Self-assembly of an Organometallic Fe₉O₆ Cluster from Aerobic Oxidation of (tmeda)Fe(CH₂^tBu)₂

Abstract

Aerobic oxidation of (tmeda)Fe(CH₂^tBu)₂ in toluene or THF solution leads to the self-assembly of a magic-sized all-ferrous oxide cluster containing the Fe₉O₆ subunit and bearing organometallic and diamine ligands. Mössbauer studies of the cluster are consistent with an all-ferrous assignment and magnetometry reveals complex intracluster and intercluster magnetic interactions.

Graphical Abstract

Molecular clusters are profoundly important species, playing fundamental roles in materials science, nanocatalysis and as models of heterogeneous surfaces and protein active sites.¹ Since aggregation steps are generally unselective and fast, synthetic routes towards large atomic assemblies that result in monodisperse samples are rare. Originally limited to time-controlled hydrothermal methods, recent developments have enabled the syntheses of atomically-precise ligand-stabilized clusters of transition metals and semiconductors,^{2,3,12,4–11} wherein conditions are controlled to selectively aggregate monomers and minimize bulk material growth. As a result of these studies of mesoscale entities, the change in physical properties of materials as they progress from the molecular scale to the bulk can be studied in great detail. Since the development of nanoengineered materials (e.g. metal-organic frameworks) depends heavily on the available design elements,¹³ modular clusters with a variety of chemical and physical properties are highly desirable.

Metal oxide clusters are of particular interest due to the electrical and magnetic properties of the bulk compounds. Transition metal oxide clusters have nearly always been studied as gas-phase species generated by laser ablation of metals in an oxygen atmosphere.^14–17 Synthetic routes towards chemically isolable atomically-precise clusters of these highly ionic materials are strongly hampered by the tendency for oxides to aggregate in the absence of polydentate ligand frameworks,^18,19 and syntheses are largely limited to hydrothermal conditions that result in carboxylate- or alkoxide-capped clusters and have limited opportunities for further elaboration into nanomaterials.^20–23 Moreover, since syntheses of hybrid MOF-like materials are often also hydrothermal,¹³ the incorporation of oxide-based clusters can be complicated by annealing (changing the structure of the oxide subunit) or agglomeration. The present work reports the synthesis and unique magnetic properties of an atomically-precise ferrous oxide cluster that bears highly reactive organometallic ligands, and represents the first synthetic route towards a member of the ‘magic-sized’ class of metal oxide clusters.^15,16

While researching unrelated chemistry of first-row transition metal dialkyl compounds, we observed that exposure of organoiron compounds to sub-stoichiometric oxygen led to the formation of a hydrocarbon-soluble complex that crystallized on standing (Figure 1). Treatment of colorless (tmeda)Fe(CH₂^tBu)₂ with 0.5 molar equivalents of O₂ (from air) in toluene solution results in the formation of a red-brown solution/suspension. After stirring at room temperature for 4 hours and removal of solvent in vacuo, the brown/black residue was extracted with hydrocarbon solvents (e.g. pentane, hexane, heptane, etc.) to yield a brown solution from which deep red crystals of the nine-iron cluster [Fe₉O₆]•solvent formed (17 % isolated yield based on [Fe] with pentane solvate).

Figure 1.

Synthesis (top), solid state structure of [Fe₉O₆] (bottom left) and views of the trigonal [(Np)FeO₂] and pseudotetrahedral [(tmeda)FeO₂] (bottom right) subunits. Selected distances and angles: d(Fe1-O1) = 1.907(2) Å, d(Fe1-O1’) = 1.903(2) Å, d(Fe1-C1) = 2.060(3) Å, d(Fe2-O1) = 1.872(2) Å, d(Fe2-N1) = 2.232(2) Å, d(Fe1-Fe1’) = 3.052(1) Å, d(Fe1-Fe2) = 3.249(1) Å, ∠(O1-Fe1-O1’) = 110.08(9)º, ∠(O1-Fe2-O1’) = 130.38(9)º.

Although the yield is relatively low, it is consistent for reaction scales up to ~ 1g of (tmeda)Fe(CH₂^tBu)₂ and is reasonable for a one-pot reaction that does not require ligands of additional complexity. The product is readily separated from the mother liquor by decantation; crystals of [Fe₉O₆] are attracted to an external permanent magnet with sufficient strength that this can be used to aid physical separation. NMR spectral analysis of the mother liquor reveals the presence of peaks consistent with neopentanol, pivalaldehyde, and residual (tmeda)Fe(CH₂^tBu)₂ which can be re-isolated via crystallization (~ 10 – 15% of the starting [Fe]). The relative amounts of oxidized organic products vary considerably from batch to batch and did not seem correlated with yields.

The cluster crystallizes in the chiral rhombohedral space group R3₂, resulting in crystallographic D₃ symmetry, and possesses two distinct iron sites in either an axial or equatorial disposition (Figure 1, bottom). The coordination environment of the axial iron sites is best described as trigonal planar (Σ(angles) = 359.93º), whereas the equatorial iron sites are best described as distorted tetrahedral (τ₄ = 0.81).²⁴ The cluster is highly air- and moisture-sensitive, but is stable for days under ambient conditions in solution (toluene, benzene, THF) and indefinitely stable in the solid state. The cluster crystallizes on both the three-fold and perpendicular two-fold rotation axes of the crystal lattice, such that one sixth of the molecule is the asymmetric unit, and clusters pack such that they are mutually aligned on the z-axis (Figure 2). There are large voids between molecules (~ 50 ų) that contain disordered solvent molecules (Figure 2, left); isostructural crystals have been grown that contain C_nH_2n+2 hydrocarbons (n = 5 – 9) and toluene, suggesting that crystal growth is somewhat insensitive to the nature of the guest.

Figure 2.

The solid-state packing of [Fe₉O₆] along the c-axis (right) highlight the axial alignment and three-fold symmetric packing of [Fe₉O₆] molecules. A depiction of the solid-state packing of [Fe₉O₆] highlighting the voids in the crystal lattice occupied by disordered solvent (left, or clarity, the Fe₉O₆ core is shown in yellow and the remaining ligands are shown in grey).

Redox-active metal centers in cluster compounds, particularly those with covalent bridging ligands (i.e. ‘soft’ anions), are notoriously challenging to assign to oxidation states.^25,26 In the present case, standard charges on ligands would predict that the average oxidation state of the nine iron atoms would be +2 (i.e. all ferrous). Zero-field Mössbauer spectroscopy at 5.5 K reveals two quadrupole doublets in a 2:1 ratio (labeled δ₁ and δ₂, respectively), as expected for the two distinct iron environments of [Fe₉O₆] (Figure 3, top). The major doublet (δ₁ = 0.41(2) mm/s, ΔE_Q = 1.57(2) mm/s), consistent with the trigonal planar environment on the axial sites, has spectral parameters similar to the high-spin trigonal-planar ferrous complex (nacnac)FeMe reported by Holland and coworkers (δ = 0.48 mm/s, ΔE_Q = 1.74 mm/s).²⁷ The minor doublet (δ₂ = 0.92(4) mm/s, ΔE_Q = 1.18(4) mm/s) should thus correspond to the pseudotetrahedral environment on the equatorial sites, although the relative paucity of Mössbauer spectra available for such an environment required the synthesis of a model complex. Treatment of (tmeda)Fe(CH₂^tBu)₂ with two equivalents of triphenylmethanol yields the high-spin ferrous alkoxide (tmeda)Fe(OCPh₃)₂, which exhibits a Mössbauer spectrum (Figure 3, bottom) that is indeed consistent with the minor component in the spectrum of [Fe₉O₆] (δ = 0.97(2) mm/s, ΔE_Q = 1.12(2) mm/s). These data taken together support an all ferrous assignment.

Figure 3.

5.5 K-Mössbauer spectra of crystalline [Fe₉O₆]•heptane (top, 0 T) and (tmeda)Fe(OCPh₃)₂ (bottom, 70 mT). Data are shown in black hash-marks; the red lines are spectral fits with the parameters discussed in the text. Linewidths (FWHM) are 0.33 mm/s for [Fe₉O₆] and 0.30 mm/s for (tmeda)Fe(OCPh₃)₂. The maroon arrow in the bottom panel indicates a broad, unidentified impurity accounting for less than 2 % of the total spectral area.

The low coordination geometry and weak ligand fields in [Fe₉O₆] should lead to high-spin configurations for all ferrous ions, but the magnetic behavior of the cluster is not as easy to predict. Coupling between magnetic ions mediated by a diamagnetic bridge would generally favor antiferromagnetic coupling,^28,29 but the odd number of iron atoms would lead to a net magnetic moment.³⁰ Accordingly, Evans method measurements (Figure 4, left) reveal that solutions of [Fe₉O₆] at 298 K exhibit an effective magnetic moment that is lower than that expected for nine non-interacting high-spin Fe(II) centers (7.9 instead of 14.7), but considerably higher than expected for nine antiferromagnetically-coupled high-spin Fe(II) centers (net S = 2, expected μ_eff value ≅ 4.9).

Figure 4.

Left: Effective magnetic moment of [Fe₉O₆] in the solid state (SQUID, 1 T, purple) and in THF-d₈ solution (400 MHz ¹H NMR, maroon). Right: variable-temperature magnetometry of crystalline [Fe₉O₆]•heptane (SQUID, left) at 1T.

The solution-phase magnetic moment is weakly temperature dependent over the range from 200 K to 298 K, suggesting the complex has an S = 2 ground state with low-lying excited states that are thermally-accessible. Electronic structure calculations of a truncated model of the cluster (Fe₉O₆(H)₆(NH₃)₆) using the broken-symmetry formalism^31,32 are consistent with strong antiferromagnetic coupling leading to an S = 2 ground state (5 α Fe(II) + 4 β Fe(II)), with an energetically-accessible S = 6 excited state (6 α Fe(II) + 3 β Fe(II), +6.8 kcal/mol, expected μ_eff value = 13.0). The two electronic configurations differ in the relative spin orientation of one of the axial Fe atoms, wherein the first excited state has all the axial Fe environments antiferromagnetically coupled to the equatorial environments (Figure S4).

Magnetic susceptibility measurements performed on solid samples of [Fe₉O₆]•heptane reveal a higher effective magnetic moment value than found in solution (9.5 vs. 7.9 at 298 K). Upon decreasing the temperature, the effective magnetic moment value decreases monotonically, eventually reaching 0.8 at 2 K. This very low value implies that complexes are coupling with each other to lead to an overall singlet ground state for the compound. The temperature dependence of the molar magnetic susceptibility (χ_para) reveals Curie-Weiss behavior from 300 K to ~ 60 K (θ_P = –186 K at 1.0 T), whereupon the susceptibility levels out and then begins to decline below 20 K (Figure 4, right). The leveling-out behavior appears to be due to two discrete magnetic ordering events, occurring at 60 K and 20 K, which is unusual for molecular materials exhibiting simple antiferromagnetic coupling. Crystals of [Fe₉O₆] are axially symmetric (i.e. x = y ≠ z) (Figure 2) and interactions between clusters in the solid state may lead to complex magnetic behavior in polycrystalline samples, including multiple ordering events to relieve frustration. Using the higher ordering temperature (T_o) of 60 K, the degree of frustration (as f = –θ_P/T_o)³³ of crystalline [Fe₉O₆]•heptane is calculated to be 3.10 at 1.0 T, indicating that this system is moderately frustrated. Magnetometry data could be fit using a four-parameter model, but the limited data results in an underdetermined mathematical system with several possible solutions (details in ESI); more detailed studies are underway.

Although the mechanism of formation of [Fe₉O₆] is difficult to elucidate, decades of research into the autoxidation of organometallic compounds, generally studied within the context of catalytic substrate oxidation, reveals that such reactions almost always favor free-radical chains.³⁴ Even noble metal organometallic systems, which usually terminate in alkoxide products from formal O-atom insertion into M-C bonds,^35–37 have been shown to proceed via radical steps initiated by binding of O₂ and formation of O-centered radicals. First row organometallics tend to yield O-atom insertion products,^38,39 and there is evidence that these proceed by way of the formation of metal dioxo intermediates.^40,41 The neopentanol and pivalaldehyde observed could be the result of rapid oxidation of neopentyl radicals produced upon coordination of O₂ to (tmeda)Fe(CH₂^tBu)₂.^42–44 Oxidations under high concentrations of (tmeda)Fe(CH₂^tBu)₂ seem essential, as the yield of [Fe₉O₆] was maximized only when the amount of residual (tmeda)Fe(CH₂^tBu)₂ remained high (10 – 15%) – that is, reactions run to complete consumption of the dialkyl precursor actually resulted in lower yields of [Fe₉O₆] (< 5%). It is also notable that oxidation of (tmeda)Fe(CH₂^tBu)₂ produces isolable amounts [Fe₉O₆] in either toluene or THF, but performing the same oxidation in pentane yields no isolable organometallic products. Given the weak C–H bonds in both toluene and THF, it seems plausible that these solvents are intercepting transient LFe^III-(O₂^–) species via formal H-atom transfer and minimizing the rate of aggregation and precipitation bulk iron oxide (Scheme 1).

Scheme 1.

One of the many possible oxidation cascade reactions leading to Fe-O bond formation during self-assembly of [Fe₉O₆], where X indicates either neopentyl or alkoxide anionic ligands.

Among the most notable aspects of this reaction is that since oxidation of Fe(II) precursors yields only Fe(II) oxide materials, it is conceivable that the iron-carbon bonds that are supplying the reducing equivalents. Low-coordinate organometallic complexes have considerable ionic character and have been used as formal sources of reduced metals,⁴⁵ and it seems possible that isolation of organometallic oxide clusters under substoichiometric oxidation conditions may be generalizable. Given the relatively low yield, the production of [Fe₉O₆] may be more a result of adventitious crystallization from a mixture of related clusters then the selective formation of just that core, but it is notable that the core bears the same Fe_nO_m stoichiometry as one of the ‘magic-sized’ clusters first observed by Sun et al. in gas phase experiments.^15,16 Such patterns are observed in noble metal and semiconductor cluster chemistry,^4,12 wherein part of the ability to isolate monodisperse clusters relies on magic numbers of elements that have a particular stability as monomers aggregate. Despite the presence of ligands and different oxidation state, the topology of the metal oxide core in [Fe₉O₆] is the same as predicted for the unligated cluster.

The self-assembly of a complex, reduced, low-coordinate organometallic cluster provides unique opportunities in synthetic chemistry. Gentle heating (65 ºC) of [Fe₉O₆] over the course of several hours in benzene, for example, results in the formation of dark precipitate and the formation of neopentane, suggesting [Fe₉O₆] is capable of C–H activation in a similar manner as related low-coordinate iron complexes.⁴⁶