Group 14 Zintl ions have been investigated extensively in solution chemistry and nanomaterials during the last 20 years [1-5]. Homoatomic deltahedral clusters of E₉⁴⁻ (E = Ge, Sn, Pb) are considered to be important precursors to react with the transition-metal reagents in an aprotic solvent, and previous studies have been mainly focused on the reactions of empty nine-atom cages with low-valent transition-metal compounds. One of the efficient strategies is that [E₉]⁴⁻ anions are used as six-electron donors to replace the ligand of organometallic complexes leading to the formation of ten-vertex closo-clusters, in which the square plane of [E₉]⁴⁻ is capped by organometallic fragments, for instance, [Sn₉M(CO)₃]⁴⁻ (M = Cr, Mo, W) [6], [Ge₉Ni(CO)]³⁻ [7], [Ge₉Pd(PPh₃)]³⁻ [8], [Sn₉Ir(COD)]³⁻ [9], [Ge₉Cu(PR₃)]³⁻ (R = ⁱPr, Cy) [10] and [E₉ZnR]³⁻ (R = Mes, ⁱPr, Ph) [11]. Moreover, as analogs of fullerene, nonatetrelide is capable of accommodating one transition metal atom by methods of solution reaction or high-temperature solid-state synthesis [12-16]. Nevertheless, due to the unique reactivity of endohedral clusters, only several [M@E₉]^q− can be further reacted with suitable organometallic compounds to form "Rudolph's complexes", namely [M@E₉ML]ⁿ⁻ (M = transition metal, L = ligand) [15, 17, 18].

A series of polymers, [Ge₉-Ge₉]⁶⁻ [19], [Ge₉=Ge₉=Ge₉]⁶⁻ [20] and nanorod [Ge₉=Ge₉=Ge₉=Ge₉]⁸⁻ [21], were obtained by oxidative coupling reactions of monomer Ge₉⁴⁻. And a family of clusters containing Ge₉⁴⁻ and bridged metal atom has also been isolated, such as [Ge₉InGe₉]⁵⁻ [22], one-dimension chain _∞¹[MGe₉]²⁻ (M = Zn, Hg) [23, 24]. As for tin clusters, only three related examples, [Sn₉HgSn₉]⁶⁻ [25], [Sn₉AgSn₉]⁷⁻ [26] and [Ag(Sn₉-Sn₉)]⁵⁻ [27] were discovered. Recently our group reported the first example of an endohedral bridged Zintl cluster [(Ni@Sn₉)In(Ni@Sn₉)]⁵⁻ [28], which contains a bridged main group metal atom. Given the less explored chemistry of [M@Sn₉]^q− and their similarity with the empty counterpart when they react with transition metals, based on the previous work, further research was carried out. Here, we present the synthesis and characterization of two new Zintl clusters [(Sn₉)Cd(Sn₉)]⁶⁻ (1a) and [(Ni@Sn₉)Cd(Ni@Sn₉)]⁶⁻ (2a) (Fig. 1), which proves that both empty and centered nine-atom tin clusters behave similarly to CdMes₂.

To investigate the different reactivities of the [Sn₉]⁴⁻ and [Ni@Sn₉]⁴⁻, we have tried to adjust the reaction conditions to isolate the corresponding complex. Since the synthesis process of 1 and 2 are in the same way, only the synthesis of 1 is taken as an example for description. When the required amount of CdMes₂ was added directly to a homogeneous solution of K₄Sn₉ with the chelating agent 2.2.2-crypt, the color of the solution rapidly changed from reddish-brown to dark red. After stirring for 3 h, the solution turned reddish-brown again. Carefully layered with toluene, crystals of compound 1 were obtained after two weeks. The electrospray ionization (ESI) mass spectra of the crystal sample of compound 2 shows the signals of [KNiSn₉]⁻ (m/z = 1166.0240), [NiSn₉]⁻ and [NiSn₁₀]⁻ without the corresponding peaks of parent anion, similarly only the signal of [Sn₉]⁻ is detected in the mass spectra of compound 1. Such phenomenon, which was also observed in [Ge₉SnGe₉]⁴⁻ [29], suggests that the Cd-Sn bonds are relatively weak in contrast to the previously reported In-Sn bond [28]. From the perspective of experimental results, the reactivity of [Sn₉]⁴⁻ and [Ni@Sn₉]⁴⁻ with CdMes₂ did not differ too much. Hence, given the above studies and the former conclusions, the formation of 1 and 2 is likely to undergo in solution first a simple ligand exchange reaction in which Sn₉⁴⁻ replaces one Mes ligand of CdMes₂ to form [Sn₉CdMes] units by a further replacement of Sn₉⁴⁻ unit and finally compound 1 was formed [15, 26, 28, 29, 32].

Although compound 1 was plagued by continuous positional disorder at the [(Sn₉)Cd(Sn₉)]⁶⁻ anion sites, the combination between X-ray single-crystal result of compound 2 and energy dispersive X-ray Spectroscopic (EDX) analyses confirm the presence of (Ni), Cd, and Sn and the ratio of them. In anions [(Sn₉)Cd(Sn₉)]⁶⁻ and [(Ni@Sn₉)Cd(Ni@Sn₉)]⁶⁻, the Cd atom is coordinated by two [Sn₉] and [Ni@Sn₉] subunits in an η³: η³ coordination mode forming sandwich complexes, and both have perfect inversion center at the cadmium atom position, which is analogous to [Ge₉ZnGe₉]⁶⁻ and [Sn₉HgSn₉]⁶⁻ [25, 29].

Two [Sn₉]⁴⁻ or [Ni@Sn₉]⁴⁻ subunits adopt a mono-capped square antiprism structure. The Sn-Sn distances of 1a are in the range of 2.9056(14) Å and 3.2826(15) Å, which are slightly shorter than those in [Sn₉]⁴⁻ (2.928(6)-3.308(5) Å). For cluster 2a, the Sn-Sn contacts and the Ni-Sn bond distances are normally compared to the reported [Ni@Sn₉]⁴⁻ [13].

All of the Cd-Sn distances (2.8913–3.3246 Å in 1a and 2.9658–3.2060 Å in 2a) are significantly longer than the sum of single-bond covalent radii 2.76 Å and the Cd-Sn single bonds in [MsSi{SiMeN(p-Tol)}₃Sn]₂Cd (2.676 Å) [30] and [Cd(SnB₁₁H₁₁)₄]⁶⁻ (av. 2.756 Å) [31], which indicates relatively weak Cd-Sn interactions. Interestingly, the distances of Cd-Sn8 in anion 1a are shorter than those in 2a, in contrast, the bond lengths of Cd-Sn2 and Cd-Sn3 are longer than those in 2a. This trend of bond length averaging in 2a may be due to the insertion of the Ni atoms. We also compared the bond lengths of 2a with those in [(Ni@Sn₉)In(Ni@Sn₉)]⁵⁻ and found that all of the Sn-Sn and Ni-Sn bonds in 2 are longer than those in [K(2.2.2-crypt)]₅[(Ni@Sn₉)In(Ni@Sn₉)].DMF (Full details of bond lengths are given in Table 1).

It is worth noting that cluster 2a is similar to the known [(Ni@Ge₉)Ni(Ni@Ge₉)]⁴⁻ [33] reported by Sevov from the view of the coordination way. Specifically, the central Cd and Ni ions of these two clusters coordinate respectively with two triangle faces of Ni@E₉ units (E = Sn and Ge) to form an approximately octahedral environment. To understand the relationship between the geometries and electronic structures of these above-mentioned clusters, we performed density functional theory at the PBE0/def2tzvp level of theory [34, 35]. The PCM polarizable conductor calculation model was performed with the solvent data for ethylenediamine (ξ = 12.9) to simulate the confining effects of the cationic environment in the solid-state in all cases [36, 37]. Default convergence criteria were used for the optimization tasks with Gaussian 09 program. The initial structures for optimizations of cluster 2a and [(Ni@Ge₉)Ni(Ni@Ge₉)]⁴⁻ were taken from the experimentally observed structures (single-crystal X-ray diffraction analysis) with a restriction of D_3d symmetry by adjusting the coordinates, but their optimized results show C_2h and C_i symmetries, respectively. Full details of the Cartesian coordinates and the optimized bond lengths are given in Supporting information. The deviation between optimized bond lengths and those corresponding experimental values are within a reasonable 2% range, which is a common observation in calculations on zintl chemistry that may be due to the absence of an explicit crystal environment in the computational model. Frequency calculations confirmed that these optimized structures correspond to their respective minima in the potential energy surfaces. A comparison of the optimized geometries exhibits the superficial similarities between these related clusters, but also some subtle differences (Fig. 2). The most obvious difference is that the Cd-Ni contacts of 4.31 Å effectively confirm non-bonding interactions while the Ni-Ni distances of 2.43 Å compare well with bonding distances in the corresponding clusters containing Ni-Ni bonds [38, 39]. The more subtle differences appear in the shape of triangles that coordinated with the Cd/Ni atoms and the E-E distances between the two Ni@E₉ units. In cluster 2a, the two Sn-Sn distances are 3.26 Å and 5.42 Å respectively, and the ratio between the two is 0.60. In contrast, the ratio of the two corresponding Ge-Ge distances is 1.04. In the D_2h-symmetric [Cu₄@Sn₁₈]⁴⁻ optimized at the same level of theory, representing the two E₉ cages are complete fusion, the corresponding ratio is 1.55 (Fig. S16 in Supporting information) [40]. Hence, cluster 2a, representing the completely separated state, can be viewed as a starting stage of fusion, which is further to form partly fusion cluster, such as [(Ni@Ge₉)Ni(Ni@Ge₉)]⁴⁻, and finally to achieve the total coalescence defined by [Cu₄@Sn₁₈]⁴⁻.

As for their electronic structures, the total cluster valence electrons of cluster 2a and [(Ni@Ge₉)Ni(Ni@Ge₉)]⁴⁻ are 110 and 106, respectively, indicating that two orbitals are occupied in the former but unoccupied in the latter. Hence, finding the two orbitals is very meaningful to understand those differences in structural chemistry. The HOMO of cluster 2a is an anti-bonding orbital constituted by the HOMOs of two Ni@Sn₉ units, which is an in-phase combination of Sn 5p orbitals in each triangular face coordinated to the bridging Cd atom. In a related matter, the LUMO+1 of the Ni₃ cluster is mainly localized on the Ge₉ fragments. Although the LUMO+1 orbital is an out-of-phase combination between each coordination triangular face, the net bond order is only 1/6, indicating some interactions between the two Ge₉ units and between the three atoms of the coordination triangular faces. Hence, if two electrons are removed, the triangular face would be expanded, which is an abovementioned difference in these clusters. For the HOMO-1, HOMO-2, and HOMO-3 of cluster 2a, they do not show any interactions between the endohedrally encapsulated Ni ions and the bridging Cd atom, but a strong anti-bonding interaction in LUMO of Ni₃ cluster was presented (the in-phase combination of Ni 3d orbitals with the lower energy which means they are also occupied). Hence, these orbitals on the Ni-Ni bonding are consistent with those experimental observations. This is another structural difference mentioned above between the two clusters.

To further understand the chemical bonding pattern of the title clusters, the adaptive natural density partitioning algorithm (AdNDP) was performed by using Multiwfn 3.8 program [41, 42]. Clusters 1a and 2a have a similar bonding situation because two embedded Ni ions can be considered as a zero-electron-donor to cluster bonding. Hence, we only analyze the bonding pictures of cluster 1a for an example. Cluster 1a has 90 valence electrons in total (18 × 4 |e| from 18 Sn atoms, 12 |e| from Cd atom and 6 |e| from the 6- cluster charge), providing 45 two-electron AdNDP bonding elements. The algorithm starts to search one-center two-electron elements, giving five d-type lone pairs (LPs) on Cd atom (Fig. S17 in Supporting information) and 18 s-type LPS on each Sn atom (Fig. 3A) with the occupied number (ON) of 1.99–2.00 |e| and 1.99–2.00 |e| respectively. Further search results found two 3c-2e σ bonds in Sn₃ triangular faces with ON = 1.97 |e| (Fig. 3B) and two 4c-2e σ bonds attributed to the two triangular faces coordinated with the transition metal Cd ions (Fig. 3C). Such 4c-2e σ bonds are reminiscent of cop-per-containing nonagermanide clusters such as Cu(NHC)[Ge₉{P(NH₂)₂}₃]⁻ and Cu[Ge₉{P(NH₂)₂}₃] and the first sandwich compound with the heteroatomic metal planar fragment, {(Ge₉)₂[η⁶-Ge(PdPPh₃)₃]}⁴⁻ [43, 44]. Finally, the remaining 36 electrons can be localized on 18 5c-2e Sn5 σ bonds with ON = 1.77–1.97 (Fig. 3D). In general, the title clusters can be viewed as the Cd ions bridging two Sn₉/Ni@Sn₉ fragments [42].

In conclusion, we present two metal clusters, [Sn₉CdSn₉]⁶⁻ (1a) and [(Ni@Sn₉)Cd(Ni@Sn₉)]⁶⁻ (2a) which contain two Sn₉⁴⁻ and Ni@Sn₉⁴⁻ subunits bridged by the Cd atom. The Ligand-free ternary Zintl anion [(Ni@Sn₉)Cd(Ni@Sn₉)]⁶⁻ is the first heterometallic cluster that transition metal atom connects two [M@E₉]^q− subunits. In our view, the Sn₉⁴⁻ and Ni@Sn₉⁴⁻ gradually replaced the Mes-ligands of the CdMes₂ to form clusters 1a and 2a with sandwich structures. DFT calculations indicate that larger cluster fusion can be modulated by electronic structure.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

This work was supported by the National Natural Science Foundation of China (Nos. 92161102 and 21971118) and the Natural Science Foundation of Tianjin City (Nos. 20JCYBJC01560 and B2021202077) to Z.-M. Sun.

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.cclet.2022.02.013.