Main

The isolation of ferrocene and the explanation of its molecular structure marked the origin of modern organometallic chemistry11,12,13. In the wake of this milestone discovery, sandwich complexes and their derivatives have had an unparalleled impact on the development and theoretical foundations of chemistry itself1,2. Sandwich compounds with two cyclopentadienyl ligands (Cp, C5H5−) are classically referred to as metallocenes. However, the strict definition of metallocenes is not obligatory in scientific terminology nowadays. Particularly in the case of the f elements, compounds with larger ligand systems, such as cyclooctatetraenediide, are also referred to as metallocenes.

The simplest molecular sandwich compounds, [M(CnHn)2]+/0/− (CnHn with a π-aromatic ring), can formally be considered to be the smallest entity. The next larger scaffold are triple-decker compounds [M2(CnHn)3]+/0/− (refs. 14,15). Following this thought, multidecker compounds of the general type [Mn(CnHn)n+1]+/0/− can then be considered the next (evolutionary) step of the sandwich complex family. Here, more than two metal ions are complexed by n − 1 bridging as well as two capping π-perimeter ligands yielding one-dimensional (1D) coordination oligomers. Today, the field of multidecker sandwich compounds is subject to intense research due to their intriguing properties and structural motifs resulting in possible applications such as nanowires3,4,5,6,7,8,9,10.

The concept of using higher dimensional metal-containing networks or specifically shaped structures is not new to organometallic chemistry16,17. However, despite more 70 years of research, sandwich compounds, in which the structure is exclusively assembled by direct metal–ligand interactions, are still limited to simple chains or related motifs18. Apart from this, some examples of circular oligomeric metallopolymers based on ansa-ferrocene or -ruthenocene derivatives are known. Here, interconnecting frameworks provided by ligand-ligand bonds by means of SiMe2– (refs. 19,20,21), CHR– (R = H, Me) (refs. 22,23,24) or C–C bonds have been reported25. However, these species do not feature any multidecker structural motifs.

Herein, we showcase the design, synthesis and characterization of an isomorphous series of circular sandwich compounds solely consisting of multiple, equivalent direct metal–ligand interactions. Reminiscent of classical metallocenes, we propose the term ‘cyclocenes’ for this fundamentally new compound class.

Synthesis of cyclic sandwich compounds

In lanthanide chemistry, only 1D-multidecker sandwich-based oligomers could so far be constructed as supramolecular structures. Typical examples are [MII/IIIn(Cot)m] polydecker (Cot = C8H82−) compounds (M = Sc, Y, La, Ce, Nd, Eu, Tb, Ho, Tm, Yb; up to n = 30 for Eu; m = n, n − 1, n + 1), which are commonly prepared by laser evaporation and molecular beam methods and obtained as a mixture of compounds with varying chain lengths3,4,5,6,7. Isolation and structural characterization on a preparative scale have not been reported. Another approach is based on the direct reduction of neutral Cot derivatives by elemental lanthanides in liquid ammonia, yielding [LnIIn(Cot)m] (Ln = Eu, Yb) species6,26,27. In the absence of a suitable donor, intermolecular assembly to polymeric 1D-multidecker stacks was claimed6. Despite manifold publications describing the physical properties of such multidecker compounds, no solid-state structures have been reported so far.

To form a ring-shaped sandwich compound, n metal ions would require complexation by n π-perimeter ligands to give a closed topology. Two simple prerequisites for a circular sandwich compound can be formulated: (1) a cyclic system would ideally show an overall neutral charge. Thus, a homoleptic system with an equal charge magnitude on the metal centres and ligand moieties appears expedient. (2) A constant bending should be present in each sandwich subunit to construct the target assembly. The [LnIIn(Cot)m] (Ln = Eu, Yb) polydeckers mentioned above meet the first criterion. Strong evidence that the second prerequisite can also be satisfied was recently reported by our group. We showed that the introduction of bulky Cot derivatives into the samarium quadrupledecker [SmIII/II/III3(CotTIPS)4] (CotTIPS = 1,4-(iPr3Si)2C8H62−) induces substantial unidirectional bending. In particular, the divalent subunit of this complex shows a pronounced bent structure with a Ct–Sm–Ct (Ct, ligand centroid) angle accounting for roughly 160° (ref. 28). Assuming that the magnitude and unidirectional character of this bending is retained in an analogue [LnIIn(CotTIPS)n] complex, a circular system of about 18 units might be formed if only a slight deviation from a 160° bending angle takes place.

Reactions of the divalent alkaline earth and lanthanide iodides SrIII2 or [MIII2(thf)2] (M = Sm, Eu, Yb) with [K2(CotTIPS)] in tetrahydrofuran (THF) resulted in monomeric THF-solvated [MII(thf)3(η8-CotTIPS)] complexes (1, M = Sr, Eu, Yb). All complexes tend to lose the coordinated THF molecules rapidly on removing the mother liquor (Fig. 1; further details on this and on the synthesis can be found in the Supplementary Information). The samarium species could not be obtained as crystalline material. The molecular structures in the solid-state of 1 resemble the piano stool motif described previously for the related complex [YbII(pyr)3(η8-Cot)] (pyr, pyridine)26 (the Supplementary Information gives structural details of 1). Attempts to fully characterize 1 were unsuccessful because 1-Sr and 1-Eu lose all solvent molecules rapidly on drying the isolated crystalline materials. In the case of 1-Yb, one equivalent of THF is retained. This is a first indication that THF free compounds of the type [LnIIn(CotTIPS)n] are formed on drying 1-Sr or 1-Eu. Consequently, THF coordination must be suppressed during crystallization to allow the formation of oligomeric systems, exclusively containing MII-Cot interactions. However, several crystallization attempts involving non-coordinating solvents, such as toluene or n-heptane, exclusively yielded amorphous material. Addition of small amounts of THF to toluene crystallization batches eventually facilitated the isolation of single crystals (Supplementary Information). Single-crystal X-ray diffraction analysis revealed that the metal ions of similar ionic radii SrII, SmII and EuII form cyclo-octadeca-sandwich complexes [cyclo-MII(μ-η8:η8-CotTIPS)]18 (2, Fig. 2).

Fig. 1: Cyclocene synthesis.
figure 1

Synthesis of [MII(thf)3(η8-CotTIPS)] (1), [cyclo-MII(μ-η8:η8-CotTIPS)]18 (2, M = Sr, Sm, Eu) and [cyclo-YbII(μ-η2:η8-CotTIPS)(thf)]4 (3-Yb). RT, room temperature.

Full size image
Fig. 2: Cyclocene structures.
figure 2

a, Molecular structures in the solid state of [cyclo-SmII(μ-η8:η8-COTTIPS)]18 (2-Sm). b, [cyclo-SrII(μ-η8:η8-CotTIPS)]18 (2-Sr). c, [cyclo-EuII(μ-η8:η8-CotTIPS)]18 (2-Eu). d, [cyclo-YbII(μ-η2:η8-CotTIPS)(thf)]4 (3-Yb). Hydrogen atoms are omitted, and SiiPr3-groups are shown transparently for better clarity. Note that the cyclo-octadeca-sandwich complexes [cyclo-MII(μ-η8:η8-CotTIPS)]18 2-Sm, 2-Sr and 2-Eu are depicted from different perspectives for improved visualization of the circular structures. However, they are isostructural.

Full size image

The M–Ct distances are similar throughout the series and show only slight variations within the corresponding cyclocene (2-Sr 2.209(2)–2.263(1) Å, 2-Sm 2.189(1)–2.2518(11) Å, 2-Eu 2.1783(7)–2.2276(7) Å). All Ct–M–Ct angles show values marginally above the postulated 160°. This is caused by the ring plane being slightly puckered (2-Sr 160.85(6)–166.11(7)°, 2-Sm 161.61(6)–165.13(6)°, 2-Eu 161.89(4)–165.27(4)°). Likewise, all M–M–M angles are of a similar magnitude (2-Sr 157.36(4)–163.46(4)°, 2-Sm 157.36(3)–161.83(3)°, 2-Eu 157.52(2)–161.40(2)°), with inner angle sums close to 360° (2-Sr 366.04°, 2-Sm 363.12°, 2-Eu 364.61°), resulting in almost ideally circular structures, with a slight elliptical deformation. The inner diameters of the ring systems account for roughly 17 Å, whereas the outer diameters amount to roughly 38 Å (Supplementary Tables 9–11).

By contrast, by using the smaller YbII ion a formal cyclo-tetra-ytterbocen [cyclo-YbII(μ-η2:η8-CotTIPS)(thf)]4 (3-Yb, Fig. 2, bottom right) was obtained in which the open face of the YbII coordination sphere is saturated by one THF ligand. We assume that the κ1-THF-YbII bonds are strengthened compared to SrII, SmII and EuII due to the higher Lewis acidity of YbII. Thus, the contracted motif is favoured over a cyclo-octadeca-sandwich structure (see below for quantum chemical calculations).

To probe the structural composition of compounds 2 in solution, diffusion ordered nuclear magnetic resonance spectroscopy (DOSY-NMR) experiments were performed on the diamagnetic complex 2-Sr and the paramagnetic complex 2-Eu. In THF-d8, these experiments yielded hydrodynamic radii of Rh = 6.9 Å (2-Sr) and 5.3 Å (2-Eu), indicative of monomeric species of the type [MII(thf)3(η8-CotTIPS)] (1) being present. In the non-coordinating solvent toluene-d8 the hydrodynamic radii account for Rh = 5.3 Å (2-Sr) and 5.0 Å (2-Eu) (Supplementary Information). Although a comparison of DOSY measurements in different solvents should be considered with caution, these values significantly deviate from what would be expected if the cyclo-octadeca-sandwich structures were retained in solution. Consequently, the cyclic motif is most probably formed by means of self-assembly pathways on crystallization. Similarly, plumbocene [Pb(Cp)2]29,30 can form different chains or a hexameric cyclic scaffold with Pb-η5-Cp-Pb bridges, depending on the crystallization conditions. However, the lead atom is coordinated by three η5-Cp-rings and thus does not resemble a classical sandwich motif.

Photoluminescence properties

The solid cyclocenes 2-Eu and 3-Yb demonstrate orange and green photoluminescence (PL) centred at 570 and 546 nm, respectively, which slightly redshifts to 583 and 550 nm at cryogenic temperatures (Fig. 3). The excitation onset at about 520 nm corresponds to the visual appearance of the samples as yellow (2-Eu) or yellow-greenish (3-Yb) polycrystalline powders, as well as to their absorption spectra (Supplementary Fig. 33, note that the PL excitation (PLE) and absorption spectra overemphasize the intensity of weak bands due to a high optical thickness of the sample preparations and non-validity of the Beer–Lambert law). The emission of 2-Eu occurs with a high quantum efficiency of ΦPL = 60% at room temperature (estimated as approaching 100% at 5 K) and decays monoexponentially on ns-pulsed laser excitation at 337 nm with τ = 2.1/1.7 µs at 5/295 K (Supplementary Fig. 31). Accordingly, it shows only a moderate temperature dependence. A similar PL has been observed for other complexes of divalent europium, including sandwich compounds, and can be assigned to a 4f65d1 → 4f7 transition of the Eu2+ ions in 2-Eu (refs. 6,31). The low-energy absorption or PLE bands also relate to the f–d transitions of Eu2+ as no contribution of molecular orbitals of the CotTIPS ligands is expected above roughly 400 nm absorption or PL excitation wavelength32. The energies of emissive f–d transitions of divalent europium and ytterbium in related compounds are typically comparable. In consistence, the emissions of 2-Eu and 3-Yb are close spectrally (Fig. 3)33.

Fig. 3: Photoluminescence properties.
figure 3

Excitation (PLE) and emission spectra of 2-Eu and 3-Yb in solid (polycrystalline) state at different temperatures between 5 and 295 K. The PLE and PL spectra were recorded and excited at 580 and 350 (2-Eu), and 550 and 350 nm (3-Yb), respectively. The inserts show the integral emission intensity (normalized to unity) as a function of temperature.

Full size image

Other PL properties differ, however, notably, reflecting in part the different complexation patterns discussed above. In contrast to 2-Eu, the PL intensity and kinetics of 3-Yb show a strong temperature dependence. The PL of the latter is relatively weak at ambient temperature, but increases roughly 40-fold on cooling, with ΦPL rising from 1.0% at 295 K to about 37% at 5 K as estimated from the emission spectra. The PL decay is clearly non-monoexponential within the whole temperature interval probed. The effective (average) decay time derived from biexponential fits varies from 0.78 μs at room temperature to 23 μs at 170 K and 220 μs at 5 K (Supplementary Fig. 32). A large radiative lifetime (roughly 600 and 80 μs at 5 and 295 K) indicates a spin-forbidden 4f135d1 → 4f14 emissive transition in 3-Yb. Its efficient non-radiative quenching is probably contributed to by the presence of THF molecules in the coordination sphere (Fig. 3) and their vibrational coupling to the Yb2+ ions. This might also explain the different temperature effects on the absorption onset seen in the excitation spectra (note to Supplementary Fig. 33).

Quantum chemical calculations

For a more detailed understanding of the ring formation, density functional theory calculations were performed using TURBOMOLE34,35. Details are given within the Supplementary Information, and optimized structures of all compounds in the appendix of the Supplementary Information.

At first, the bonding situation in the new class of sandwich complexes was investigated. For this, the smallest unit, namely [MII(CotTIPS)] (M = Sr, Sm, Eu, Yb), was extracted from the cyclocene structure, and its electronic properties were investigated. The bonding between the metal ions and the CotTIPS units in the cyclocenes is predominantly of electrostatic nature. This is evident from the absence of significant covalent interactions, as visible, for example, from Foster–Boys localization36, which did not yield any orbital with noteworthy contributions of both the metal and any of the ring carbons.

In fact, natural population analysis37 charges of the metal ions amount to 1.76/1.27/1.28/1.26 elementary charges for M = Sr, Sm, Eu, Yb (Supplementary Table 12), similar to analogue charges for metal halides of the type MIIX2 (M = Sr, Sm, Eu, Yb; X = Cl, Br, I), which are classically considered ionic compounds containing divalent metal ions (Supplementary Table 14). Also, the absence of s valence electrons at the metal ions (from the same analysis, Supplementary Table 12) in the [MII(CotTIPS)] fragments is in accordance with chemical intuition. Explicitly including scalar-relativistic effects and spin-orbit coupling does not change this picture. Virtually the same results were obtained for the larger fragments [MII3(CotTIPS)4]2− (for M = Sr, Sm, Eu) considered below (Supplementary Table 13).

The pronounced bending observed in the cyclocene structures is due to steric reasons, which becomes evident when studying the Ct–M–Ct bending angle for a section of the ring consisting of three metal ions and four ring moieties using substituents with different steric demands. In detail, the steric demand of the ring substituents was varied from H ([MII3(Cot)4]2−) to TMS (TMS is trimethylsilane, Me3Si) ([MII3(CotTMS)4]2−, CotTMS = 1,4-(Me3Si)2C8H62−) to TIPS ([MII3(CotTIPS)4]2−). The Ct–M–Ct angle of the central ring-M-ring unit is shown in Table 1. For Cot, a linear arrangement with a Ct–M–Ct angle close to 180° is found, allowing for the formation of polymeric 1D stacks as outlined above. When TMS groups are introduced, a bending is induced, leading to Ct–M–Ct angles of roughly 178°. The transition to the more sterically demanding TIPS groups leads to a more pronounced bending with Ct–M–Ct angles of 161–163°, in good agreement with the experimentally obtained values, thereby allowing for the formation of 18-membered ring structures. This indicates that the cyclocene size might be tailored by modifying the bulk of the Cot substituents38.

Table 1 Calculated Ct–M–Ct angles
Full size table

Obviously, the steric demand of the substituents is responsible for the bending of the individual ring-M-ring units. However, this sterically induced bending leads to the formation of the observed cyclic structures only when it occurs in a unidirectional manner. Otherwise, the formation of chain-type structures is also conceivable. The difference between these two options was investigated by considering the Sr model systems partially depicted in Fig. 4. Relevant energy differences based on these structures are given in Supplementary Tables 15 and 16. The following picture emerges from the calculations: (1) arrangements representing sections of the ring molecule are energetically preferred by 6 to 27 kJ mol−1 over the lowest energy arrangement unsuitable for ring formation. (2) The energy gain for ring formation is about 140 kJ mol−1, estimated from the reaction energy of [Sr(CotTIPS)]n−1 + [Sr(CotTIPS)] → [Sr(CotTIPS)]n for n = 2 to n = 17. For the ring closure (n = 18), it is more than twice as large, 322 kJ mol−1. (3) Dispersion interactions are only slightly larger (2–16 kJ mol−1) for arrangements suitable for ring formation than for unsuitable arrangements. However, dispersion interactions make up more than half of the reaction energy, making them crucial for stabilizing the structures.

Fig. 4: Quantum chemical calculations.
figure 4

Sr-based model systems to investigate the difference between the formation of a cyclic and chain-type structure. H atoms and iPr groups are omitted for clarity. Sr is drawn in green, Si in beige, C in grey. Given energies (in kJ mol−1) refer to single-point calculations at the PBE0/def2-TZVP/D4 level of theory for structures optimized at the PBE/SV(P)/D4 level. Further details and computational settings are given within the Supplementary Information.

Full size image

To summarize, the ability to form a ring is due to the specific design of the CotTIPS ligand, but the driving force for the ring formation is the energy gained by its closure, which is missing when a non-cyclic structural motif is formed.

Finally, we rationalized the formation of a four-membered ring in case of Yb, contrary to the formation of 18-membered rings for Sr, Sm and Eu. As discussed above, cyclocene formation is mainly determined by two factors, namely ionic interactions between the metal atom and the Cot derivative as well as its steric demand. Whereas the latter is the same for all CotTIPS compounds, the ionic radius of YbII (1.14 Å) differs significantly from that of SrII (1.26 Å), SmII (1.27 Å) and EuII (1.25 Å) (coordination number 8). This in turn reveals that the observed structural differences are apparently induced by the smaller ionic radius of YbII compared to SrII, SmII and EuII (ref. 39). The extra nuclear charge of Yb (Z = 70) compared to Sm (Z = 62) and Eu (Z = 63) is only weakly shielded by the extra f electrons, leading to a significant increase in the effective nuclear charge Zeff and thus a decrease in the ionic radius. A smaller ionic radius indicates two consequences that are relevant here: (1) the Lewis acidity of YbII is increased relative to SrII, SmII and EuII, leading to a higher tendency to form Lewis acid–base adducts when a Lewis base is present. This is evidenced, for instance, by the fact that the solvates 1 readily lose all solvent ligands on drying in vacuo in the case of 1-Sr and 1-Eu, whereas for 1-Yb one equivalent of THF is retained. In fact, the calculated binding energy of THF to [MII(CotTIPS)] (M = Sr, Sm, Eu, Yb) fragments is about 15 kJ mol−1 higher for YbII than for the other metals (Supplementary Table 17). (2) The ligands in a sandwich complex show smaller M-ring distances, in line with smaller inter-ring distances, leading to increased steric repulsion between two ring ligands. This, together with the presence of an extra THF prevents the formation of an 18-membered ring structure for Yb and rationalizes the formation of a smaller ring.

Conclusion

The isolation and structural characterization of the cyclocene family [cyclo-MII(μ-η8:η8-CotTIPS)]18 (M = Sr, Sm, Eu) opens a fundamental, new chapter in the chemistry of sandwich compounds. The formation of these fascinating species has been rationalized aided by quantum chemical methods and translated into basic design principles towards future ventures into the synthesis of cyclocenes. The pronounced bending observed in the cyclocene structures is caused by the steric demand of the Cot substituents. The driving force for the ring formation is the energy gained by its closure, which is largely influenced by dispersion interactions, and missing when a non-cyclic structural motif is formed. The stability and size of cyclocenes is assumed to be adjustable by specifically tailoring the used ligand systems. Considering this prospect in the light of classical sandwich compounds and their well-established properties, this textbook example of the cyclocene motif is expected to find further application in functional organometallic chemistry.

Methods

Synthetic method

Experiments were carried out under a dry, oxygen-free argon atmosphere using Schlenk-line and glovebox techniques. All solvents and reagents were rigorously dried and deoxygenated before use. All compounds were characterized by single-crystal X-ray diffraction studies. Further details are available in the Supplementary Information.

Accession codes

The X-ray crystallographic coordinates for structures reported in this article have been deposited at the Cambridge Crystallographic Data Centre (CCDC), under deposition nos. CCDC 2194266 (2-Eu), 2194267 (2-Sm), 2194268 (2-Sr) and 2194309 (1-Eu), 2194310 (1-Sr) and 2194311 (1-Yb).