Introduction

Phosphorescent emitter materials1 provide high efficiency in phosphorescent organic light-emitting diodes (OLEDs) because of the full utilization of triplet excitons; these materials are still dominated by Ir, Pt, and Os compounds2,3,4. Circularly polarized-LED (CP-LED) devices for 3D display and human-computer interactions are promising for future technology5,6,7,8 and require chiral materials with bright circularly polarized luminescence (CPL); these materials include circularly polarized fluorescence (CPF)9, circularly polarized thermally activated delayed fluorescence (CP-TADF)10,11, and circularly polarized phosphorescence (CPP)12,13. Recently, ultrasmall chiral metal clusters have demonstrated superior CPF, CP-TADF, and CPP with high quantum yield (QY)14,15,16,17,18, and combined with their relatively low toxicity and cost, excellent photostability, and solution processability, these clusters are promising materials for use in CP-LED devices19,20. More importantly, these structurally well-defined compounds serve as a foundation for comprehending the underlying photophysical mechanisms, facilitating the rational synthesis of the clusters with enhanced and customized CPL properties. The vast array of structures and compositions in these metal clusters, with numerous metal atoms, delicate metal-metal bonds, and an array of organic ligands, present boundless opportunities for luminescent materials. However, this diversity also introduces complexities in the luminescence mechanism because of the particularly intricate energy level density and subtle influencing factors. Thus, establishing design principles for achieving high-efficiency luminescence in metal clusters, including chiral variants, has significant challenges.

Similar to molecular materials, a consensus has been reached on the importance of suppressing nonradiative decay and enhancing the intersystem crossing (ISC) rate to achieve efficient phosphorescence in metal clusters. However, while this criterion is commonly applied to explain the high quantum yield (QY) in metal clusters with multiple metal atoms and complex structures, it is less frequently used for predicting the properties of molecular materials21,22,23,24. These super-small metal clusters with a size of 1–3 nm, build an intermediate connection between the molecular and quantum dots25,26,27. Photon-excited materials, such as quantum dots, generate excitons for emission. These excitons, which are electron-hole pairs bound by the Coulombic interactions, are crucial for LED and CP-LED devices28,29,30,31. Additionally, exciton binding energy (Eb) has become an important parameter in evaluating nanoscale light-emitting materials/devices and photovoltage devices32,33,34 In quantum dots, a higher exciton binding energy suggests that excitons can remain bound for longer, thereby increasing the probability of radiative recombination. Metal clusters, which share fundamental similarities with quantum dots in their quantum confinement effects, exhibit discrete energy levels and can enhance exciton binding. Inspired by this, we hypothesized that this same principle also applies to metal clusters as well, where higher binding energy can indicate efficient radiative transitions. Consequently, this could simplify the design principles for highly emitting metal clusters and provide clearer guidance for their development. Moreover, strong electron‒phonon coupling or interactions related to molecular motions lead to fast nonradiative decay, resulting in a high nonradiative energy loss35,36. However, understanding the influence of Eb and electron‒phonon interactions on QY of metal clusters remains challenging. Exploiting the extensive structural variability and intricate emissive states of metal clusters to engineer excitons and electron‒phonon interactions could provide a design principle for enhancing the luminescence brightness of metal clusters.

Cu-doping in metal clusters has been shown to enhance luminescent properties. For instance, the phosphorescent Au16Cu6(tBuPhC ≡ C)18 cluster reported by Wang, Zhou and their coauthors exhibits notable luminescence37. Differently, Au2Cu6 clusters demonstrate weak fluorescence around 680 nm and strong phosphorescence38,39. While, chiral R/S-AuCu4 has been reported to exhibit high-efficiency thermally activated delayed fluorescence19. These studies reveal that different clusters exhibit distinct emissive mechanisms, and there is no consensus on the general enhancement of luminescence through Cu-doping. In addition, the highly emitting chiral Cu-doped metal clusters are much less underexplored, with the R/S-AuCu4 cluster being a notable exception19. The integration of high-efficiency luminescence with chirality would lead to CPL, offering potential applications in chiral optics, including circularly polarized light-emitting diodes (CP-LEDs).

In this study, we reported three enantiomeric pairs of bimetallic rhomboid clusters of Au2Cu2 with alternating metal arrangements (Au-Cu-Au-Cu), R/S-Au2Cu2(L)4 by using a thiazolidine-2-thione ligand (L) with different substituents (methyl short for mtt, isopropyl short for iptt, and phenyl short for ptt). Compared to their homogold counterparts (R/S-Au4(L)4)40, the crystalline Au2Cu2 clusters demonstrate significant increases in their photoluminescence quantum yields (PLQYs) of 36.7 → 81.2% ( ~ 2-fold greater in Au2Cu2(iptt)4), 3.4 → 82.1% ( ~ 24-fold greater in Au2Cu2(mtt)4), and 7.9 → 96.4% ( ~ 12-fold greater in Au2Cu2(ptt)4) (Fig. 1a). For instance, from Au4(ptt)4 to Au2Cu2(ptt)4, the radiative recombination rate displayed a modest 1.3-fold increase, while the nonradiative recombination rate experienced a significant 241.1-fold decrease at room temperature (RT). With these clear comparisons, we have identified the exciton binding energy (Eb) and electron‒phonon interactions as the primary contributors to the bright luminescence in the metal clusters. The exciton binding energy also increased by 1.9 times from Au4(ptt)4 to Au2Cu2(ptt)4. The electron‒phonon interactions determined by calculating the mode-resolved reorganization energies were 1.21 eV for R-Au4(ptt)4 and 0.015 eV for R-Au2Cu2(ptt)4, which supported the experimental trends of the PLQYs. Therefore, we hypothesized that substituting two copper atoms for two gold atoms within the rhomboid cluster structure would alter the exciton binding energy and influence the vibration mode of the peripheral ligands. This finding provides additional insights that could deepen our understanding of the emission mechanism of the clusters. CPL behaviors were attributed to the chirality of the cluster structures combined with the bright emission. Based on the high-efficiency CPL characteristics, cluster-based light-emitting diode (LED) devices were fabricated with a maximum EQE of 15.51% for R/S-Au2Cu2(iptt)4.

Fig. 1: Au4 and Au2Cu2 cluster structures.
figure 1

a Structures of R-Au4(L)4 and R-Au2Cu2(L)4; optical microscopy images of the crystals under UV light. Color codes: C, gray; S, yellow; N, blue; H atoms are omitted for clarity. b Metal atom distances of the optimized rhomboid cores in R-Au2Cu2(iptt)4, R-Au2Cu2(mtt)4, R-Au2Cu2(ptt)4, and R-Au4(ptt)4. For R-Au2Cu2(iptt)4, the other three Au-Cu lengths are 2.84, 2.89, and 2.90 Å. For R-Au4(ptt)4, the other three Au-Cu lengths are 3.14, 3.10, and 3.07 Å. Color codes: Au, orange; Cu, lavender.

Full size image

Results

Synthesis and Structure Analysis

The three enantiomeric Au2Cu2 clusters R/S-Au2Cu2(iptt)4, R/S-Au2Cu2(mtt)4, and R/S-Au2Cu2(ptt)4 were synthesized by the reaction of chiral ligands R/S-4-isopropylthiazolidine-2-thione (R/S-iptt) (Supplementary Figs. 1, 2), R/S-4-methylthiazolidine-2-thione (R/S-mtt) (Supplementary Figs. 3, 4), and R/S-4-phenylthiazolidine-2-thione (R/S-ptt) (Supplementary Figs. 5, 6), respectively, with Me2SAuCl and Cu(CH3CN)4(PF6) in a dichloromethane/acetonitrile (CH2Cl2/CH3CN) mixture in the presence of triethylamine at room temperature. The synthesis details are provided in the experimental section (Supplementary Fig. 7). Moreover, these Au2Cu2 clusters rapidly formed on a large scale in less than a minute ( < 1 minute) (Supplementary Fig. 8), which provides a strong foundation for their practical application.

As revealed by single-crystal X-ray diffraction (SCXRD) analysis, the three enantiomeric Au2Cu2 clusters consisted of two Au atoms, two Cu atoms, and four chiral ligands. Among the cluster molecules, the top and bottom quadrilateral faces of the Au2Cu2 kernel were capped by two chiral ligands. Both gold atoms were solely bonded to the thiol sulfur atoms, while the copper atoms were exclusively coordinated with the nitrogen atoms (Fig. 1a). R-Au2Cu2 clusters are used as representatives in the following description. Clusters R/S-Au2Cu2(iptt)4 were crystallized in the monoclinic chiral space group P21 (No. 4) (Supplementary Table 1) with Flack parameters of −0.020(3) and −0.017(5) at 200 K, indicating homochiral cluster molecule packing. Among the smallest asymmetric units of single-crystal structures, there are two types of R-Au2Cu2(iptt)4 molecules (clusters A and B) with the same composition existed with slightly different Au2Cu2 kernels (Supplementary Fig. 9). In the type A kernel, all Au and Cu atoms were linked by intracluster Au···Cu metallophilic interactions with distances in the range of 2.8382–2.8686 Å, and the angle of the two Au-Cu-Au planes was 160.779°. The type B kernel contained not only intracluster Au···Cu metallophilic interactions with distances ranging from 2.8243–2.8884 Å but also a 2.9355 Å intracluster Cu···Cu cuprophilic interaction. The angle of the two Au-Cu-Au planes was 167.988°. In addition, both type A and type B cluster molecules were integrated and assembled into one-dimensional (1D) supramolecular chains (chains A and B) by intercluster Au···Au aurophilic interactions with lengths of 3.3527 Å and 3.2280 Å, respectively (Supplementary Fig. 10). Clusters R/S-Au2Cu2(mtt)4 crystallized in an orthorhombic crystal system with the chiral space group P212121 (No. 19) (Supplementary Table 2) and Flack parameters of 0.016(11) and −0.014(2) at 200 K, respectively. The R-Au2Cu2(mtt)4 molecule also had a slightly twisted quadrilateral Au2Cu2 kernel gathered via intracluster Au···Cu and Cu···Cu interactions with distances in the range of 2.8625–3.0534 Å, and the angle of the two Au-Cu-Au planes was 170.542° (Supplementary Fig. 11). Clusters R/S-Au2Cu2(ptt)4 were crystallized in the orthorhombic chiral space group C2221 (No. 20) (Supplementary Table 3) with Flack parameters of 0.016(11) and −0.014(2) at 200 K, respectively. In the R-Au2Cu2(ptt)4 structure, cluster types A and B crystallized in unit cells arranged at an approximately vertical angle (Supplementary Fig. 12). Compared to cluster type B with a regular rhombus Au2Cu2 kernel, cluster type A had a twisted Au2Cu2 skeleton constructed by an intracluster Au···Cu metallophilic with lengths of 2.8659 and 2.8665 Å, and the torsion angle between the two Au-Cu-Au planes was 162.566°. The different and abundant interactions around the clusters were dominated by the surface ligands and potentially caused deformation in the Au2Cu2 kernels. Cu doping shortened the Au···Au, Au···Cu, and Cu···Cu metallophilic interactions in Au2Cu2 clusters compared to those in the Au4 cluster; this resulted in a significant shrinkage of the rhomboid core (Fig. 1b).

Electrospray ionization mass spectrometry (ESI-MS) in positive ion mode in DMF solution and elemental analysis were employed to confirm the chemical formula of all three Au2Cu2 clusters (Supplementary Fig. 13). The peaks at approximately m/z = 1000.8857, 917.7809, and 1103.8250 corresponded to the molecular fragments [R-Au2Cu2(iptt)4-iptt]+, [R-Au2Cu2(mtt)4-mtt]+, and [R-Au2Cu2(ptt)4-ptt]+, respectively; thus, these clusters possessed similar isolated Au2Cu2 species in solution. 1H and 13C NMR spectra of R-Au2Cu2(iptt)4, R-Au2Cu2(mtt)4, and R-Au2Cu2(ptt)4 in CDCl3 were collected to determine the formula of these metal clusters (Supplementary Figs. 14–19). Powder X-ray diffraction (PXRD) and thermogravimetric analysis (TGA) were further used to test the phase purity and thermal stability of the crystalline samples (Supplementary Figs. 20, 21). For R-Au2Cu2(iptt)4, SCXRD analysis was performed at 200 K, and one CH2Cl2 guest molecule was observed. According to TGA, no guest molecules were found; thus, the CH2Cl2 guest molecules in R-Au2Cu2(iptt)4 likely escaped from the crystal lattice, which could also be further verified by elemental analysis.

Photophysical properties

UV−vis absorption spectra of these Au2Cu2 clusters in CH2Cl2 (1 × 10−5 M) were recorded at room temperature (Supplementary Fig. 22). In solution, these isolated clusters had analogous absorption behaviors, indicating that the different substituent groups in the surface ligands had nearly no impact on the electronic transition of these isolated clusters. These solid-state Au2Cu2 clusters emitted dazzling yellow, green, and cyan photoluminescence with corresponding maximum peaks at 570, 539, and 519 nm upon UV irradiation of R-Au2Cu2(iptt)4, R-Au2Cu2(mtt)4, and R-Au2Cu2(ptt)4, respectively, under ambient conditions (Fig. 1a and Supplementary Figs. 23–25). In addition, these clusters could be excited with a broad excitation wavelength range, showing excitation wavelength-independent emission behavior (Supplementary Fig. 27). The room-temperature photoluminescence showed ultrahigh PLQYs of 81.2% for R-Au2Cu2(iptt)4 with an emission decay lifetime (τ) of 7.0 μs, 82.1% for R-Au2Cu2(mtt)4 (τ = 12.3 μs), and 96.4% for R-Au2Cu2(ptt)4 (τ = 24.5 μs) (Fig. 2a and Supplementary Figs. 23–26). The long emission lifetime indicated that the bright luminescence was related to the spin-forbidden lowest triplet excited state (T1), also known as phosphorescence (Ph)37,41,42. Moreover, as the temperature increased from 10 K to 290 K, the emission intensity exhibited a monotonic decrease without noticeable blueshifts in the emission peaks of these Au2Cu2 clusters. Consequently, the thermally activated delayed fluorescence (TADF) mechanism was prevented, and only intrinsic Ph from the T1 state was observed. The frozen dilute solution containing three pairs of Au2Cu2 clusters exhibited bright phosphorescence, with nearly identical or approaching emission peak positions observed in the crystalline samples (Supplementary Fig. 28). The redshift in crystallinity was likely related to intercluster dipole interactions. Consequently, the crystallization did not alter the inherent emission properties of R-Au2Cu2(iptt)4. In contrast, R-Au2Cu2(mtt)4 and R-Au2Cu2(ptt)4 exhibit significant differences in their emission characteristics when analyzed in both frozen solutions and crystalline solid states, which can be attributed to the distinct ligand substituents (Supplementary Fig. 29). In subsequent studies, we utilized crystalline samples to compare their phosphorescence performance with those of the Au4 clusters.

Fig. 2: Temperature-dependent photophysical properties.
figure 2

Temperature-dependent lifetimes and PLQYs (a), and krad and knon of powders R-Au4(ptt)4 (excitation wavelength: 350 nm) and R-Au2Cu2(ptt)4 (excitation wavelength: 380 nm) (b), R-Au4(mtt)4 (excitation wavelength: 350 nm) and R-Au2Cu2(mtt)4 (excitation wavelength: 380 nm) (c), R-Au4(iptt)4 (excitation wavelength: 370 nm) and R-Au2Cu2(iptt)4 (excitation wavelength: 400 nm) (d) from 10 to 290 K.

Full size image

In contrast to the crystalline Au4 clusters with identical ligands, the crystalline Au2Cu2 clusters showed a remarkable increase in the PLQY and a redshift in the emission peak intensity (Fig. 1). To elucidate the underlying reasons for the high PLQY, we initially conducted measurements on the temperature-dependent excitations, emissions, and decay lifetimes of these solid-state Au4 and Au2Cu2 clusters. The emission intensity of the Au4 and Au2Cu2 clusters exhibited a consistent decrease as the temperature increased from 10 to 290 K, with a nearly constant emission maximum. As shown in Supplementary Fig. 23, R-Au4(ptt)4 and R-Au2Cu2(ptt)4 experienced 1.4 and 1.2 times decreases in their emission intensities, respectively, over this temperature range. The emission lifetimes of R-Au4(ptt)4 and R-Au2Cu2(ptt)4 were measured to be 9.5 to 2.6 μs and 27.9 to 24.5 μs, respectively, across the temperature range of 10 to 290 K (Fig. 2a and Supplementary Figs. 23 and 26). As shown in Supplementary Fig. 24, a decrease in the emission intensities of 37.7 and 2.0 times for R-Au4(mtt)4 and R-Au2Cu2(mtt)4, respectively, was observed as the temperature decreased from 10 to 290 K. This reduction was accompanied by a decrease in the emission lifetime from 9.8 to 1.2 μs and 18.8 to 12.3 μs (Supplementary Figs. 24 and 26). Similar behaviors were also observed in Supplementary Fig. 25 for R-Au4(iptt)4 and R-Au2Cu2(iptt)4, where the emission intensity decreased by 2.6 and 1.6 times, respectively. Additionally, the emission lifetimes decreased from 4.2 to 3.1 μs and 8.8 to 7.0 μs within the same temperature range (Supplementary Figs. 25 and 26).

According to the Eq. (1), the ratio of the radiative rate (krad) to the sum of the radiative and nonradiative rates (knon) determines the PLQY32:

$${{\rm{PLQY}}}=\frac{{k}_{{{\rm{rad}}}}}{{k}_{{{\rm{rad}}}}+{k}_{{{\rm{non}}}}}$$
(1)

The temperature-dependent PLQYs of these Au4 and Au2Cu2 clusters were estimated from the ratio of the integration of the peak areas at different temperatures. In the case of R-Au2Cu2(ptt)4, the PLQYs remained consistently high, approaching unity, when excited at 380 nm across the temperature range of 10 to 290 K (Fig. 2a). This was notably higher than the PLQY of R-Au4(ptt)4, which was 9.9% at 10 K and 7.9% at 290 K. As shown in Supplementary Fig. 26, upon reducing the temperature from 290 to 10 K, both R-Au2Cu2(mtt)4 and R-Au2Cu2(iptt)4 showed only slight changes in the PLQYs from 82.1% and 81.2% to near unity, respectively. However, significant enhancements in the PLQYs were observed for R-Au4(mtt)4, increasing from 3.4% to 64.8%, and for R-Au4(iptt)4, increasing from 36.7% to 85.9%, over the same temperature range. Moreover, krad and knon in the temperature range of 10 to 290 K were also examined. As depicted in Fig. 2b and Table 1, the krad of R-Au2Cu2(ptt)4 calculated as 0.35 × 105 s−1 at 10 K and 0.39 × 105 s−1 at 290 K was greater than those of 0.10 × 105 s−1 at 10 K and 0.30 × 105 s−1 at 290 K for R-Au4(ptt)4. Notably, the knon of R-Au4(ptt)4 was two orders of magnitude greater than that of R-Au2Cu2(ptt)4, 131.4 times greater at 10 K and 241.1 times greater at 290 K. From R-Au4(mtt)4 to R-Au2Cu2(mtt)4, although the krad was almost consistent at low temperatures and exhibited a 2.3-fold increase at 290 K, the knon showed an approximately 55.3-fold decrease at 290 K (Fig. 2c). When the ligands were changed to iptt, minimal changes in the krad values were observed over the temperature range of 10 to 290 K from R-Au4(iptt)4 to R-Au2Cu2(iptt)4, and an approximately 10-fold decrease in knon was found at 290 K (Fig. 2d and Table 1). Overall, in the case of the Au2Cu2 clusters, the krad values consistently exceeded the knon values across all temperatures. On the other hand, the krad values of R-Au4(ptt)4 remained significantly lower than the knon values from 290 to 10 K. Additionally, the knon values of R-Au4(mtt)4 and R-Au4(iptt)4 far surpassed the krad values as temperatures decreased below 220 K and 190 K, respectively (Fig. 2c and d). The consistently low knon values of the Au2Cu2 clusters contributed to their high PLQYs, both at low and room temperatures.

Table 1 The measured PLQY, lifetime, krad, knon, calculated exciton binding energy, exchange interaction, and electron‒phonon interaction of R-Au2Cu2(iptt)4, R-Au2Cu2(mtt)4, R-Au2Cu2(ptt)4, and R-Au4(ptt)4
Full size table

Density functional theory calculations

To gain more insight into the origin of high PLQY of the Au2Cu2 clusters, we carried out density functional theory (DFT) and time-dependent DFT (TD-DFT) calculations. As shown in Supplementary Fig. 30 and Supplementary Table 4, all Au4 and Au2Cu2 clusters possessed similar highest-occupied molecular orbital (HOMO) and lowest-unoccupied molecular orbital (LUMO) characteristics. The HOMO was mainly localized on the metal atoms with a small distribution over the ligands (especially on the S and N atoms). The LUMO was contributed by both metal atoms and ligands, and the electron density in the LUMO was dominantly localized in the central region of the metal skeleton in both the Au2Cu2 and Au4 clusters, which was likely induced by the intracluster metallophilic interactions. As shown in Supplementary Table 4, the HOMO of R-Au4(ptt)4 consisted of 70.02% Au atoms and 29.98% ligand, and the LUMO consisted of 25.44% Au atoms and 74.56% ligand. By Cu doping, the HOMO of R-Au2Cu2(ptt)4 included 24.16% Au atoms, 51.25% Cu atoms and 24.59% ligand contributions, and the LUMO consisted of 14.08% Au atoms, 18.05% Cu atoms, and 67.87% ligand contributions (Supplementary Table 4). Similar HOMO and LUMO compositions were also observed for R-Au2Cu2(iptt)4 and R-Au2Cu2(mtt)4. In these cases, the substituent groups (mtt, iptt, ptt) of the ligand contributed minimally to the frontier molecular orbitals, thereby not altering the inherent transitions of the cluster molecules. The calculated UV-vis absorption spectra of these clusters matched well with the experimental ones in the CH2Cl2 solution (1 × 10−5 M) (Supplementary Fig. 31), indicating the rationality of the orbital calculations at the molecular level.

The exciton binding energy (Eb), which is defined as an electron and a hole attracted to each other by the Coulomb force, was calculated by two kinds of models in our systems. Model 1 is derived from the energy difference between the fundamental gap (Efund) and the optical gap (Eopt). While Efund is defined as the energy difference between the ionization potential (IP) and the electron affinity (EA), Eopt represents the energy required to transition from the ground state (S0) to the lowest singlet excited state (S1). This optical gap is typically determined experimentally through measurements of the absorption band edge. However, given the greater degree of degeneracy in the molecular orbitals of metal clusters, the Eb calculated based on single molecular orbital pair excitation is less accurate; thus, Model 2, which is based on the electrostatic interaction between the densities of electrons (ρe) and holes (ρh), is more suitable for these calculations. The ρe and ρh densities can be constructed as described below43,44:

$${E}_{b}=\iint \frac{{\rho }^{h}({r}_{1}){\rho }^{e}({r}_{2})}{\left|{r}_{1}-{r}_{2}\right|}d{r}_{1}d{r}_{2}$$
(2)

where the density distribution of holes and electrons can be defined as follows:

$${\rho }^{h}(r)= {\rho }_{\left({loc}\right)}^{h}(r)+{\rho }_{\left({cross}\right)}^{h}(r)={\sum}_{i\to a}{({w}_{i}^{a})}^{2}{\varphi }_{i}(r){\varphi }_{i}(r) \\ -{\sum}_{i\to a}{\sum}_{j\ne i\to a}{w}_{i}^{a}{w}_{j}^{a}{\varphi }_{i}(r){\varphi }_{j}(r)$$
(3)
$${\rho }^{e}(r)= {\rho }_{\left({loc}\right)}^{e}(r)+{\rho }_{\left({cross}\right)}^{e}(r)={\sum}_{i\to a}{({w}_{i}^{a})}^{2}{\varphi }_{a}(r){\varphi }_{a}(r) \\ -{\sum}_{i\to a}{\sum}_{i\to b\ne a}{w}_{i}^{a}{w}_{j}^{b}{\varphi }_{a}(r){\varphi }_{b}(r)$$
(4)

where φ denotes the molecular orbital wavefunction, wi denotes the configuration coefficient of orbital i, and “loc” and “cross” represent the contributions of the local term and cross term to the hole and electron distributions, respectively. More details about the definition of Model 2 can be found in ref. 45. Additionally, the long-range corrected functional wB97XD, combined with the more extensive basis set def2-TZVP, was employed to ensure the proper description of electron excitation with separated holes and electrons.

As shown in Table 1 and Supplementary Table 5, the Au4 cluster consistently exhibits a smaller Eb than the Au2Cu2 clusters across both the theoretical levels and the two models, especially for the triplet excitons. Moreover, only R-Au4(ptt)4 has a significantly larger singlet exciton binding energy than the triplet (Table 1); thus, a singlet exciton can be more easily dissociated into free charges31. Therefore, possible de-excitation occurs in the Fl channel rather than in the Ph channel. A higher exciton binding energy tends to result in more stable excitons, which can lead to longer exciton lifetimes and potentially higher PLQYs. However, nonradiative decay pathways can also significantly influence the PLQY. Since the heavy atoms of Au and Cu are involved in the metal clusters, the spin-orbit coupling and exchange interactions that determine the ordering of the electronic spin states of materials are discussed.

The exchange interaction, which influences the arrangement of the electronic spin states in materials, is a crucial factor in describing exciton behavior35. When the exciton becomes localized, the exchange interaction increases, leading to a greater splitting between the singlet and triplet states46. Based on the calculated singlet and triplet excited energies (see Supplementary Table 6), the exchange interactions of the Au2Cu2 clusters and R-Au4(ptt)4 can be determined by considering the singlet-triplet splitting (singlet-triplet splitting = 2 × exchange interaction). Table 1 and Supplementary Table 7 demonstrate that doping with Cu atoms results in stronger exchange interactions in Au2Cu2 clusters than in R-Au4(ptt)4. The hole and electron distributions of the S1 and T1 excited states in the Au2Cu2 clusters and R-Au4(ptt)4 are presented in Supplementary Fig. 32. For the Au2Cu2 clusters, the hole and electron densities in both the S1 and T1 excited states are consistent, with minor differences (Supplementary Fig. 32). These results indicate that the distortion between S1 and T1 is minimal across different spin states. The holes in both states are predominantly located on the metal, with a smaller number located on the coordinated S and N atoms. The electrons tend to be transferred to the central position of the Au2Cu2 core, with some also residing on the metal atoms, coordinated S and N atoms, and the connecting C atoms. In contrast, for R-Au4(ptt)4, the hole densities of S1 are similar to those of Au2Cu2, but the electrons are primarily transferred to the coordinated S and N atoms as well as the connecting C atoms. However, the hole densities of T1 are mostly distributed on the metal atoms and coordinated S and N atoms, with the electron densities distributed on the coordinated S and N atoms, and nearby C atoms; this arrangement is significantly different from those of Au2Cu2. These findings indicate that the higher exchange interactions in Au2Cu2 clusters enhance exciton localization, potentially due to the presence of Cu-Au bonds.

Generally, a robust spin-orbit coupling matrix element (SOCME) enhances the Ph (phosphorescence) output, and the heavier material Au would exhibit larger SOC than Cu and therefore potentially promote the higher PLQY30,34,35. However, the PLQY of the Au4 cluster decreased significantly relative to that of the Au2Cu2 cluster, although the four metal clusters all had large SOCME ( > 300 cm−1) between S1 and the low-lying triplet state Tn (n ≤ 3) (Supplementary Fig. 33, Supplementary Tables 8 and 9). Considering the strong electron‒phonon coupling induced by molecular motion significantly contributed to the intrinsic electron-transfer barrier and suppressed the ISC process, thereby inhibiting radiative decay47. Here, we evaluated electron‒phonon interactions by calculating reorganization energy components in these four metal clusters. According to the mode-specific model48,49, the electron-phonon coupling energies calculated from the mode-resolved reorganization energies were 0.018 eV for R-Au2Cu2(iptt)4, 0.012 eV for R-Au2Cu2(mtt)4, 0.015 eV for R-Au2Cu2(ptt)4, and 1.21 eV for R-Au4(ptt)4 (Table 1 and Fig. 3). In the Au2Cu2 clusters, the three highest vibrations were all in the low-frequency range of less than 220 cm−1. For R-Au4(ptt)4, the three highest vibrations were located in the high-frequency region at 683, 720, and 1496 cm−1. Moreover, the vibrational frequency intensity of the Au4 cluster in the excited state was two orders of magnitude greater than that of the Au2Cu2 clusters. The large displacement of the vibration between S1 and T1 would lead to a clear structural distortion. Clearly, the root mean square deviation (RMSD) calculated from S0/S1, S0/T1, and S1/T1 (Supplementary Figs. 34, 35) indicated significant differences compared to R-Au4(ptt)4, with values nearly ten times larger than those of Au-Cu binaries.

Fig. 3: Reorganization energy contribution.
figure 3

Reorganization energy components as a function of frequency at the PBE0/def2-SVP level in (a) R-Au2Cu2(iptt)4, (b) R-Au2Cu2(mtt)4, (c) R-Au2Cu2(ptt)4, and (d) R-Au4(ptt)4, the top three normal modes that contribute to the reorganization energy for the four metal clusters are labeled in the red box.

Full size image

Interestingly, the vibrations of the organic ligands are metal-skeleton-dependent or metal-dependent, which has not been reported in the literature (Supplementary Fig. 36). Therefore, we attributed the predominant nonradiative decay, which largely influenced the PLQY, to metal-skeleton-dependent vibrations of functional groups in the organic ligands. In summary, for R-Au4(ptt)4, the low PLQY was closely related to the strong electron‒phonon coupling, while for the Au2Cu2 clusters, the introduction of Cu atoms led to excitons that were more strongly bound, thereby enhancing the krad and suppressing the knon. These findings indicated that Cu doping significantly stabilized triplet excitons and reduced electron‒phonon coupling.

Transient absorption spectroscopy

Transient absorption (TA) measurements were further performed to gain deep insights into the differences in electron dynamics in their excited states. We selected R-Au4(ptt)4 and R-Au2Cu2(ptt)4 for comparative illustration in the film state considering the solution state is less emissive, and the crystalline samples are not adaptable for TA measurements. The probing region was selected as 400–600 nm. As shown in Fig. 4a–d, the global fitting results produced three dynamic signals of τ1 = 4.2 ps, τ2 = 28.7 ns and τ3 = 2.1 μs for R-Au4(ptt)4. In contrast, for R-Au2Cu2(ptt)4, only two dynamic processes with τ1 = 5.7 ps and τ2 = 16.9 μs were observed. TCSPC measurements of the film-state revealed decay lifetimes of 5.4 μs for R-Au4(ptt)4 and 16.9 μs for R-Au2Cu2(ptt)4 at room temperature (Supplementary Fig. 37). Therefore, the microsecond relaxation processes of 2.1 μs and 16.9 μs in the TA measurements for R-Au4(ptt)4 and R-Au2Cu2(ptt)4 could be considered as the Ph relaxations from T1. For R-Au4(ptt)4, the shortest τ1 = 4.2 ps could be attributed to internal conversion (IC) from Sn (high-lying singlet excited state, n > 1) to S1; τ2 = 28.7 ns was likely responsible for ISC processes from S1 to Tn (n = 1–10). While for R-Au2Cu2(ptt)4, a longer lifetime of nanosecond scale disappeared, only a 5.7 ps followed by the Ph emission. Considering the above-discussed higher Eb and lower electron‒phonon interactions of this bimetallic cluster, the 5.7 ps process was combined by IC and ISC processes from S1 to T2 (n = 1–2) (Supplementary Fig. 33). In addition, the extremely low reorganization energy of 0.015 eV also supported a rapid intersystem crossing (ISC) process in R-Au2Cu2(ptt)4. In contrast, the larger value of 1.2 eV in R-Au4(ptt)4 led to a delayed ISC process with a lifetime of approximately 29 ns.

Fig. 4: Transient absorption.
figure 4

a–d Transient absorption data and corresponding global analysis models of R-Au4(ptt)4 and R-Au2Cu2(ptt)4.

Full size image

Chiroptical properties

The absolute configurations of these enantiomer clusters have been determined by SXRD analysis. To gain further insight into the relationship between chiroptical behavior and structure, CD spectra were measured. All enantiomers of R/S-Au2Cu2(iptt)4, R/S-Au2Cu2(mtt)4, and R/S-Au2Cu2(ptt)4 showed perfect mirror CD signals with similar spectral characteristics in both dilute solution and the solid-state, apart from a small shift (Fig. 5a and Supplementary Fig. 38). As shown in Supplementary Fig. 39, the isolated clusters exhibited completely different CD signals from those of the corresponding chiral ligands in the CH2Cl2 solution. R/S-iptt and R/S-Au2Cu2(iptt)4 were selected as representatives for the below-detailed analysis. For S-Au2Cu2(iptt)4, five CD peaks at 365 nm (-), 297 nm (+), 268 nm (-), 243 nm (+), and 229 nm (-) were observed; however, S-iptt had five CD signal peaks at 345 nm (+), 280 nm (-), 265 nm (+), 254 nm (-), and 232 nm (+). We postulated that the chiroptical properties of these isolated clusters were related to both the surface ligands and the Au2Cu2 kernel, whose chiroptical activity was endowed by chiral ligands through the Au-S/Cu-N interface. Moreover, R/S-mtt/R/S-Au2Cu2(mtt)4 and R/S-ptt/R/S-Au2Cu2(ptt)4 displayed similar chiroptical behaviors. Contrasting the chiroptical properties of each enantiomer in dilute solution and the solid-state, no dramatic difference was observed for R/S-Au2Cu2(iptt)4, but significantly different CD bands were observed for R/S-Au2Cu2(mtt)4 and R/S-Au2Cu2(ptt)4 (Supplementary Fig. 39). These results indicated that the chiroptical activity in the ground state of solid-state R/S-Au2Cu2(iptt)4 mainly originated from the isolated cluster and was independent of the intercluster interactions. In contrast, the intercluster interactions in the crystal states of R/S-Au2Cu2(mtt)4 and R/S-Au2Cu2(ptt)4 played important roles in transmitting chiroptical activity from isolated clusters to macroscopic structures. CPL spectra of R/S-Au2Cu2(iptt)4, R/S-Au2Cu2(mtt)4, and R/S-Au2Cu2(ptt)4 in powders were recorded. As shown in Fig. 5b, the powder samples of R/S-Au2Cu2(iptt)4, R/S-Au2Cu2(mtt)4, and R/S-Au2Cu2(ptt)4 had mirror image CPL signals centered at 578, 542, and 520 nm with luminescence anisotropy factors (glum = 2(IL − IR)/(IL + IR)) of −5.4/5.1 × 10−3, −3.9/3.4 × 10−3, and −7.5/7.0 × 10−3 (Supplementary Fig. 40), respectively; these values were comparable to those of reported chiral nanoclusters. These results also indicated that the excited states of these clusters were chiral.

Fig. 5: Chiroptical properties and CP-LED.
figure 5

a–b Solid-state CD spectra and CPL with DC spectra of R/S-Au2Cu2(iptt)4, R/S-Au2Cu2(mtt)4, and R/S-Au2Cu2(ptt)4 (solid line: S- cluster; dashed line: R- cluster). c Energy level diagram of the LED devices. d EL and CPEL spectra of the LED devices. Inset: The turn-on device photos at 8 V (insets). e External quantum efficiency-luminescence relationships of the LEDs.

Full size image

Circularly polarized cluster light-emitting diodes

With an ultrahigh PLQY and solution-processable property in these pairs of enantiopure Au2Cu2 clusters, we fabricated a prototype light-emitting diode (LED) device through spin coating, with an optimized device architecture of indium tin oxide (ITO)/poly(3,4-ethylenedioxythiophene): poly(styrenesulfonate) (PEDOT:PSS; 40 nm)/1,3-di-9-carbazolylbenzene (mCP): cluster (10%wt., 40 nm)|bis[2-(diphenylphosphino)phenyl]ether (DPEPO, 10 nm)|1,3,5-tri[(3-pyridyl)-phen-3-yl]benzene (TmPyPB, 50 nm)|LiF (1 nm)|Al (100 nm) (Fig. 5c). PEDOT:PSS and LiF served as hole and electron injecting layers, respectively, mCP was the host dopant in the emissive layer, DPEPO was used as an exciton blocking layer, and TmPyPB was employed as an electron transporting layer (Supplementary Fig. 41). The R-Au2Cu2(iptt)4, R-Au2Cu2(mtt)4, and R-Au2Cu2(ptt)4 LEDs emitted green electroluminescence (EL) with emission peaks at 543, 541, and 538 nm and turn-on voltages of 5, 6, and 6 V, respectively (Fig. 5d and Supplementary Table 10). Compared to R-Au2Cu2(mtt)4 and R-Au2Cu2(ptt)4 LEDs, the R-Au2Cu2(iptt)4 LED showed better EL performance, with maximum efficiencies of 59.60 cd A−1 for the current efficiency (CE, ηCE), 27.12 lm W−1 for the power efficiency (PE, ηPE), and 15.51% for the EL external quantum efficiency (EQE, ηEQE) (Fig. 5e, Supplementary Figs. 42–43, and Supplementary Table 10). In addition, these CP-LEDs also exhibited circularly polarized electroluminescence (CPEL) behaviors with dissymmetry factor (gEL) of −7.6/ + 2.6 × 10−3 for R/S-Au2Cu2(iptt)4, −1.4/ + 0.7 × 10−3 for R/S-Au2Cu2(mtt)4, and −1.7/ + 1.1 × 10−3 for R/S-Au2Cu2(ptt)4 (Fig. 5d and Supplementary Fig. 44). The outstanding CPL properties demonstrate promising potential for the application of these metal clusters in the field of circularly polarized light-emitting diodes (CP-LEDs) (Supplementary Table 11).

Discussion

In summary, we developed three enantiomeric pairs of rhomboid gold-copper clusters by incorporating two copper atoms in the mother Au4 cluster and three different substituents (methyl, isopropyl, and phenyl) in the thiazolidine-2-thione-derived ligands. The introduction of the copper atoms facilitated the localization of the excitons in the triplet state, leading to significant increases in the exciton binding energies; consequently, a notable maximum 1.3-fold increase in the krad was observed. Furthermore, electron‒phonon interactions and reorganization energy were dramatically reduced, resulting in a maximum 241.1-fold decrease in the knon. These combined effects culminated in achieving a phosphorescence quantum yield of over 96% at room temperature and excellent CPL. A solution-processed CP-LED prototype fabricated using Au2Cu2 clusters demonstrated a maximum external quantum efficiency of 15.51% in green and a maximum dissymmetry factor |gEL| of 7.6 × 10−3. Our study provides valuable insights into the crucial roles of the exciton-binding energy and electron‒phonon interactions in shaping the emission efficiency of luminescent metal clusters. Our results introduce an approach to designing bright chiral metal clusters for optoelectronic applications, particularly in the realm of circularly polarized optoelectronics.

Methods

Synthesis of R/S-Au2Cu2(iptt)4

Me2SAuCl (7.38 mg, 0.025 mmol) and R/S-iptt (8.05 mg, 0.05 mmol) were dissolved in 3 mL of CH2Cl2 and 1 mL of CH3CN. After stirring for 5 minutes, Cu(CH3CN)4(PF6) (9.32 mg, 0.025 mmol) and triethylamine (10 μL) were added. After stirring for 10 minutes, the solution was filtered and then kept in the dark for one day to afford block crystals suitable for X-ray diffraction analysis. Yields: 66.5% for R-Au2Cu2(iptt)4 based on Au. 1H NMR (600 MHz, CDCl3): δ 5.15-5.06 (m, 4H), 3.63-3.52 (m, 4H), 3.24–3.18 (m, 4H), 2.53–2.40 (m, 4H), 0.95–0.90 (m, 12H), 0.91-0.87 (m, 12H); 13C NMR (151 MHz, CDCl3) δ 183.0, 80.3, 36.2, 32.9, 19.7, 16.1. Analysis (calcd., found for C24H40Au2Cu2N4S8): C (24.80, 24.82), N (4.82, 4.45), S (22.07, 22.13), H (3.47, 3.42).

Synthesis of R/S-Au2Cu2(mtt)4

Me2SAuCl (7.38 mg, 0.025 mmol) and R/S-mtt (6.65 mg, 0.05 mmol) were dissolved in 3 mL of CH2Cl2 and 1 mL of CH3CN. After stirring for 5 minutes, Cu(CH3CN)4(PF6) (9.32 mg, 0.025 mmol) and triethylamine (10 μL) were added. After stirring for 10 minutes, the solution was filtered and then kept in the dark for one day to afford block crystals suitable for X-ray diffraction analysis. Yields: 72.7% for R-Au2Cu2(mtt)4 based on Au. 1H NMR (600 MHz, CDCl3): δ 5.08-5.00 (m, 4H), 3.89–3.82 (m, 4H), 3.12-3.07 (m, 4H), 1.35-1.31 (m, 12H); 13C NMR (151 MHz, CDCl3) δ 182.5, 70.5, 44.0, 21.0. Analysis (calcd., found for C16H24Au2Cu2N4S8): C (18.30, 18.37), N (5.34, 4.88), S (24.43, 24.89), H (2.30, 2.26).

Synthesis of R/S-Au2Cu2(ptt)4

Me2SAuCl (7.38 mg, 0.025 mmol) and R/S-ptt (9.75 mg, 0.05 mmol) were dissolved in 3 mL CH2Cl2 and 1 mL CH3CN. After stirring for 5 minutes, Cu(CH3CN)4(PF6) (9.32 mg, 0.025 mmol) and triethylamine (10 μL) were added, respectively. After stirring for 10 minutes, the solution was filtered and then kept in the dark for one day to afford block crystals suitable for X-ray diffraction analysis. Yields: 76.4% for R-Au2Cu2(ptt)4 based on Au. 1H NMR (600 MHz, CDCl3): δ 7.34–6.92 (m, 20H), 6.22-6.04 (m, 4H), 4.13-3.87 (m, 4H), 3.55-3.26 (m, 4H); 13C NMR (151 MHz, CDCl3) δ 184.0, 140.1, 129.0, 128.6, 127.5, 78.3, 45.2. Analysis (calcd., found for C36H32Au2Cu2N4S8): C (33.31, 33.24), N (4.32, 4.05), S (19.76, 19.46), H (2.48, 2.44).

Synthesis of R-Au4(mtt)4 and R-Au4(ptt)4

R-Au4(mtt)4 and R-Au4(ptt)4 crystals were collected according to previous methods37. Me2SAuCl (14.76 mg, 0.05 mmol) and R-mtt (13.3 mg, 0.1 mmol) or R-ptt (19.5 mg, 0.1 mmol) were dissolved in 3 mL CH2Cl2 and 1 mL CH3CN. After stirring for 5 minutes, triethylamine (10 μL) was added. After stirring for 10 minutes, the solution was filtered and then kept in the dark for one day to afford block crystals suitable for X-ray diffraction analysis. Yields: 58.2% for R-Au4(mtt)4 based on Au; 51.3% for R-Au4(ptt)4 based on Au.

Preparation of PMMA films containing R-Au4(ptt)4 and R-Au2Cu2(ptt)4

Polymethyl Methacrylate (PMMA) (100 mg) and Au4 clusters (5 mg) were each dissolved in 1 mL of CH2Cl2. Subsequently, these solutions were combined and dripped onto a clean glass plate. After allowing the mixture to dry slowly at room temperature, PMMA films containing Au4 clusters were obtained.

Crystallographic data collection and refinement of the structure

SCXRD measurements of clusters S/R-Au2Cu2(iptt)4 and S/R-Au2Cu2(ptt)4 were performed at 200 K on a Rigaku XtaLAB Pro diffractometer with Cu-Kα radiation (λ = 1.54184 Å), and S/R-Au2Cu2(mtt)4 and R-Au4(mtt)4 were performed at 200 K on a Rigaku XtaLAB Pro diffractometer with Mo-Kα radiation (λ = 0.71073 Å). Data collection and reduction were performed using the program CrysAlisPro. All structures were solved with direct methods (SHELXS) and refined by full-matrix least-squares on F2 using OLEX2, which utilizes the SHELXL-2015 module. All atoms were anisotropically refined, and hydrogen atoms were placed in calculated positions with idealized geometries and assigned fixed isotropic displacement parameters. Detailed information regarding the X-ray crystal data, intensity collection procedure, and refinement results for all cluster compounds is summarized in Supplementary Tables 1–3.

Photoluminescence measurements at room temperature

UV-vis absorption spectra of the solution were recorded using a TU-1901 spectrophotometer in the range of 200–800 nm, and UV-Vis absorption spectra of the solid were recorded using a Hitachi UH4150 in the range of 200–800 nm. Emission and excitation spectra at room temperature were recorded with a HORIBA Scientific Fluorolog-3 spectrofluorometer. Luminescence decay was measured on a HORIBA Scientific Fluorolog-3 fluorescence spectrometer in time-correlated single-photon counting (TCSPC) mode. The photoluminescence quantum efficiency was measured using an integrating sphere on a HORIBA Scientific Fluorolog-3 spectrofluorometer. The crystal luminescence microscopy images were recorded on an Olympus BX53 microscope.

Photoluminescence measurements at various temperatures

The temperature-dependent excitation and emission spectra of the solid-state samples were recorded by using a HORIBA Scientific Fluorolog-3 spectrofluorometer with a PHYSIKE ACRYO-4K-HT-101-O closed cycle cryostat system to control the sample temperature. The temperature-dependent lifetimes of the solid-state samples were also measured with a HORIBA Scientific Fluorolog-3 spectrofluorometer in TCSPC mode with a PHYSIKE ACRYO-4K-HT-101-O closed cycle cryostat system that controlled the sample temperature.

Circularly polarized luminescence measurements

CPL and electroluminescence (EL) spectra were measured on a JASCO CPL-300 spectrometer.

Circular dichroism measurement

CD spectra were recorded on a JASCO J-1500 spectrometer.

Transient absorption measurement

The TA spectra were recorded on a commercial pump-probe system (Helios-EOS, Ultrafast Systems LLC) in combination with a femtosecond laser system (Astrella, Coherent). Laser pulses ( ~ 800 nm center wavelength, <100 fs duration, 1 kHz repetition rate, ~7 mJ/pulse) were generated by a Ti:sapphire-based regenerative amplified laser system. The laser pulses were split to generate pump and probe beams. The pump pulses at 320 nm were delivered by an optical parametric amplifier that was excited by a portion of the 800-nm laser pulses. The pump pulse energy in each measurement was ~25 μJ/cm2 at the sample cell. For femtosecond TA (100 fs-7 ns delay region) measurements, the probe and reference beams were generated by focusing the 800-nm beam (split from the amplifier, with a tiny portion) onto a CaF2 crystal to generate white-light continuum pulses (350 − 650 nm). The pump−probe delay was controlled by an optical delay line. For the nanosecond TA (1 ns−400 μs delay region) measurement, the pump beam was generated in the same way as in the femtosecond TA experiment described above. The probe and reference beams were provided by an additional supercontinuum laser (370 − 900 nm, ~0.5 ns duration, 2 kHz repetition rate). The pump−probe delay time was electronically controlled in the nanosecond TA measurements. In femtosecond and nanosecond TA spectroscopy, the intensities of the probe and reference beams were detected by a pair of linear array detectors to obtain the TA signal. The global and target analyzes were performed with Glotaran software.

Quantum chemical calculations

All metal clusters were first optimized with density functional theory (DFT) of the PBE0/def2-SVP level with the Gaussian 16 program50, and then the absorption spectra of R-Au2Cu2(iptt)4, R-Au2Cu2(mtt)4, R-Au2Cu2(ptt)4, R-Au4(iptt)4, R-Au4(mtt)4, and R-Au4(ptt)4 were simulated by time-dependent density functional theory (TD-DFT) at the same theoretical level as the ground state. The spin-orbit coupling matrix elements between the singlet and triplet excited states were calculated by the ORCA 5.0.3 program51. Additionally, the long-range corrected functional wB97XD, combined with the more extensive basis set def2-TZVP, was employed to ensure the inclusion of long-range interactions for the Eb calculations. The Eb calculations and molecular orbitals were generated and visualized by the Multiwfn and VMD programs52,53.

Fabrication of CP-OLED devices and testing

The device configuration was ITO | PEDOT:PSS (40 nm)|mCP:cluster (10%wt., 40 nm)|DPEPO (10 nm)|TmPyPB (50 nm)|LiF (1 nm)|Al (100 nm). Poly(3,4-ethylenedioxythiophene): poly(styrenesulfonate) (PEDOT:PSS) and LiF were the hole and electron injecting layers, respectively. MCP, DPEPO, and TmPyPB served as the host, exciton blocking, and electron transporting layers, respectively.

ITO substrates were cleaned with detergents, deionized water, acetone, and ethanol, dried in an oven at 120 °C for 3 h, treated with oxygen plasma for 15 minutes, and then transferred to a super-clean bench. A PEDOT:PSS layer was spin-coated on the ITO-coated glass substrate, and then baked at 120oC for 15 minutes. The emitting layer was spin-coated from chlorobenzene at a concentration of 10 mg/mL, and the ITO-coated glass substrates were then heated at 50oC for 15 minutes. After spin-coating, the glass substrate was transferred to a high vacuum evaporation system, and the DPEPO, TmPyPB, LiF, and Al layers were deposited under a base pressure of ~ 10−4 Pa. The external quantum efficiency of the devices was calculated with the assumption of a Lambertian distribution. The electroluminescent spectra were collected via a calibrated PR655 spectrometer.