. 34 and 35). Similarly, the cis-configuration was accurately constructed through the combined effects of analogous supramolecular interactions between the methyl group in the LC and Phang and an additional charge-separation mechanism.

Conclusion

In this paper, we demonstrate that asymmetric 90° Pt(II) acceptors direct the formation of [Pt2L2] metallacycles exhibiting cis–trans isomerization. Based on this phenomenon, we have developed a method for the stereoselective synthesis of discrete metallacycles with specific cis or trans configurations. Self-assemble configurations were fully characterized by NMR, ESI-TOF-MS, DFT calculations and SCXRD. Multiple supramolecular interactions were observed in their crystal structures, suggesting that the cis–trans configurations can be accurately constructed on the base of synergistic supramolecular interactions. Moreover, hierarchical self-assembly in the crystalline state reveals stark structural divergences between cis- / trans- metallacycles. This work explains the self-assembly mechanism of supramolecular cis–trans isomeric metallacycles, deepens the concept of self-assembly based on asymmetric acceptors, opens new avenues for the precise construction of complex supramolecular isomeric systems, and offers fundamental insights into the hierarchical assembly behaviors of supramolecular isomeric systems.

Methods

 General procedure

All reagents and deuterated solvents were used as purchased without further purification. Ligands LA was prepared according to the literature procedures41,42, and phosphangulene (Phang) was prepared according to the literature procedures32,33,34. Column chromatography was conducted using silica gel (200 − 300 mesh). 31P{1H}, 1H, 13C NMR spectra data were recorded at 298 K on a Bruker Avance 400 MHz and 600 MHz spectrometer. 1H NMR chemical shifts were recorded relative to residual solvent signals. Mass spectra were recorded on the Waters Synapt G2 tandem mass spectrometer using electrospray ionization, Thermo Scientific Q Exactive mass spectrometer using electrospray ionization and Agilent 6545 Q-TOF using electrospray ionization. X-ray diffraction analysis was conducted a Bruker D8 VENTURE PHOTON II MetalJet, in which crystals were frozen in paratone oil inside a cryoloop under a cold stream of N2. An empirical absorption correction using SADABS was applied for all data. The structures were solved and refined to convergence on F2 for all independent reflections by the full-matrix least squares method using the OLEX2 1.543.

Synthesis of LB

To a dry solution of n-BuLi (2.5 M in hexanes; 4.00 mL, 10 mmol, 1.0 equiv) in Et2O (50 mL) at –78 oC under N2, 4-bromo-2,6-dimethylpyridine (1.32 mL, 10.0 mmol, 1.0 equiv) was added dropwise. After the reaction was stirred for 30 min at the same temperature, a solution of methyl isonicotinate (1.18 mL, 10 mmol, 1.0 equiv) in dry Et2O (5 mL) was added slowly. The reaction was kept at –78 oC for another 30 min, and then warmed slowly to –20 oC over 3 h before quenched with 6 M NaOH(aq). The mixture was then warmed to room temperature, and extracted with ethyl acetate. The organic extracts were combined, dried over Na2SO4, filtered, and evaporated in vacuo. To a vigorously stirred solution of the former crude product and 2-bromoethanol (0.71 mL; 10 mmol) in DMF (15 mL) and THF (5 mL) at -60 oC under argon, solution of t-BuOK (1.68 g; 15 mmol) in DMF (8 mL) was added dropwise for 30 min. Then mixture was stirred for 90 min and solution of 5% NH4Cl(aq). (10 mL), brine (20 mL) and water were added (water was necessary to dissolve inorganic precipitates). Mixture was extracted with ethyl acetate (5×50 mL), combined organic phases were washed with brine (3×70 mL) and dried with Na2SO4, and then purified by silica gel column (n-hexane/ethyl acetate = 4:1, v/v) to obtain LB (827.8 mg, 32.3%) of the white solid. The 1H NMR spectrum of LB is shown in Supplementary Fig. 1 & 48.1H NMR (400 MHz, CDCl3) δ 8.59 – 8.53 (m, 2H), 7.41 – 7.35 (m, 2H), 7.04 (s, 2H), 4.05 – 3.99 (m, 4H), 2.48 (s, 6H). The 13C NMR spectrum of LB is shown in Supplementary Fig. 2 & 62. 13C NMR (151 MHz, CDCl3) δ (ppm): 159.29, 152.87, 152.57, 150.43, 122.35, 118.83, 108.04, 66.59, 23.82. ESI-HRMS is shown in Supplementary Fig. 3: m/z = 257.1282 [LB + H+]+, calcd. for [C15H16N2O2 + H+]+, 257.1285.

Synthesis of LC

Ligand LC is synthesized in the same procedure as ligand LB. The 1H NMR spectrum of LC is shown in Supplementary Fig. 4 & 49. 1H NMR (400 MHz, CDCl3) δ 7.07 (s, 4H), 4.05 (s, 4H), 2.53 (s, 12H). The 13C NMR spectrum of LC is shown in Supplementary Fig. 5 & 63. 13C NMR (151 MHz, CDCl3) δ (ppm): 159.18, 153.13, 118.83, 108.15, 66.47, 23.80. ESI-HRMS is shown in Supplementary Fig. 6 : m/z = 285.1596 [LC + H+]+, calcd. for [C17H20N2O2 + H+]+, 285.1598.

Synthesis of Pt-A1

PtCl2 (266 mg, 1 mmol) and Phang (304 mg, 1 mmol) were weighed sequentially in a Schlenk reaction flask, ultra-dry CH2Cl2 (30 mL) was taken, and after repeated freezing and thawing with liquid nitrogen for strict deoxidation, it was transferred to the Schlenk flask with a double-ended needle, protected by N2, and added to the reaction flask with PMe3 (1 M in THF, 1.1 mL) The reaction was carried out at room temperature and protected under the darkness for 3 days. At the end of the reaction, the solvent was removed by nitrogen flow. The residue was added into the crude silica gel powder and mixed well, and then purified by silica gel column (dichloromethane/n-hexane/ethyl acetate = 2:1:1, v/v/v) to obtain Pt-A1 (407 mg, 61.2%) of white powdery solid. The 1H NMR spectrum of Pt-A1 is shown in Supplementary Fig. 7 & 50. 1H NMR (400 MHz, CDCl3) δ 7.42 (t, J = 8.3 Hz, 3H), 7.16 (dd, J = 8.3, 3.7 Hz, 6H), 1.46 (d, J = 11.7 Hz, 9H). The 31P{1H} NMR spectrum of Pt-A1 is shown in Supplementary Fig. 8 & 56. 31P{1H} NMR (162 MHz, CDCl3) δ(ppm): -21.54 (d, 2JP-P = 18.00 Hz, 195Pt satellites, 1JPt-P = 1693.24 Hz), -94.58 (d, 2JP-P = 18.00 Hz, 195Pt satellites, 1JPt-P = 2106.75 Hz). ESI-HRMS is shown in Supplementary Fig. 9 : m/z = 685.9360 [M + K+]+, calcd. for [C21H18Cl2O3P2Pt + K+]+, 685.9449.

Synthesis of Pt-A2 & Pt-A2-NO3 -

PtCl2 (266 mg, 1 mmol) and Phang (304 mg, 1 mmol) were weighed sequentially in a Schlenk reaction flask, ultra-dry CH2Cl2 (30 mL) was taken, and after repeated freezing and thawing with liquid nitrogen for strict deoxidation, it was transferred to the Schlenk flask with a double-ended needle, protected by N2, and added to the reaction flask with PEt3 (0.15 mL) The reaction was carried out at room temperature and protected under the darkness for 3 days. At the end of the reaction, the solvent was removed by nitrogen flow. The residue was added into the crude silica gel powder and mixed well, and then purified by silica gel column (dichloromethane/n-hexane/ethyl acetate = 2:2:1, v/v/v) to obtain Pt-A2 (482 mg,70%) of white powdery solid. The 1H NMR spectrum of Pt-A2 is shown in Supplementary Fig. 10 & 51. 1H NMR (400 MHz, CD2Cl2, 25°C) δ 7.45 (t, J = 8.3 Hz, 3H), 7.18 (dd, J = 8.3, 3.6 Hz, 6H), 1.70 (dq, J = 10.4, 7.5 Hz, 6H), 1.04 (dt, J = 17.9, 7.6 Hz, 9H). The 31P{1H} NMR spectrum of Pt-A2 is shown in Supplementary Fig. 11 & 57. 31P{1H} NMR (162 MHz, CD2Cl2, 25°C) δ(ppm): 12.16 (d, 2JP-P = 17.60 Hz, 195Pt satellites, 1JPt-P = 1664.20 Hz), -95.90 (d, 2JP-P = 17.60 Hz, 195Pt satellites, 1JPt-P = 2106.10 Hz). ESI-TOF-MS is shown in Supplementary Fig. 12 : m/z = 653.0545 [M – Cl-]+ and 711.0108 [M + Na+]+, calcd. for m/z = 653.0529 [C24H24Cl2O3P2Pt – Cl-]+ and 711.0106 [C24H24Cl2O3P2Pt + Na+]+.

Pt-A2 (200 mg, 0.29 mmol) and AgNO3 (147.79 mg, 0.87 mmol, excess) were placed in a 10 mL reaction flask. Subsequently, ultra-dry CH2Cl2 (5 mL) were added, the vial was sealed, and the contents were stirred for a period of 24 h. The reaction vessel should be wrapped in tinfoil and protected from light. Upon completion of the reaction, the reaction solution should be subjected to centrifugation, and the upper layer was filtered through a degreased cotton wool. The dichloromethane should then be removed under reduced-pressure evaporation to obtain Pt-A2-NO3- (cis-Pt(NO3)2(PEt3)(Phang), 210.7 mg, 98%) as a light yellow powdery solid. The 1H NMR spectrum of Pt-A2-NO3- is shown in Supplementary Fig. 13 & 52. 1H NMR (400 MHz, CD2Cl2) δ 7.54 (t, J = 8.3 Hz, 3H), 7.25 (dd, J = 8.3, 4.0 Hz, 6H), 1.66 (dq, J = 10.5, 7.6 Hz, 6H), 1.11 (dt, J = 18.5, 7.6 Hz, 9H). The 31P{1H} NMR spectrum of Pt-A2 NO3- is shown in Supplementary Fig. 14 & 58. 31P{1H} NMR (162 MHz, CD2Cl2, 298 K) δ(ppm): 9.99 (d, 2JP-P = 27.40 Hz, 195Pt satellites, 1JPt-P = 1780.02 Hz), -101.56(d, 2JP-P = 27.40 Hz, 195Pt satellites, 1JPt-P = 2298.12 Hz). ESI-TOF-MS is shown in Supplementary Fig. 15 : C24H24N2O9P2Pt, Calcd m/z = 681.0862 [M – NO3-]+, Found m/z = 681.0948 [M – NO3-]+.

Self-Assembly of M1 & M1’ (trans & cis- Pt2LA 2)

The metallacycles M1 & M1’ was synthesized by a one-pot method. Pt-A1 (12.926 mg, 0.02 mmol), AgNO3 (13.58 mg, 0.08 mmol) were placed in a 2-dram vial, followed by addition of 1.5 mL of DMSO, and the vial was then sealed with Teflon tape at room temperature under the darkness for 12 h. And then ligand LA (4.56 mg,0.02 mmol) was added to the vial which was then sealed at room temperature for 12 h. The reaction mixture was filtered, and then ethyl ether (8 mL) and ethyl acetate (5 mL) were added to give a precipitate, which was collected by centrifugation to metallacycles M1 & M1’ as a white solid. The 1H NMR spectrum of M1 & M1’ is shown in Supplementary Fig. 16 & 53. 1H NMR (400 MHz, CD3OD) δ 8.94 – 8.79 (m, 4H), 8.54 (ddd, J = 6.7, 5.1, 2.9 Hz, 4H), 7.78 (ddd, J = 8.0, 6.7, 1.6 Hz, 4H), 7.72 – 7.65 (m, 6H), 7.65 – 7.61 (m, 4H), 7.31 (ddd, J = 8.3, 5.9, 4.1 Hz, 12H), 4.30 (s, 2H), 4.22 – 4.19 (m, 2H), 4.17 – 4.14 (m, 2H), 4.02 (s, 2H), 1.45 (dd, J = 12.3, 8.0 Hz, 18H). The 31P{1H} NMR spectrum of M1 & M1’ is shown in Supplementary Fig. 17 & 59. 31P{1H} NMR (162 MHz, CD3OD, 298 K) δ(ppm): -25.72 (dd, 2JP-P = 22.20 Hz, 195Pt satellites, 1JPt-P = 1508.60 Hz), -96.89 (dd, 2JP-P = 22.20 Hz, 195Pt satellites, 1JPt-P = 1891.01 Hz). ESI-TOF-MS is shown in Supplementary Fig. 18 : theoretical m/z = 401.5633 [C68H60N4O10P4Pt2]4+, experimental m/z = 401.5634 [M - 4NO3-]4+.

Self-Assembly of M2 (trans-Pt2LB 2)

The metallacycles M2 was synthesized by a one-pot method. Pt-A1 (12.926 mg, 0.02 mmol), AgNO3 (6.79 mg,0.04 mmol) were placed in a 2-dram vial, followed by addition of 1.5 mL of DMSO, and the vial was then sealed with Teflon tape at room temperature under the darkness for 12 h. And then ligand LB (5.126 mg,0.02 mmol) was added to the vial which was then sealed at room temperature for 12 h. The reaction mixture was filtered, and then ethyl ether (8 mL) and ethyl acetate (5 mL) were added to give a precipitate, which was collected by centrifugation to metallacycle M2 as a white solid (15.7 mg, 82%). The 1H NMR spectrum of M2 is shown in Supplementary Fig. 20a & 54. 1H NMR (400 MHz, CD3OD) δ 8.84 – 8.74 (m, 4H), 7.84 (dd, J = 6.8, 1.4 Hz, 4H), 7.73 (t, J = 8.3 Hz, 6H), 7.45 (s, 4H), 7.35 (dd, J = 8.3, 4.1 Hz, 12H), 4.21 – 4.11 (m, 8H), 2.79 (s, 12H), 1.49 (d, J = 12.3 Hz, 18H). The 31P{1H} NMR spectrum of M2 is shown in Supplementary Fig. 21 & 60. 31P{1H} NMR (162 MHz, CD3OD, 298 K) δ(ppm): -23.71 (d, 2JP-P = 23.8 Hz, 195Pt satellites, 1JPt-P = 1501.02 Hz), -99.99 (d, 2JP-P = 24.20 Hz, 195Pt satellites, 1JPt-P = 1851.01 Hz). ESI-TOF-MS is shown in Supplementary Fig. 22 : theoretical m/z = 415.5790 [C72H68N4O10P4Pt2]4+ and 574.7681 [C72H68N5O13P4Pt2]3+, experimental m/z = 415.5789 [M – 4NO3-]4+, m/z = 574.7666 [M – 3NO3-]3+.

Self-assembly of M5 (cis-Pt2LCLD)

Pt-A2-NO3- (7.42 mg, 0.01 mmol), LD (1.05 mg, 0.005 mmol) and ligand LC (1.42 mg, 0.005 mmol) were placed in a 2-dram vial, followed by addition of 0.6 mL of acetone and 0.2 mL of water, and the vial was then sealed with Teflon tape. The mixture was stirred at room temperature for 12 h, and then all solvents were removed by N2 flow and 1 mL acetone was added to the residue to react another 12 h. Then, the mixture was filtered to remove insoluble materials. After concentrating the clear solution to 0.4 mL, the resulting metallacycle M5 was precipitated with diethyl ether, which was collected by centrifugation as a white solid (7.9 mg, 89.5%). The 1H NMR spectrum of M5 is shown in Supplementary Fig. 29 & 55. 1H NMR (600 MHz, CD3OD) δ 7.71 (dd, J = 7.7, 1.8 Hz, 2H), 7.69 (t, J = 8.3 Hz, 6H), 7.45 (s, 4H), 7.33 (dd, J = 8.4, 3.9 Hz, 12H), 7.25 (t, J = 7.7 Hz, 1H), 6.90 (t, J = 1.7 Hz, 1H), 4.22 (s, 4H), 2.68 (s, 12H), 1.90 – 1.84 (m, 12H), 1.20 (dd, J = 18.5, 7.5 Hz, 18H). The 31P{1H} NMR spectrum of M5 is shown in Supplementary Fig. 30 & 61. 31P{1H} NMR (162 MHz, CD3OD, 298 K) δ(ppm): 11.73 (d, 2JP-P = 24.5 Hz, 195Pt satellites, 1JPt-P = 1591.86 Hz), -102.85 (d, 2JP-P = 24.4 Hz, 195Pt satellites, 1JPt-P = 2027.95 Hz). ESI-TOF-MS is shown in Supplementary Fig. 31 : theoretical m/z = 841.1646 [C73H72N2O12P4Pt2]2+, experimental m/z = 841.1640 [M – 2NO3-]2+.

Computational Studies

All calculations were carried out with the Gaussian 16 software. The PBE0 functional44 was adopted for all calculations. For geometry optimization and frequency calculations, the mixed basis set (SDD ECP basis set45 for Pt atom and the def2-SVP basis set46 for other atoms) was used, and the optimal geometry for each compound was determined. The single point energy calculations were performed with a larger basis set the def2-TZVP basis set. The DFT-D3 dispersion correction with BJ-damping47 was applied to correct the weak interaction to improve the calculation accuracy. The SMD implicit solvation model was used to account for the solvation effect47. Finally, the single point energy of each compound was added to the free energy correction terms calculated before to obtain the Gibbs free energy. The calculated DFT results are shown in Supplementary Fig. 36, 37 & 38.