Snap-shots of cluster growth: structure and properties of a Zintl ion with an Fe3 core, [Fe3Sn18]4−

The endohedral Zintl-ion cluster [Fe3Sn18]4− contains a linear Fe3 core with short Fe–Fe bond lengths of 2.4300(9) Å. The ground state is a septet, with significant σ and π contributions to the Fe–Fe bonds. The Sn18 cage is made up of two partially fused Sn9 fragments, and is structurally intermediate between [Ni2CdSn18]6−, where the fragments are clearly separated and [Pd2Sn18]4−, where they are completely fused. It therefore represents an intermediate stage in cluster growth. Analysis of the electronic structure suggests that the presence of the linear Fe–Fe–Fe unit is an important factor in directing reactions towards fusion of the two Sn9 units rather than the alternative of oligomerization via exo bond formation.


Introduction
The chemistry of Zintl ions, and in particular those containing endohedral metals, has been the subject of several recent reviews, [1][2][3][4][5][6][7][8][9] and applications in catalysis and materials chemistry are beginning to emerge. 10,11The vast majority of these clusters are relatively small (14 main-group atoms or fewer) and contain a single transition metal ion, oen with a closed-shell d 10 conguration: classic examples include the icosahedral triad [Ni/Pd/PtPb 12 ] 2− , 12 but the range of encapsulated metals now includes much of the d block.Larger clusters containing multiple transition metals are much less common but they offer the possibility of unusual magnetic phenomena and/or metalmetal bonding.Amongst the few known examples, 13 the Ge 18 series [Ni 2 InGe 18 ] 5− , 14 [Ni 3 Ge 18 ] 4− , 15 and [Pd 2 Ge 18 ] 4− (ref.16) (Fig. 1) maps out a progressive fusion of the two Ge 9 polyhedra which are well separated in [Ni 2 InGe 18 ] 5− but fully coalesced in the Pd cluster.[Ni 3 Ge 18 ] 4− appears to be an intriguing intermediate case, where the fusion is only partially complete.It is far from clear how these clusters are actually formed in situ, but it is certainly plausible that the stepwise fusion of pre-formed polyhedral E 9 or ME 9 fragments is involved.Indeed, Sevov and Goicoechea proposed the fusion of NiGe 9 and Ni 2 Ge 9 units as a possible route to formation of [Ni 3 Ge 18 ] 4− , 17 and Dehnen's analysis of fragmentation patterns for [TaGe 4 As 8 ] 3− and [TaGe 6 As 6 ] 3− identied cluster fragments such as [Ge 2 As 2 ] 2− and [Ge 3 As] 3− that may play a role in growth. 18The challenge from a synthetic perspective is that these component polyhedra typically carry high negative charges, and so their close approach incurs a high coulombic penalty.Transition metal ions that can bridge two polyhedral units may, therefore, play an important role in fusion by buffering these repulsions and also, potentially, by removing excess electron density through Fig. 1 Cluster fusion vs. cluster oligomerisation of E 9 polyhedra, E = Ge, Sn. [14][15][16][19][20][21][22][23][24][25][26][27][28][29] the extrusion of metal in the elemental form. A furtr complication is that the oxidative fusion of clusters is, at least in principle, in competition with oxidative oligomerisation via the formation of exo E-E bonds (Fig. 1(b)).This phenomenon is well established in Ge chemistry where linked chains of Ge 9 are known.17,19 A deeper understanding of the factors that control cluster growth and the balance between fusion and oligomerisation may provide access to a wider range of element combinations and compositions, and to tailored structural, magnetic and catalytic properties.
In this paper, we extend our recent work on the Zintl-ion chemistry of tin by reporting the synthesis of a new cluster, [Fe 3 Sn 18 ] 4− , which has a linear Fe 3 chain and Fe-Fe bond lengths of 2.4300( 9

Structure and properties of [Fe 3 Sn 18 ] 4−
The reaction of ethylenediamine (en) solutions of K 4 Sn 9 with [K(thf)Fe(O t Bu) 3 ] 2 (thf = tetrahydrofuran) results in the formation of the tri-iron cluster [Fe 3 Sn 18 ] 4− in the form of its Sn 18 ] (1).Electrospray ionisation mass spectrometry (ESI-MS) of freshly-prepared DMF (DMF = dimethylformamide) solutions of 1 reveals a peak attributable to the dianion [Fe 3 Sn 18 ] 2− (m/z 1152.0323note the peak-to-peak separations of 0.5 between isotopologues that conrm the −2 charge, Fig. 2(c)) and also a very weak signal assigned to the cation-dianion pair [K(2.2.2-crypt)Fe 3 Sn 18 ] − (m/z 2719.2300).It is common to observe only singly charged anions in the ESI-MS of Zintl clusters, but the large size of the Fe 3 Sn 18 unit reduces the coulomb repulsion in the dianion to the extent that it is not ionized under the prevailing conditions. 1 crystallises in the monoclinic space group P2 1 /c and the unit cell contains a single anionic [Fe 3 Sn 18 ] 4− cluster with four [K(2.2.2-crypt)] + cations (Fig. 2(a) and (b), CCDC 2170116).The Sn 18 unit adopts a D 3d -symmetric structure based on two Sn 9 polyhedra in a staggered, face-to-face arrangement, with a chain of three Fe centers aligned along the principal axis.In this section and the following discussion of the electronic structure, we focus rst on the Fe 3 chain, where Fe-Fe bonding is the primary interest, before turning to the Sn 18 cage which we try to place in the wider context of Zintl-ion chemistry.The Fe-Fe bond lengths of 2.4300(9) Å in 1 are remarkably short, much shorter than those in the other known Fe 2 -containing Zintl cluster, [Fe 2 Ge 16 ] 4− (2.636(3) Å).Even shorter bonds are known in Fe 2 dimers such as the Fe I Fe I paddlewheel complex 30 (2.127 Å) and the (as-yet unknown) Fe 2 C 30 (2.10 Å). 31 Direct comparison with other Fe 3 chains is restricted to classical coordination compounds such as Guillet's bis[(trimethylsilyl)amido]pyridine complex (Fig. 3(a), referred to henceforth as Fe 3 L 3 ) where the Fe-Fe bond lengths are 2.4416(5) Å (ref.32) and to the [Fe 3 (DpyF) 4 ] 2+ complex (DpyF = dipyridylformamide) rst synthesised by Cotton and Murrillo 33 and subsequently studied by Hillard and co-workers, 34 where the Fe-Fe bond lengths are longer, at 2.7838(5) Å.These two Fe II Fe II Fe II complexes share a common S = 6 ground state and a common formal s bond order of 0.25 (per Fe-Fe bond), but differ in the distribution of electrons in the levels of p symmetry, with only Fe 3 L 3 having an additional p component to the Fe-Fe bond.Correlations between bond order and bond length are notoriously difficult when bridging ligands are present, but nevertheless the similar bond lengths in [Fe 3 Sn 18 ] 4− and Fe 3 L 3 offers an initial indication that Fe-Fe p bonding may also be signicant in the former.We return to this question in the following discussion of the electronic structure of the cluster.Turning our focus now to the structure of the Sn 18 cage, we note rst that the cluster can be viewed as two FeSn 9 units, bridged by a third Fe center.We can make useful comparison to the pair of closely-related clusters identied in Fig. 3

Electronic structure
Geometry optimisations using the PBE functional indicate that the lowest energy state for [Fe 3 Sn 18 ] 4− is a spin septet (S = 3), 7 A 2g , with optimised Fe-Fe bond lengths of 2.45 Å, in excellent agreement with the available X-ray data (Table 1).The Fe-Sn and Sn-Sn bond lengths are also fully consistent with experiment.Despite multiple attempts, we have been unable to measure reproducible magnetic susceptibilities to conrm the paramagnetism of [Fe 3 Sn 18 ] 4− : this likely reects the challenges in producing a homogeneous sample, and in avoiding oxidative degradation during the course of the experiment.The spin-polarised Kohn-Sham eigenvalues and eigenfunctions of the 7 A 2g ground state are collected in Fig. 4: levels that are localised primarily on the Fe 3 chain are shown in green while those localised primarily on Sn are in grey.The same data, in the form of projected density of states (PDOS) and overlap projected density of states (OPDOS), is presented and discussed in the ESI, Fig. S8.† Of the 100 valence electrons of the cluster, we can identify 22, colored green, that are distributed over the 15 linear combinations of Fe 3d orbitals in Fig. 4 (4e g , 5a 1g , 4e u , 7a 2u , 6e g , 5e u , 7e g , 8e g and 9a 1g in the a set).It is notoriously difficult to assign oxidation formal states in endohedral Zintl clusters, where transition-and main-group metal orbitals are typically well mixed, but the presence of 22 valence electrons indicates a Fe 2 3 + chain, and hence a Sn 18 cluster in a −6 oxidation state.
Of the 15 metal-based orbitals, only one, the strongly Fe-Fe-Fe s anti-bonding 9a 1g orbital, is vacant in both spin-a and spin-b manifolds, while the complementary s bonding and non-bonding orbitals, 5a 1g a, 7a 1g b and 5a 2u a, 7a 2u b, are doubly occupied: the s 2 s nb2 s* 0 conguration gives a net s bond order of 0.5 per Fe-Fe bond.Fe-Fe p and d interactions are mixed in the orbitals of e g and e u symmetry, but the p interactions are primarily contained in 4e g , 4e u and 8e g in the a manifold, 5e g , 7e u and 10e g in b.The prominent positive and negative peaks in the OPDOS shown in ESI, Fig. S8, † corresponding to 5e g b (p bonding) and 10e g b (p antibonding), respectively, conrm the very signicant p overlap.The p 4 p nb4 p* 2 conguration then denes a formal Fe-Fe p bond order of 0.5 per bond.There are no large peaks in the OPDOS for the orbitals with dominant Fe-Fe d symmetry (5e u a, 7e g a, 9e g a, 8e u b, 8e g b, 9e g b), so d bonding can be assumed to be negligible, as might be expected at a distance of 2.4300(9) Å.The overall formal Fe-Fe bond order is therefore 1.0 per Fe-Fe   + unit and, hence, a Ge 18 cluster in the −6 oxidation level.Switching our focus now to the Sn 18 cage, we can identify a single vacant orbital, 8a 2u , picked out in red in Fig. 4, that has Sn-Sn s* character between the Sn 3 faces bound to the central Fe atom (Sn8−Sn9 ′ in Fig. 2).This orbital, along with its doubly-occupied Sn-Sn bonding counterpart, generates a 6-center-2-electron bond that links the two Sn 9 units.

Cluster fusion vs. cluster oligomerisation
In the previous section we have established a link between the new cluster [Fe 3 Sn 18 ] 4− and [Ni 3 Ge 18 ] 4− through their common oxidation level of −6 for the E 18 cluster unit.In this section, we try to identify broader relationships between the family of clusters with 18 tetrel vertices (E 18 ) but rather different structures.Amongst these, we can pick out the two pairs, [Ni 4− appear to be precisely intermediate between the two limits, with two partially but not fully coalesced E 9 units, consistent with the formal charge assignment of Sn 18 6− .
If we wish to analyse the electronic origins of these structural trends we are faced with the immediate problem that in some cases the clusters contain 3 transition metal ions but in others only 2. In order to circumvent this difficulty, we choose to focus on the electronic structure of the empty cage, Sn 18 , and explore its dependence on charge state: −8 / −6 / −4.The relationship between structure and charge state can be made explicit by the Walsh diagram for the isolated E 18 cluster shown in Fig. 5 (calculated using extended Hückel theory).This gure is constructed by extracting the structures of the Sn 18 units from DFT optimisations of [Ni 2 CdSn 18 ] 6− , [Fe 3 Sn 18 ] 4− and [Pd 2 Sn 18 ] 4− and interpolating between these three geometries.A comment on the choice of reaction coordinate is necessary here.The fusion of the two Sn 9 units proves to be highly asynchronous: the structural impact of the rst 2-electron oxidation is very different from the second 2-electron oxidation.In such circumstances, no single structural parameter can adequately capture the changes occurring across the entire spectrum, from 2 × Sn 9 4− on the le to Sn 18 4− on the right.We therefore choose to identify two distinct Sn-Sn distances that serve as independent measures of structural change.The Sn8−Sn9 ′ distance is closely related to the distance between the centroids of the two Sn 9 units, and it varies rapidly as we go from 2 × Sn 18 4− to Sn 18 6− , and then more slowly from Sn 18 6− to Sn 18 4− .The Sn8−Sn9 ′ distance, in contrast, varies strongly in the le half of the diagram, but is relatively constant as we move from Sn 18 6− to Sn 18 4− .We can, therefore associate the rst 2-electron oxidation with a relative motion of the two Sn 9 units towards each other, such that both Sn4−Sn9 ′ and Sn8−Sn9 ′ contract.The second 2-electron step is then associated almost exclusively with the formation of the Sn4−Sn9 ′ bonds, with little further change in Sn8−Sn9 ′ .At the separated limit (le hand side of Fig. 5) there is a total of 40 low-lying valence orbitals (up to 7a 2u ) that can accommodate 80 valence electrons, the count for Sn 18 8− .The transition from this separated limit to the intermediate structure typical of [Fe 3 Sn 18 ] 4− (or [Ni 3 Ge 18 ] 4− ) involves a reduction in the separation between the centroids of the two Sn 9 units, resulting in contraction of both the Sn8−Sn9 ′ and Sn4−Sn9 ′ distances.The result is the rapid destabilisation of a single orbital, 7a 2u , that is antibonding across Sn8−Sn9 ′this is the Sn-Sn antibonding orbital discussed previously in the context of Fig. 4 (where it was labelled 8a 2u due to the presence of a lower-lying Fe/Ni-based level of the same symmetry that is obviously absent in the empty cluster).In the second step, from the intermediate structure found in [Fe 3 Sn 18 ] 4− to the fully coalesced one in [Pd 2 Sn 18 ] 4− , the Sn4−Sn9 ′ distance contracts from 4.22 Å to 3.10 Å, causing a rapid destabilisation of a second cluster-based orbital, 7a 1g , which is bonding with respect to the Sn8−Sn9 ′ contact but strongly anti-bonding with respect to Sn8−Sn9 ′ .We note here that Lin and co-workers have also analysed the fusion of two PdSn 9 units from the perspective of the 'principal interacting orbital' model, [37][38][39] where they identied a s-symmetry interaction between 'principal interacting orbitals' localised on the Sn4 and Sn9 ′ atoms.To the extent that the structurally characterised clusters illustrated in Fig. 1 can be viewed as snapshots of the oxidative coalescence of two separated Sn 9 clusters, it seems that the 4-electron oxidation of Sn 18 8− to Sn 18 4− is a rather asynchronous one, with the two units coming together rst via the formation of Sn8−Sn9 ′ bonds (shown in red in Fig. 5), followed by a distinct second 2-electron oxidation step that leads to formation of the Sn4−Sn9 ′ bonds (shown in blue in Fig. 5), completing the fusion of the two units.§ In the introduction we noted that there is, in principle, a competing pathway for oxidative coupling of Zintl clusters that leads to oligomerisation via exo bond formation rather than fusion to form a single ellipsoidal cage.This is precisely what is observed in the oxidation of [AgSn 18 ] 7− (ref.26) to [AgSn 18 ] 5− (Fig. 1) where the charge on the Sn 18 unit (assuming a redox non-innocent Ag + ion) is reduced from −8 to −6, precisely the same as in the [Ni 2 CdSn 18 ] 6− to [Fe 3 Sn 18 ] 4− comparison.What, then, are the factors that determine the preference for fusion of the two Sn 9 units in [Fe 3 Sn 18 ] 4− with retention of three-fold rotational symmetry but oligomerisation in [AgSn 18 ] 5− ?From an electronic perspective, the Sn 18 units in the two clusters both have a 2-electron bond linking the two Sn 9 unitsthe only difference is that in [Fe 3 Sn 18 ]   38,39 and Sn 36 nanorod 27 where again there is no underlying metal-metal bonded framework to oppose the bending.

Summary and conclusions
In this paper, we have reported the synthesis and structure of a new Zintl-ion cluster, [Fe 3 Sn 18 ] 4− , containing a linear Fe 3 chain with short Fe-Fe bond lengths of 2.4300(9) Å. Electronic structure analysis indicates the presence of both Fe-Fe s and p bonding, with a formal net bond order of 1.0 ( leading rst to partial fusion of the two cages and then to their complete coalescence.One of the obvious challenges in forming ever larger Zintl ions from smaller fragments is that the latter carry high negative charges, so their close approach necessarily involves a substantial coulomb barrier.The identication of [Fe 3 Sn 18 ] 4− as an intermediate stage of cluster fusion presents the intriguing possibility that the central metal cation may act as a buffer, templating the close approach of the anionic components.Transfer of electron density from the main-group cage to the transition metal may then drive the fusion of the two polyhedral fragments, with concomitant reduction of the cations and their expulsion as metal atoms, as is observed, for example, in NaSi, where elevated pressures lead to the formation of Na metal with concomitant amorphisation of the Si 4 cluster units through Si-Si bond formation. 40The presence of Fe-Fe bonding appears to play an important role in this process by preventing bending at the central metal atom, directing the reaction towards cluster fusion rather than the competing oxidative oligomerisation observed in the [AgSn 18 ] 7−/5− pair.

Materials and reagents
All manipulations and reactions were performed under a nitrogen atmosphere using standard  ).In a 10 mL vial, K 4 Sn 9 (122 mg, 0.100 mmol) and 2.2.2-crypt (113 mg, 0.3 mmol) were dissolved in en (ca. 3 mL) and stirred for 30 min, resulting a dark brown solution.Then [K(thf)Fe(O t Bu) 3 ] 2 (33 mg, 0.043 mmol) was dispersed in toluene (0.5 mL), producing a light pink suspension, and then added dropwise to the above mixture.Aer stirring for 3 hours at room temperature, the resulting brown solution was ltered through glass wool and transferred to a test tube, then carefully layered by toluene (ca. 3 mL) to allow for crystallisation.Small brown block-like crystals of 1 (10% yield based on the K 4 Sn 9 precursor) were isolated aer two weeks.

X-ray crystallography
Crystallographic data for 1 were collected on Rigaku XtalAB Pro MM007 DW diffractometer with graphite monochromated Cu Ka radiation (l = 1.54184Å).The crystal structure was solved using direct methods and then rened using SHELXL-2014 (ref.42) and Olex2, 43 with all non-hydrogen atoms rened anisotropically during the nal cycles.All hydrogen atoms of the organic molecule were placed by geometrical considerations and were added to the structure factor calculation.The SQUEEZE procedure 44 to remove the solvent molecules which ) Å.The Fe 3 chain is of signicant interest in its own rightthere are few examples of metal-metal bonded units encapsulated inside Zintl clusters, and the short Fe-Fe distances are a clear a priori indication of strong bonding.Of equal interest is the structure of the Sn 18 cage because the degree of fusion of the two Sn 9 polyhedra appears to be mid-way between the completely separated limited, as observed in [Ni 2 -CdSn 18 ] 6− , and the completely fused limit in [Pd 2 Sn 18 ] 4 .[Fe 3 Sn 18 ] 4− is, therefore, the Sn analogue of the Ge 18 unit in [Ni 3 Ge 18 ] 4− .Our analysis of the electronic structure indicates that the E 18 cages in [Ni 3 Ge 18 ] 4− and [Fe 3 Sn 18 ] 4− share a common −6 charge state, as does the Sn 18 unit in [AgSn 18 ] 5− , 20 where the two Sn 9 units are not fused but rather oligomerised via an exo bond (Fig. 1(b)).A comparison of the different structural chemistry of these isoelectronic species offers a fascinating insight into the factors that control the balance between fusion in [Fe 3 Sn 18 ] 4− and [Ni 3 Ge 18 ] 4− and oligomerisation in [AgSn 18 ] 5− .

3 + 3 +
Returning to the comparison with the coordination complexes [Fe(DpyF) 4 ] 2+ (ref.34) and Fe 3 L 3 , 32 identied in Fig. 3, the overall oxidation state of the Fe 3 unit is lower in [Fe 3 Sn 18 ] 4− (Fe 2 vs. Fe 6 ) and the Fe-Fe s* orbital is unoccupied, both of which contribute to the lower multiplicity (S = 3 vs.S = 6) and stronger Fe-Fe bonding in the cluster compared to the coordination complexes.

Fig. 4
Fig. 4 (a) Kohn-Sham orbitals for [Fe 3 Sn 18 ] 4− in its 7 A 2g ground state.Levels shown in green are the primarily Fe-based orbitals while the remainder, in grey, have dominant Sn character.Orbitals with similar spatial characteristics are joined by a dashed line.

Fig. 5
Fig. 5 Walsh diagram showing the coalescence of the two Sn 9 polyhedra to a single ellipsoidal Sn 18 unit.The figure is generated by interpolating between the optimised structures of the Sn 18 unit as it is found in the optimised geometries of [Ni 2 CdSn 18 ] 6− , [Fe 3 Sn 18 ] 4− and [Pd 2 Sn 18 ] 4− .

Table 1
Selected bond lengths from crystallographic and DFT-optimised structures for the [M 3 E 18 ] 4− family (all distances in Å).
Sn 18 ] 4− (and also in [Ni 3 Ge 18 ] 4− ) is connected to the presence of the underlying Fe-Fe-Fe or Ni-Ni-Ni bonded framework, which provides a rigid 'strut' that resists the bending at the central metal necessary to form a localised exo Sn-Sn bond.Where metal-metal bonding is absent, as it necessarily is in [AgSn 18 ] 5− , bending to form a localised 2center-2-electron bond is the preferred outcome: a series of DFT calculations on different isomers of [AgSn 18 ] 5− conrms a 0.2 eV preference for the bent structure shown in Fig. 1 over the alternative D 3d -symmetric [Fe 3 Sn 18 ] 4− -like alternative.Taking the argument a step further, a second 2-electron oxidation step could, in principle, generate clusters with two exo bonds linking the Sn 9 units as an alternative to forming the coalesced cage typical of [Pd 2 Sn 18 ] 4−doubly bonded E 9 units of this kind have been identied in the Ge 27 p) per Fe-Fe bond.The cluster is structurally similar to the [Ni 3 Ge 18 ] 4− anion reported previously by Sevov and co-workers, although the Ni-Ni bonding in that case lacks the p component.The structure of the Sn 18 unit in [Fe 3 Sn 18 ] 4− is intermediate between that in [Ni 2 CdSn 18 ] 6− , where the two Sn 18 units are almost completely separated, and [Pd 2 Sn 18 ] 4− , where they are completely fused to form a continuous ellipsoidal Sn 18 unit.