DOI:
10.1039/C9SC03431E
(Edge Article)
Chem. Sci., 2019, Advance Article
Electronic structures of bent lanthanide(III) complexes with two Ndonor ligands†
Received
11th July 2019
, Accepted 17th September 2019
First published on 18th September 2019
Low coordinate metal complexes can exhibit superlative physicochemical properties, but this chemistry is challenging for the lanthanides (Ln) due to their tendency to maximize electrostatic contacts in predominantly ionic bonding regimes. Although a handful of Ln^{2+} complexes with only two monodentate ligands have been isolated, examples in the most common +3 oxidation state have remained elusive due to the greater electrostatic forces of Ln^{3+} ions. Here, we report bent Ln^{3+} complexes with two bis(silyl)amide ligands; in the solid state the Yb^{3+} analogue exhibits a crystal field similar to its three coordinate precursor rather than that expected for an axial system. This unanticipated finding is in opposition to the predicted electronic structure for twocoordinate systems, indicating that geometries can be more important than the Ln ion identity for dictating the magnetic ground states of low coordinate complexes; this is crucial transferable information for the construction of systems with enhanced magnetic properties.
Introduction
The remarkable optical, magnetic and catalytic properties of the lanthanides (Ln) have provided numerous technological applications,^{1} and design criteria now exist to build complexes with precise geometrical features that maximize these attributes.^{2–10} Highly axial Ln^{3+} complexes have recently become desirable targets for the singlemolecule magnet (SMM) community as such geometries can provide maximum anisotropy for several Ln^{3+} ions;^{2–5,11–13} indeed, we have previously predicted that a hypothetical nearlinear Dy^{3+} cation [Dy{N(Si^{i}Pr_{3})_{2}}_{2}]^{+} could exhibit a record energy barrier to the reversal of magnetization, providing the inspiration for this work.^{14} Some of us^{15–18} and others^{19,20} have recently shown that isolated axial Ln^{3+} metallocenium cations [Ln(Cp^{R})_{2}]^{+} (Cp^{R} = substituted cyclopentadienyl) can be prepared by halide abstraction from [Ln(Cp^{R})_{2}(X)] precursors by using the silylium reagent [H(SiEt_{3})_{2}][B(C_{6}F_{5})_{4}].^{21} The axial [Dy(Cp^{R})_{2}]^{+} members of this family^{15,19,20} together with a linear Tb^{2+} metallocene^{22} exhibit the current highest blocking temperatures for SMMs.
The isolation of low coordinate Ln complexes is often synthetically challenging, as the predominantly ionic bonding regimes in these systems favour high coordination numbers to maximize the number of electrostatic interactions between ligand donor atoms and relatively large Ln cations.^{8} Seminal work by Bradley in the early 1970s provided the trigonal pyramidal Ln complexes, [Ln{N(SiMe_{3})_{2}}_{3}], which exhibit additional Ln⋯Cγ–Siβ interactions that stabilize the coordinatively unsaturated Ln^{3+} centres.^{23,24} In the interim, numerous trigonal pyramidal and planar Ln^{3+} and Ln^{2+} complexes have been accessed by using a combination of sterically demanding ligands and strict anaerobic conditions.^{25,26} In contrast, there are only a handful of structurally characterised monomeric Ln^{2+} complexes with only two formally monodentate ligands; the majority contain intramolecular πarene contacts,^{27–31} whilst bent [Ln{C(SiMe_{3})_{3}}_{2}] (Ln = Sm, Eu, Yb)^{32–34} and nearlinear [Ln{N(Si^{i}Pr_{3})_{2}}_{2}] (1Ln; Ln = Sm, Eu, Tm, Yb)^{14,35} have additional electrostatic interactions between the ligand σbonding frameworks and Ln^{2+} centres. Ln^{3+} complexes with only two monodentate ligands have remained elusive to date as more Lewis acidic Ln^{3+} centres favour higher coordination numbers.^{1}
In 2018, some of us showed that 1Sm can be easily oxidized by a variety of reagents to afford heteroleptic Sm^{3+} halide complexes [Sm{N(Si^{i}Pr_{3})_{2}}_{2}(X)] (X = F, Cl, Br, I).^{36} Herein we report the synthesis of the bent Ln^{3+} complexes [Ln{N(Si^{i}Pr_{3})_{2}}_{2}][B(C_{6}F_{5})_{4}] (2Ln; Ln = Sm, Tm, Yb) by an analogous halide abstraction from [Ln{N(Si^{i}Pr_{3})_{2}}_{2}(X)], (3Ln; X = Cl, Ln = Sm,^{36} Tm; X = F, Ln = Yb) using [H(SiEt_{3})_{2}][B(C_{6}F_{5})_{4}]; 3Tm and 3Yb are prepared by the oxidation of [Ln{N(Si^{i}Pr_{3})_{2}}_{2}] (1Ln; Ln = Tm, Yb) with ^{t}BuCl and [FeCp_{2}][PF_{6}], respectively. We have probed the electronic structures of these exotic yet structurally simple complexes by magnetic and EPR methods, supported by ab initio calculations. This allows us to probe the effect of approximately linear, bent or planar geometries on the ligand field splitting. Simple electrostatic arguments^{5} based on aspherical electron density distributions in the Russell Saunders sublevels^{37} predict that 2Ln and 3Ln should have opposite senses of magnetic anisotropy for a given 4f^{n} configuration: we find that this is not always the case, and in fact can vary markedly with the degree of bending of the N–Ln–N angle.
Results and discussion
Synthesis
Oxidation of the Ln^{2+} complexes 1Ln with either ^{t}BuCl (Ln = Sm,^{36} Tm) or [FeCp_{2}][PF_{6}] (Yb) in toluene gave the heteroleptic Ln^{3+} complexes 3Ln in good yields (58–72%) following recrystallization from hexane (Scheme 1); similar oxidative procedures on Ln^{2+} bis(silyl)amide complexes have recently been applied by Anwander and coworkers.^{38} The Eu^{3+} analogue 3Eu could not be accessed by analogous methods, with crystals of 1Eu the only isolable product from numerous attempts to oxidize 1Eu with ^{t}BuCl, [FeCp_{2}][PF_{6}] and Ph_{3}CCl. This can be attributed to the preference of Eu to exhibit the +2 oxidation state over all other Ln, as illustrated by standard reduction potentials, E^{θ}, Ln^{3+} → Ln^{2+}: −0.35 V (Eu), −1.15 V (Yb), −1.55 V (Sm), −2.3 V (Tm).^{39} Halide abstraction of 3Ln using [H(SiEt_{3})_{2}][B(C_{6}F_{5})_{4}] in benzene (Sm, Tm) or toluene (Yb) yielded the bent Ln^{3+} complexes, 2Ln, in moderate yields (46–70%) after recrystallization from DCM layered with hexane (Scheme 1). The silylium reagent was selected for its solubility in noncoordinating solvents and for the provision of a large thermodynamic driving force for the reaction.^{40}

 Scheme 1 Synthesis of 2Ln and 3Ln. See ref. 36 for the synthesis of 3Sm.  
NMR spectroscopy
The paramagnetic Ln^{3+} centres in 2Ln and 3Ln engender large pseudocontact shifts and significant signal broadening in NMR spectra;^{41,42} the spectra that exhibited signals are compiled in ESI Fig. S4–S13.†^{1}H NMR spectra were recorded from +200 to −200 ppm and for 2Sm peaks were observed at 0.43 ppm and −5.27 ppm, corresponding to the methyl and methine protons, respectively, of the bis(silyl)amide ligand. For both 2Tm and 2Yb only one broad peak was observed at 25.04 ppm and 11.02 ppm, respectively, which we tentatively assign to the methyl protons as these are more numerous than methine protons. No signals were observed for 2Ln by ^{29}Si{^{1}H} and ^{13}C{^{1}H} NMR spectroscopy. Similarly, no signals were observed for the [B(C_{6}F_{5})_{4}]^{−} anion in the ^{13}C{^{1}H} NMR spectra of 2Ln; however for 2Sm, 2Tm and 2Yb, the ^{11}B{^{1}H} NMR spectra displayed sharp peaks at −16.76, −12.35 and −14.67 ppm, respectively. The ^{19}F{^{1}H} NMR spectra of 2Sm and 2Yb each displayed three signals characteristic of the [B(C_{6}F_{5})_{4}]^{−} anion (−133.17, −163.71 and −167.60 ppm for 2Sm and −131.58, −162.00 and −165.15 ppm for 2Yb), but only one signal was observed in the ^{19}F{^{1}H} NMR spectrum of 2Tm (−128.51 ppm). No signals corresponding to 3Ln could be seen in the ^{1}H or ^{13}C{^{1}H} NMR spectra for all 3Ln, with only diamagnetic impurities observed; no features were seen in the ^{19}F NMR spectrum of 3Yb. Given the paucity of information that could be extracted by NMR spectroscopy for 2Ln and 3Ln, we did not conduct variable temperature studies as these did not prove fruitful for 1Ln previously;^{35} instead we have analysed metal–ligand interactions by computational methods (see below).
Single crystal XRD
The solid state structures of 2Ln and 3Ln were determined by single crystal Xray diffraction. Complexes 2Tm and 3Tm are depicted in Fig. 1 and selected metrical parameters are compiled in Table 1; see ESI Fig. S1–S3† and ref. 35 for the structures of other complexes. Complexes 2Ln are structurally analogous, though 2Sm and 2Yb both adopt the P2_{1}/n space group and 2Tm crystallizes in P, and one molecule of DCM was present in the crystal lattice for both 2Tm and 2Yb, but is absent in crystals of 2Sm. The [Ln{N(Si^{i}Pr_{3})_{2}}_{2}]^{+} cations in 2Ln exhibit bent geometries defined by the two Ln–N bonds, with N–Ln–N angles of 131.02(8)° for 2Sm, 125.49(9)° for 2Tm, and 127.7(2)° for 2Yb, which are in contrast to the nearlinear geometries seen for 1Ln (range 166.01(14)–175.5(2)°).^{14,35} We attribute the bent geometries of 2Ln to the Ln^{3+} cations being more Lewis acidic than the Ln^{2+} centres in 1Ln,^{14,35} as this permits the more electron deficient Ln^{3+} centres to form additional stabilizing electrostatic contacts with methyl and methine groups of the {N(Si^{i}Pr_{3})_{2}} ligands. A permanent dipole is formed between the two formally anionic N^{−} centres and Ln^{3+} ion upon bending; such dipolar stabilization mechanisms have previously been used to explain the pyramidal geometries of some fblock trissilylamides.^{43} Crystal packing forces and interligand dispersion forces also likely make important contributions.^{44} This subtle interplay of forces is particularly apparent for 2Yb (see below).

 Fig. 1 Molecular structures of (a) 2Tm and (b) 3Tm at 100 K with selected atom labelling. Displacement ellipsoids set at 50% probability level, solvent of crystallization and hydrogen atoms are omitted for clarity. Key: thulium, teal; silicon, orange; nitrogen, blue; fluorine, green; boron, yellow; carbon, grey.  
Table 1 Selected bond distances and angles of Ln{N(Si^{i}Pr_{3})_{2}}_{2} moieties in 2Ln and 3Ln
Complex 
Ln–N/Å 
N–Ln–N/° 
Ln–X 
2Sm

2.257(3), 2.228(3) 
131.02(8) 
— 
2Tm

2.156(2), 2.156(2) 
125.49(9) 
— 
2Yb

2.152(4), 2.144(5) 
127.7(2) 
— 
3Sm
^{36}

2.295(2), 2.317(2) 
128.24(7) 
2.5813(7) 
3Tm

2.219(2), 2.238(2) 
129.39(5) 
2.4832(5) 
3Yb

2.226(3), 2.235(3) 
138.71(9) 
1.983(2) 
As with the 1Ln series,^{14,35} the heavier Ln^{3+} centres in 2Ln exhibit more bent N–Ln–N angles, which we again ascribe to the greater charge density of smaller Ln^{3+} cations driving stronger electrostatic interactions with ligand C–H bonds. The {N(Si^{i}Pr_{3})_{2}} ligands in 2Ln are staggered with respect to each other, with the mean Ln–N bond lengths decreasing with Ln^{3+} atomic radii: 2.243(4) Å (Sm), 2.156(3) Å (Tm) and 2.148(6) Å (Yb). It may appear counterintuitive that the Ln–N bonds in 2Ln are shorter than those in 1Ln (2.483(6) Å, Sm; 2.373(2) Å, Tm; and 2.384(3) Å, Yb)^{14,35} given the decreased N–Ln–N angles in 2Ln compared with 1Ln, but shorter Ln–N bonds for 2Ln are expected from an increase in Ln oxidation state. Three Si–C bonds are oriented towards the Ln^{3+} centre in each [Ln{N(Si^{i}Pr_{3})_{2}}_{2}]^{+} cation; these are assigned as Ln⋯Cγ–Siβ agostictype interactions by analogy with those discussed for threecoordinate silylsubstituted Ln complexes.^{45–48} These interactions lead to three relatively long βSi–C bonds, three short Ln⋯Si distances, six Ln⋯C and six Ln⋯H electrostatic contacts with methyl/methine groups [e.g. for 2Tm: range Tm⋯C: 2.731(3)–3.051(3) Å; range Tm⋯H: 2.200–2.495 Å; range Tm⋯Si: 3.066(2)–3.178(2) Å; mean βSi–C: 1.938(3) Å; range other Si–C: 1.889(3)–1.917(3) Å]. The [B(C_{6}F_{5})_{4}]^{−} anions do not coordinate; the shortest Ln⋯F distance for 2Yb is 4.627(4) Å, whereas for 2Sm and 2Tm the shortest Ln⋯F distances are longer at 7.957(2) Å and 7.715(2) Å, respectively. Using the IUPAC definition of coordination number as the number of metal–ligand σbonds,^{49} the cations of 2Ln can be considered to be formally twocoordinate as they each exhibit two Ln–N bonds; we probed the numerous additional Ln⋯Cγ–Siβ electrostatic interactions further through calculations as these could affect the magnetic properties of the proposed [Dy{N(Si^{i}Pr_{3})_{2}}_{2}]^{+} cation (see below).^{12,14}
The structure of 3Sm has previously been reported,^{36} but will be discussed together with 3Tm and 3Yb as all three complexes are structurally similar. Complex 3Yb crystallizes in P, whilst 3Sm and 3Tm are in the P2_{1}/c space group. Complexes 3Ln all crystallize with distorted trigonal planar geometries, with the Ln^{3+} centres positioned out of the plane defined by the two nitrogen atoms and halide (distances of Ln from N_{2}(X) plane: 0.245(2) Å for 3Sm, 0.3292(9) Å for 3Tm and 0.312(2) Å for 3Yb). As expected the Yb–F bond length of 3Yb [1.983(2) Å] is shorter than the Ln–Cl bond lengths of 3Sm (2.5813(7) Å) and 3Tm (2.4832(5) Å) due to the smaller size of the fluoride anion; this also leads to differing N–Ln–N angles (3Sm: 128.24(7)°; 3Tm: 129.39(5)°; 3Yb: 138.71(9)°). The mean Sm–N bond length of 3Sm (2.306(3) Å) is significantly longer than the mean Ln–N bond lengths of 3Tm (2.229(3) Å) and 3Yb (2.231(4) Å), which corresponds with earlier Ln^{3+} ions being larger.^{1} The Ln–N bond lengths in 3Ln are longer than those in 2Ln, as expected from increasing the formal coordination number from two to three. Finally, as with 2Ln the coordination spheres of the Ln^{3+} centres of 3Ln are completed by multiple electrostatic contacts with methine and methyl groups. These are also likely to arise from Ln⋯Cγ–Siβ agostictype interactions, though in 3Ln there are fewer, and the Tm⋯C/H/Si distances are generally longer due to the presence of a halide [e.g. for 3Tm: range three Tm⋯C: 2.874(2)–3.261(2) Å; range three Tm⋯H: 2.324–2.467 Å; range three Tm⋯Si: 3.195(2)–3.354(2) Å; mean three βSi–C: 1.928(3) Å; range other Si–C: 1.899(2)–1.913(2) Å].
UVvisNIR spectroscopy
Dilute solutions of 2Sm, 2Tm and 2Yb in DCM are pale red, green and purple, respectively, and their electronic absorption spectra are dominated by strong ligand to metal charge transfer bands tailing in from the UV region (Fig. 2 and ESI Fig. S19–S21†). Complex 2Sm (4f^{5}) exhibits the most intense absorption in the visible region [λ_{max} = 411 nm (24300 cm^{−1}), ε = 511 M^{−1} cm^{−1}], whilst 2Tm and 2Yb exhibit weaker visible absorptions [2Tm; λ_{max} = 373 nm (26800 cm^{−1}), ε = 275 M^{−1} cm^{−1}; 2Yb: λ_{max} = 425 nm (23500 cm^{−1}), ε = 309 M^{−1} cm^{−1}, λ_{max} = 563 nm (17800 cm^{−1}), ε = 249 M^{−1} cm^{−1}]. Weak absorptions (ε < 100 mol^{−1} dm^{3} cm^{−1}) were seen for all 2Ln in the nearIR region, corresponding to Laporteforbidden f–f transitions:^{1}2Sm shows absorptions at λ_{max} = 1370 nm (7300 cm^{−1}), ε = 14 M^{−1} cm^{−1} and 1285 nm (7782 cm^{−1}), ε = 13 M^{−1} cm^{−1}, which arise due to ^{6}H_{5/2} → ^{6}F_{J} transitions; 2Tm shows absorptions at λ_{max} = 1549 nm (6456 cm^{−1}), ε = 6 M^{−1} cm^{−1} and λ_{max} = 1383 nm (7230 cm^{−1}), ε = 15 M^{−1} cm^{−1} which arise due to ligand fieldsplit ^{3}H_{6} → ^{3}H_{4} transitions; 2Yb has a broad feature at λ_{max} = 1015 nm (9552 cm^{−1}), ε = 77 M^{−1} cm^{−1} and two weaker absorptions at λ_{max} = 904 nm (11061 cm^{−1}), ε = 26 M^{−1} cm^{−1} and λ_{max} = 844 nm (11840 cm^{−1}), ε = 27 M^{−1} cm^{−1} which correspond to ^{2}F_{7/2} → ^{2}F_{5/2} transitions, showing the ligand field splitting in the excited ^{2}F_{5/2} term. These absorptions are moderately strong for f–f transitions because they are all spinallowed (ε < 200 M^{−1} cm^{−1}).^{1} The spectral pattern of one intense absorption and two weaker absorptions of approximately equal intensity at higher energy for the ^{2}F_{7/2} → ^{2}F_{5/2} manifold is a common feature for Yb^{3+} complexes; Da Re et al.^{50} and Denning et al.^{51} have discussed these transitions in considerable detail previously.

 Fig. 2 Room temperature UVvisNIR spectra of 2Ln (1 mM in DCM) from 6200–35000 cm^{−1}.  
Solutions of 3Sm, 3Tm and 3Yb are pale yellow, green and red, respectively, and as with 2Ln their absorption spectra, are dominated by ligand to metal charge transfer bands tailing in from the UV region (Fig. 3, ESI Fig. S22 and S23† and ref. 36) [3Sm: λ_{max} = 376 nm (26595 cm^{−1}), ε = 713 M^{−1} cm^{−1}; 3Tm: λ_{max} = 327 nm (30581 cm^{−1}), ε = 378 M^{−1} cm^{−1}; 3Yb: λ_{max} = 418 nm (23923 cm^{−1}), ε = 250 M^{−1} cm^{−1}, λ_{max} = 326 nm (30674 cm^{−1}), ε = 99 M^{−1} cm^{−1}]. In the near IR region f–f absorptions are observed for all complexes; 3Sm exhibits three main peaks at 7246, 7710 and 8439 cm^{−1} due to ^{6}H_{5/2} → ^{6}F_{J} transitions, however there appear to be numerous weaker transitions. Complex 3Tm shows two main absorptions at λ_{max} = 1506 nm (6640 cm^{−1}), ε = 47 M^{−1} cm^{−1} and λ_{max} = 777 nm (12870 cm^{−1}), ε = 86 M^{−1} cm^{−1}, corresponding to ^{3}H_{6} → ^{3}H_{4} and ^{3}H_{6} → ^{3}F_{4} transitions, however again these are structured due to ligand field splitting. Complex 3Yb displays two absorptions at λ_{max} = 973 nm (10277 cm^{−1}), ε = 22 M^{−1} cm^{−1} and λ_{max} = 860 nm (11627 cm^{−1}), ε = 17 M^{−1} cm^{−1} corresponding to ligand fieldsplit ^{2}F_{7/2} → ^{2}F_{5/2} transitions. The f–f transitions are at higher energy for 3Ln, presumably due to stronger ligand fields; this is most clear for the Yb pair, where for 2Yb the lowest energy transition is at 9500 cm^{−1}, whilst this is seen at 10200 cm^{−1} for 3Yb.

 Fig. 3 Room temperature UVvisNIR spectra of 3Ln (1 mM in THF) from 6200–35000 cm^{−1}. For 3Sm, an empirical absorption correction of ε + 1.9 mol^{−1} dm^{3} cm^{−1} has been applied.  
Magnetism and EPR spectroscopy
Linear and trigonalplanar environments should stabilize oblate and prolatespheroid electron density distributions, respectively, along the axis of quantization.^{2–5} This should then stabilize either the minimum or maximum m_{J} sublevels of the ^{2S+1}L_{J} Russell Saunders ground term depending on the 4f^{n} configuration.^{2–5} The ions studied here are 4f^{5} (Sm^{3+}), 4f^{12} (Tm^{3+}) and 4f^{13} (Tm^{2+}, Yb^{3+}) and in each case the electron density distribution in the maximum m_{J} states is prolate, hence an ideal linear geometry at Ln should give the minimum m_{J} = ±1/2 (Kramers) or 0 (nonKramers) ground sublevels, along with easyplane magnetic anisotropy. Correspondingly, ideal trigonalplanar geometry at Ln should give the maximum m_{J} = J ground levels and easyaxis magnetic anisotropy. These states can be probed by magnetometry and EPR spectroscopy. Room temperature solution phase magnetic moments (where χ is the molar magnetic susceptibility, T is the temperature) for 2Ln and 3Ln determined by the Evans method^{52} are in good agreement with those from solidstate SQUID magnetometry (Table 2 and ESI Fig. S24–S35†). We present the magnetic data for 2Ln and 3Ln pairs for each Ln^{3+} ion in turn.
Table 2 Room temperature χT values for 2Ln and 3Ln determined by Evans solution NMR method and solidstate SQUID magnetometry (1.0 T applied field for 2Sm and 3Sm; 0.1 T applied field for other compounds), with freeion values [g_{J}^{2}J(J + 1)/8], and values from CASSCF calculated electronic structures
χT/cm^{3} mol^{−1} K 
2Sm

2Tm

2Yb

3Sm

3Tm

3Yb

Theoretical value for ground spin orbit multiplet in the absence of a ligand field.

Freeion 
0.09^{a} 
7.15 
2.57 
0.09^{a} 
7.15 
2.57 
Evans 
0.43 
6.44 
2.13 
0.38 
6.31 
1.78 
SQUID 
0.23 
6.86 
1.98 
0.24 
6.31 
1.93 
CASSCF 
0.29 
6.88 
2.24 
0.29 
6.85 
2.24 
Complexes 2Yb and 3Yb have room temperature χT values of 1.98 and 1.93 cm^{3} mol^{−1} K, respectively (ESI Fig. S29 and S35†): these are lower than the freeion 4f^{13 2}F_{7/2} value due to substantial crystal field effects, as supported by CASSCFSO calculations which gives the total spread of the J = 7/2 term approaching 2000 cm^{−1} (ESI Table S3†). The same is true for the isoelectronic 4f^{13} Tm^{2+} analogue 1Tm.^{35} For 2Yb and 3YbχT decreases slowly on cooling, reaching 1.3 and 1.6 cm^{3} mol^{−1} K, respectively, at 2 K. At 2 K and 7 T, 2Yb and 3Yb reach saturation magnetizations of 1.80 and 1.84 μ_{B}, respectively, and the temperature dependence of the traces indicates isolated Kramers doublet ground states as expected (ESI Fig. S28 and S34†).^{35}
The similar properties of 2Yb and 3Yb were confirmed by lowtemperature EPR spectroscopy (Fig. 4 and Table 3): solid 2Yb has nearaxial gvalues of g_{1} = 6.80, g_{2} = 1.46 and g_{3} = 1.09, whilst solid 3Yb gives g_{1} = 7.11 with g_{2,3} not observed but ≪1. Approximating g_{1} = g_{‖} and g_{2,3} = g_{⊥}, this g_{‖} ≫ g_{⊥} pattern clearly demonstrates easyaxis magnetic anisotropy, consistent with a high m_{J} ground state doublet (the pure ±7/2 doublet would have g_{‖}, g_{⊥} = 8.0, 0). This is expected for trigonal planar 3Yb, but not for 2Yb which has only two Ndonors that we would expect to stabilize the low m_{J} doublet. Hence, for 2Yb it appears that the N–Yb–N angle has sufficiently deviated from linearity such that the crystal field is still quantized along the axis normal to the YbN_{2} plane despite the loss of the inplane fluoride from 3Yb. Clearly this result is very different from the easyplane isoelectronic nearlinear Tm^{2+} compound 1Tm (Fig. 4a). To further probe this finding, we examined the EPR spectra of the Yb^{3+} compounds in solution. EPR spectra of a frozen solution of 3Yb is very similar to the solid state, with g_{1} = 7.51 (g_{2,3} not observed), however, a frozen solution of 2Yb gives g_{1} = 4.38, g_{2} = 3.99 and g_{3} = 1.21 (Fig. 4), which unambiguously shows that there has been a switch to easyplane anisotropy (now approximate g_{1,2} = g_{⊥} and g_{3} = g_{‖}) as the g_{‖} ≪ g_{⊥} pattern indicates stabilization of a low m_{J} doublet (the pure ±1/2 doublet would have g_{‖}, g_{⊥} = 1.14, 4.17).^{53} Thus, the structure of 2Yb must relax in solution such that the N–Yb–N angle opens up and there is a flip of the orientation of the axis of quantization from being normal to the YbN_{2} plane to lying along the N⋯N direction. This is supported by CASSCFSO results based on the crystal structures: these give ground Kramers doublet g_{1} = 7.12, g_{2} = 1.14 and g_{3} = 0.55 for 2Yb, and g_{1} = 7.90, g_{2} = 0.10 and g_{3} = 0.07 for 3Yb (Table 3), with g_{1} (g_{‖}, defining the axis of quantization) oriented normal to the YbN_{2}(F) plane (Fig. 5). The ground doublet is 99% m_{J} = 7/2 in character for 3Yb, and slightly more mixed at 85% m_{J} = 7/2 for 2Yb due to the competing components of the crystal field (ESI Table S3†).

 Fig. 4 c.w. Xband EPR spectra. (a) 1Tm as a powder at 10 K;^{35} (b) 2Yb as a powder (in eicosane) at 10 K; (c) 2Yb in 1 mM DCM solution at 10 K (the feature at 320 mT is a background signal); (d) 3Yb as a powder (in eicosane) at 10 K; (e) 3Yb in 1 mM DCM solution at 10 K. Insert shows an expansion of the low field region of (d) and (e); these spectra are truncated as there are no features arising from 3Yb at higher fields. Experimental spectra are in black, simulations are in red.  
Table 3 Comparison of EPR data and metrical parameters for isoelectronic 1Tm, 2Yb and 3Yb
Complex 
N–Ln–N/° 
Calculated gvalues 
Measured gvalues 
Solid state 
Frozen solution 
g
_{1}

g
_{2}

g
_{3}

g
_{1}

g
_{2}

g
_{3}

g
_{1}

g
_{2}

g
_{3}

1Tm
^{35}

166.89(6) 
5.49 
3.60 
1.15 
5.71 
2.92 
1.01 
5.71 
2.92 
1.01 
2Yb

127.7(2) 
7.12 
1.14 
0.55 
6.80 
1.46 
1.09 
4.38 
3.99 
1.21 
3Yb

138.71(9) 
7.90 
0.10 
0.07 
7.11 
— 
— 
7.51 
— 
— 

 Fig. 5 Orientation of the main magnetic axis (red dashed line) for 2Yb (left) and 3Yb (right).  
Complex 2Tm has a χT value of 6.86 cm^{3} mol^{−1} K at 300 K, in good agreement with the freeion 4f^{12 3}H_{6} value. χT decreases rapidly with decreasing temperature due to depopulation effects within the multiplet, reaching ca. 0.8 cm^{3} K mol^{−1} at 2 K (ESI Fig. S27†). M(H) curves measured at 2 and 4 K are superimposable and fail to saturate (ESI Fig. S26†), suggesting a singlet nonmagnetic ground state for this nonKramers system. CASSCFSO calculations, performed on the two crystallographically nonequivalent molecules in the unit cell of 2Tm, confirm this, giving a singlet ground state which is separated from the first excited level by ca. 14.5 cm^{−1} (average for two independent molecules, ESI Table S3†). Magnetic data for 3Tm are markedly different: χT (6.31 cm^{3} mol^{−1} K at 300 K) only decreases slowly on cooling, reaching 5.48 cm^{3} mol^{−1} K at 2 K (ESI Fig. S33†), and M(H) at 2 and 4 K saturate at 3.3 μ_{B} above ca. 4 T (ESI Fig. S32†); this is direct evidence of a pseudodoublet magnetic ground state. Indeed, CASSCFSO calculations give a ground state pseudodoublet for 3Tm with an intradoublet gap of only 0.13 cm^{−1}. The pseudodoublet wave functions are mixtures of m_{J} = +6 and −6, which resolve into a pure m_{J} = +6 and m_{J} = −6 pair (98% purity) in a small applied magnetic field (ESI Table S5 and S6†). These results are supported by EPR spectroscopy of 2Tm and 3Tm in the solid state. We find that 2Tm is EPR silent at 5 K (ESI Fig. S37†), consistent with the magnetic data and as predicted by CASSCFSO, whilst 3Tm has a nearzerofield EPR transition at Xband (ca. 9.39 GHz; ESI Fig. S38†) indicating a zerofield splitting between the pseudodoublet states of ca. 0.3 cm^{−1}, in excellent agreement with magnetometry and CASSCFSO.
For 2Sm and 3Sm the room temperature χT products are 0.23 and 0.24 cm^{3} mol^{−1} K, respectively, higher than the freeion value for the 4f^{5 6}H_{5/2} multiplet (ESI Fig. S25 and S31).† This is indicative of lowlying, thermally accessible excited states as is commonly observed for Sm^{3+} (the ^{6}H_{7/2} term lies at only ca. 1000 cm^{−1}).^{54} On cooling, χT steadily decreases to 0.05 and 0.02 cm^{3} mol^{−1} K, respectively, at 2 K. For both 2Sm and 3Sm, the molar magnetization (M) at low temperatures fails to saturate as a function of applied magnetic field (H), reaching ca. 0.08 and 0.16 μ_{B}, respectively, at 2 K and 7 T (ESI Fig. S24 and S30†). In both cases, the traces for 2 and 4 K are distinct. These data are consistent with low magnetic moment Kramers doublet ground states. The ^{6}H_{5/2} ground term has a low Landé factor of g_{J} = 2/7, hence the effective gfactors for all the Kramers doublets are low. The extreme cases of pure m_{J} = 1/2 and 5/2 doublets would have g_{‖}, g_{⊥} = 0.29, 0.86 and 1.43, 0, respectively, and these would give rather similar and low magnetic moments for powders. Unfortunately, we were unable to obtain reliable EPR spectra for 2Sm or 3Sm. CASSCFSO calculations give a reasonable agreement with the experimental χT(T) and M(H) curves for both 2Sm and 3Sm (ESI Fig. S24, S25 and S31†) and indicate that the ground state gtensor for 2Sm is strongly rhombic, whereas in the case of 3Sm the main magnetic axis is perpendicular to the N_{2}(Cl) plane with strongly easyaxis character.
Comparing 2Yb with isoelectronic 1Tm, the N–Ln–N angle in 1Tm is much closer to linear at 166.89(6)° [cf. 127.7(2)° for 2Yb] and it has easyplane magnetic anisotropy as shown by EPR spectroscopy in both solid and frozen solution state with g_{1} = 5.6, g_{2} = 3.0 and g_{3} = 1.0.^{35} CASSCFSO calculations for the crystal structure of 1Tm give g_{1} = 5.49, g_{2} = 3.60 and g_{3} = 1.15, with g_{3} oriented along the N–Tm–N direction, resulting from a 99% pure m_{J} = 1/2 ground doublet.^{35} In order to test the importance of the identity of the metal ion vs. the N–Ln–N angle, we performed further CASSCFSO calculations on the structure of 1Tm [N–Ln–N 166.89(6)°] where we substitute Yb^{3+} in place of Tm^{2+}, and on the structure of 2Yb [N–Yb–N 127.7(2)°] where we substitute Tm^{2+} in place of Yb^{3+} (note the change in ion charge to maintain an f^{13} configuration in both cases). We find the former to have an m_{J} = 1/2 ground doublet (g_{1} = 5.34, g_{2} = 3.67, g_{3} = 1.16), and the latter to have an m_{J} = 7/2 ground doublet (g_{1} = 6.76, g_{2} = 1.97, g_{3} = 0.82): thus, it is the structure that dictates these differing properties for f^{13} configurations and it is not due to the identity of the metal ion. Nocton and coworkers have recently made similar observations for isoelectronic f^{13} Tm^{2+} and Yb^{3+} 18crown6 complexes.^{55} Whilst such reasoning is logical, it is not a phenomenon that has been observed frequently with realworld chemical systems.
Ab initio calculations
To clarify the dependence of the magnetic anisotropy on the N–Ln–N angle in 2Yb we have carried out a systematic ab initio investigation. CASSCFSO calculations have been performed on model structures based on the experimental structure of 2Yb in which the N–Ln–N angle has been varied between 180° and 110°. The calculated gvalues of the ground Kramers doublet of Yb^{3+} show a clear dependence of the type of magnetic anisotropy on the N–Ln–N angle, with the switching point located between 140° and 150° (Fig. 6): easyaxislike (g_{1} > g_{2,3}; g_{‖} > g_{⊥}) for N–Ln–N angles <140° and easyplanelike (g_{3} < g_{1,2}; g_{‖} < g_{⊥}) for angles> 150°. This implies that there must be a significant structural change in the N–Yb–N angle of 2Yb in the solution phase, becoming at least 150°. Optimization of the structure of 2Yb in the gas phase using densityfunctional theory (DFT) shows an increase in the N–Yb–N angle from 127 to 133° (ESI Table S13†). This indicates that the molecule tends to become more linear when removed from the solid state, suggesting that interactions with solvent molecules (absent in our gas phase calculations) stabilize larger N–Yb–N angles.

 Fig. 6 CASSCFSO calculated g_{‖} (black) and g_{⊥} (red) for the ground Kramers doublet of model structures based on 2Yb as a function of the N–Ln–N angle (lines). CASSCFSO values based on XRD experimental models (solid symbols) and experimental values (open symbols) for 2Yb (squares) and the isoelectronic 1Tm (circles). Given the rhombicity of the calculated gtensor we defined g_{‖} as the unique value that is either larger or smaller than the average of the three gvalues, while g_{⊥} is defined as the average value of the two remaining gvalues.  
We have conducted the same angulardependent study of the electronic structure of 2Tm as for 2Yb. The N–Tm–N angle has been varied between 180° and 120° (ESI Fig. S36†). Our results show that there is also a characteristic change in electronic structure for f^{12}2Tm: above 160° the singlet ground state is mainly a mixture of m_{J} = +1 and −1 functions, while below 150° the ground state is dominated by the m_{J} = 0 function (ESI Fig. S36†). The quantization axis in all cases is the direction that bisects the N–Tm–N angle; given the low symmetry of the complex and the fact that the molecule is neither linear nor trigonal this is not surprising. Therefore, the change in electronic structure from 2Tm to 3Tm appears to be in agreement with electron density arguments: the trigonalplanar coordination environment of 3Tm stabilizes a prolate ground pseudodoublet with maximum m_{J} where the quantization axis is normal to the trigonal plane, whilst the twocoordinate environment of 2Tm stabilizes an oblate singlet state dominated by m_{J} = 0; however in the latter case, far from being linear with a N–Tm–N angle of 125.49(9)°, the axis of quantization bisects the N–Tm–N angle and thus does not follow simple electron density arguments.
To examine the impact of the Ln⋯Cγ–Siβ agostictype interactions on the electronic structure of the Yb^{3+} ion in 2Yb, we have performed a CASSCFSO calculation on a model complex where the ligands have been trimmed to {N(SiH_{3})_{2}}, thus removing the Ln⋯C and Ln⋯H electrostatic contacts; these calculations reveal changes of <10% to the SO energy levels (ESI Table S10†) and a slight increase in axiality of the ground Kramers doublet (ESI Table S11†) compared to 2Yb. Although we cannot rule out changes to the Ndonor strength for the trimmed ligand versus {N(Si^{i}Pr_{3})_{2}}, these results suggest that the Ln⋯Cγ–Siβ agostictype interactions have only a slight influence on the electronic structure at the Yb(III) site and that they are far weaker than the Ln–N coordination bonds that dominate the electronic structure.
Finally, as this study was driven by our attempts to isolate a nearlinear twocoordinate Dy^{3+} complex, it is relevant to predict what the SMM properties of such a material could be now that we are far closer to a representative material with 2Tm and 2Yb, than in the previously reported nearlinear 1Sm.^{14} Hence, we have performed CASSCFSO calculations using the molecular geometry of 2Tm where Tm^{3+} has been replaced with Dy^{3+}. As predicted based on simple model compounds,^{12} even this bent geometry with a N–Dy–N angle of 125.49(9)° and equatorial agostic interactions can produce a very high barrier to magnetic relaxation of ca. 1300–1400 cm^{−1} (ESI Fig. S39 and Table S12†), and thus bent twocoordinate Dy^{3+} complexes of the type presented here are still exciting synthetic targets.
Conclusions
The preference for bent geometries in [Ln{N(Si^{i}Pr_{3})_{3}}_{2}]^{+} cations can be accredited to the formation of multiple electrostatic contacts between the highly Lewis acidic Ln^{3+} ions and the electron density associated with the ligand σbonding network, together with dipole stabilization, crystal packing forces and dispersion force interactions. By a combination of magnetic studies, EPR spectroscopy and ab initio calculations we have deduced the electronic structures of the bent Ln^{3+} cations. Interestingly, in the solid state [Yb{N(Si^{i}Pr_{3})_{3}}_{2}]^{+} expresses a similar crystal field to its threecoordinate precursor, rather than the axial crystal field that would be predicted for a twocoordinate complex. EPR spectroscopy shows that [Yb{N(Si^{i}Pr_{3})_{3}}_{2}]^{+} switches to an axial crystal field in solution, indicating that the N–Ln–N angle is less bent in the solution phase. The electronic structures of these bent Ln^{3+} cations are therefore sensitive to changes in molecular geometry.
Our synthetic results show that axial Dy^{3+} complexes such as [Dy{N(Si^{i}Pr_{3})_{2}}_{2}]^{+}, proposed as SMMs with large energy magnetization reversal barriers,^{14} are feasible chemical targets, whilst our electronic structure results show that the physical properties of target complexes for the SMM community are not trivially predictable. As a bent [Dy{N(Si^{i}Pr_{3})_{2}}_{2}]^{+} cation is predicted to show a lower effective barrier to magnetic reversal than a linear analogue, it would be of benefit to be able to predict what ligand systems would provide twocoordinate Dy^{3+} complexes that are less bent. Although the [Ln{N(Si^{i}Pr_{3})_{2}}_{2}]^{+} framework is of sufficient steric bulk, a linear geometry is not enforced as the coordination sphere is flexible enough to be rearranged to increase the strength of ligand–metal electrostatic and ligand–ligand London dipole interactions. Given that recently isolated linear Dy^{2+} and Tb^{2+} metallocene systems have been proposed to exhibit significant s–d mixing,^{22} it can be inferenced that combining electronic stabilization with similarly bulky but more rigid ligand frameworks may be a useful strategy in the future pursuit of linear twocoordinate Ln^{3+} complexes.
Experimental
Materials and methods
All manipulations were conducted under argon with the strict exclusion of oxygen and water by using Schlenk line and glove box techniques. Toluene, benzene and hexane were dried by refluxing over potassium and were stored over potassium mirrors. Dichloromethane (DCM) was dried over CaH_{2} and was stored over 4 Å molecular sieves. All solvents were degassed before use. For NMR spectroscopy C_{6}D_{6} was dried by refluxing over K and CD_{2}Cl_{2} was dried by refluxing over CaH_{2}. Both NMR solvents were vacuum transferred and degassed by three freeze–pump–thaw cycles before use. 1Ln,^{14,35} [H(SiEt_{3})_{2}][B(C_{6}F_{5})_{4}]^{21a} and 3Sm^{36} were prepared according to literature methods.
^{1}H (400 MHz), ^{13}C{^{1}H} (100 MHz and 125 MHz), ^{13}C{^{19}F} (126 MHz), ^{11}B{^{1}H} (128 MHz) and ^{19}F{^{1}H} (376 MHz) NMR spectra were obtained on an Avance III 400 MHz or 500 MHz spectrometer at 298 K. These were referenced to the solvent used, or to external TMS (^{1}H, ^{13}C), H_{3}BO_{3}/D_{2}O (^{11}B) or C_{7}H_{5}F_{3}/CDCl_{3} (^{19}F). UVvisNIR spectroscopy was performed on samples in Youngs tapappended 10 mm path length quartz cuvettes on an Agilent Technologies Cary Series UVvisNIR Spectrophotometer from 175–3300 nm. FTIR spectra were variously recorded as Nujol mulls in KBr discs on a PerkinElmer Spectrum RX1 spectrometer or as microcrystalline powders using a Bruker Tensor 27 ATRFourier Transform Infrared (ATRFTIR) spectrometer. EPR spectroscopic measurements were performed at Xband using a Bruker superhighQ Xband resonator attached to a Bruker EMX bridge, on solid state and frozen solution samples contained in flamesealed quartz EPR tubes. Elemental analysis was carried out by Mr Martin Jennings and Mrs Anne Davies at the Microanalytical service, School of Chemistry, the University of Manchester. Elemental analysis results for 2Yb reproducibly gave low carbon values; this has consistently been seen for {N(Si^{i}Pr_{3})_{2}} complexes and we have previously attributed this observation to the formation of carbides from incomplete combustion.^{14,35,36,56} However, all other analytical data obtained are consistent with the bulk purity of 2Ln and 3Ln.
[Sm{N(Si^{i}Pr_{3})_{2}}_{2}][B(C_{6}F_{5})_{4}] (2Sm).
Benzene (30 mL) was added to 3Sm (0.843 g, 1 mmol) and [H(SiEt_{3})_{2}][B(C_{6}F_{5})_{4}] (0.911 g, 1 mmol) and the resultant orange reaction mixture was stirred overnight at room temperature. The solvent was removed in vacuo and the oily red solid was washed with hexane (3 × 20 mL) and dried in vacuo for 1 h. The resultant red solid was cooled to −78 °C, dissolved in DCM (5 mL), layered with hexane (10 mL) and stored overnight at −25 °C to yield red crystals of 2Sm (1.137 g, 76%). Anal. calcd (%) for C_{60}H_{84}N_{2}Si_{4}BF_{20}Sm: C, 48.47; H, 5.69; N, 1.88; found: C, 47.25; H, 5.63; N, 1.72. χT product (Evans method, 298 K, [D_{2}]DCM): 0.43 cm^{3} mol^{−1} K. ^{1}H NMR ([D_{2}]DCM, 400 MHz, 298 K): δ = −5.27 (br, 72H, v_{1/2} ∼ 10 Hz, CH(CH_{3})_{2}), 0.43 (br, 12H, v_{1/2} ∼ 50 Hz, CH(CH_{3})_{2}). ^{11}B{^{1}H} NMR ([D_{2}]DCM, 128 MHz, 298 K): δ = −16.76 (s). ^{19}F NMR ([D_{2}]DCM, 376 MHz, 298 K): δ = −133.17 (br, oF), −163.71 (br, pF), −167.60 (br, mF). The paramagnetism of 2Sm precluded assignment of its ^{13}C{^{1}H} and ^{29}Si NMR spectra. IR (ATR, microcrystalline): 2954 (s), 2870 (s), 2813 (s), 1642 (s), 1511 (s), 1459 (s), 1384 (m), 1273 (s), 1082 (s), 978 (s), 928 (s), 881 (s), 765 (m), 693 (s), 676 (m), 543 (s), 489 (s), 415 (s) cm^{−1}.
[Tm{N(Si^{i}Pr_{3})_{2}}_{2}][B(C_{6}F_{5})_{4}] (2Tm).
Benzene (30 mL) was added to 3Tm (1.905 g, 2.21 mmol) and [H(SiEt_{3})_{2}][B(C_{6}F_{5})_{4}] (2.012 g, 2.21 mmol) and the resultant yellow reaction mixture was stirred overnight at room temperature. The solvent was removed in vacuo and the oily yellowgreen solid was washed with hexane (3 × 20 mL) and dried in vacuo for 1 h. The resultant yellowgreen solid was cooled to −78 °C, dissolved in DCM (5 mL), layered with hexane (10 mL) and stored overnight at −25 °C to yield yellowgreen crystals of 2Tm (1.540 g, 46%). Anal. calcd (%) for C_{60}H_{84}N_{2}Si_{4}BF_{20}Tm: C, 46.06; H, 5.45; N, 1.76; found: C, 46.01; H, 5.55; N, 1.70. χT product (Evans method, 298 K, [D_{2}]DCM): 6.44 cm^{3} mol^{−1} K. ^{1}H NMR ([D_{2}]DCM, 400 MHz, 298 K): δ = 25.04 (br, v_{1/2}–800 Hz, CH(CH_{3})_{2}). ^{11}B{^{1}H} NMR ([D_{2}]DCM, 128 MHz, 298 K): δ = −12.39 (s). ^{19}F NMR ([D_{2}]DCM, 376 MHz, 298 K): δ = −128.51 (br, oF). The paramagnetism of 2Tm precluded assignment of its ^{13}C{^{1}H} and ^{29}Si NMR spectra. IR (Nujol): 2359 (m), 2340 (m), 1643 (w), 1514 (m), 980 (m), 918 (w), 897 (w), 800 (w), 773 (w), 756 (w), 700 (w), 683 (w), 667 (w), 660 (w) cm^{−1}.
[Yb{N(Si^{i}Pr_{3})_{2}}_{2}][B(C_{6}F_{5})_{4}] (2Yb).
Toluene (15 mL) was added to a precooled (−78 °C) mixture of 3Yb (0.425 g, 0.5 mmol) and [H(SiEt_{3})_{2}][B(C_{6}F_{5})_{4}] (0.455 g, 0.5 mmol). The resultant dark purple reaction mixture was allowed to warm to room temperature slowly and stirred overnight. The solvent was removed in vacuo and the oily dark purple solid was washed with hexane (3 × 20 mL) and dried in vacuo for 1 h. The resultant dark purple solid was cooled to −78 °C, dissolved in DCM (1.5 mL), layered with hexane (3 mL) and stored at −35 °C overnight to yield dark purple crystals of 2Yb (0.5272 g, 70%). Anal. calcd (%) C_{60}H_{84}N_{2}Si_{4}F_{20}BYb·CH_{2}Cl_{2}: C, 45.94; H, 5.44; N, 1.76; found: C, 44.81; H, 5.18; N, 1.58. χT product (Evans method, 298 K, [D_{2}]DCM): 2.13 cm^{3} mol^{−1} K. ^{1}H NMR ([D_{2}]DCM, 400 MHz, 298 K): δ = 11.02 (br, v_{1/2} ∼ 400 Hz, CH(CH_{3})_{2}). ^{11}B{^{1}H} NMR ([D_{2}]DCM, 128 MHz, 298 K): δ = −14.67 (s). ([D_{2}]DCM, 376 MHz, 298 K): δ = −131.58 (br, oF), −162.05 (br, pF), −165.15 (br, mF). The paramagnetism of 2Yb precluded assignment of its ^{13}C{^{1}H} and ^{29}Si NMR spectra. IR (Nujol): 1267 (w), 1086 (m), 980 (m), 945 (w), 885 (w), 800 (w), 704 (m), 660 (m) cm^{−1}.
[Tm{N(Si^{i}Pr_{3})_{2}}_{2}(Cl)] (3Tm).
A solution of ^{t}BuCl (0.82 mL, 7.5 mmol) in toluene (10 mL) was added dropwise to a precooled (−78 °C) solution of 1Tm (1.240 g, 1.5 mmol). The reaction mixture was allowed to warm slowly to room temperature and was stirred at room temperature for 30 min, resulting in a colour change from dark brown to light brown. Volatiles were removed in vacuo and the product was extracted with hexane (10 mL), filtered, concentrated to 7 mL and stored at −35 °C overnight to yield pale green crystals of 3Tm (0.930 g, 72%). Anal. calcd (%) C_{36}H_{84}N_{2}Si_{4}ClTm: C, 50.17; H, 9.82; N, 3.25; found: C, 50.39; H, 10.23; N, 4.11. χT product (Evans method, 298 K, [D_{6}]benzene): 6.31 cm^{3} mol^{−1} K. The paramagnetism of 3Tm precluded assignment of its ^{1}H, ^{13}C{^{1}H} and ^{29}Si NMR spectra. IR (Nujol): 1260 (w), 1245 (w), 1077 (w), 1061 (w), 1012 (m), 991 (w), 934 (s), 879 (m), 799 (w), 728 (m), 701 (s), 667 (m), 632 (m), 598 (m) cm^{−1}.
[Yb{N(Si^{i}Pr_{3})_{2}}_{2}(F)] (3Yb).
Toluene (20 mL) was added to a precooled (−78 °C) mixture of 1Yb (1.246 g, 1.5 mmol) and [Fe(Cp)_{2}][PF_{6}] (0.496 g, 1.5 mmol) with stirring, and a white vapour was observed. The orange reaction mixture was stirred overnight at room temperature. All volatiles were removed in vacuo and ferrocene was sublimed away from the crude product at 90 °C for 1.5 hours. The remaining crude orange powder (1.029 g) was extracted with hexane (10 mL), filtered, concentrated to 7 mL and stored at −35 °C overnight to yield orangered crystals of 3Yb (0.734 g, 58%). Anal. calcd (%) C_{36}H_{84}N_{2}Si_{4}FYb·0.8C_{6}H_{14}: C, 53.36; H, 10.45; N, 3.05; found: C, 53.92; H, 10.87; N, 3.73. χT product (Evans method, 298 K, [D_{6}]benzene): 1.78 cm^{3} mol^{−1} K. The paramagnetism of 3Yb precluded assignment of its ^{1}H, ^{13}C{^{1}H}, ^{19}F and ^{29}Si NMR spectra. IR (Nujol): 1247 (w), 1214 (w), 1071 (w), 1012 (w), 996 (w), 944 (m), 882 (m), 800 (w), 703 (m), 665 (m) cm^{−1}.
Conflicts of interest
There are no conflicts to declare.
Acknowledgements
We thank the UK Engineering and Physical Sciences Research Council (EPSRC) (EP/N007034/1 for M. V., EP/K039547/1 for a single crystal Xray diffractometer, EP/P002560/1 for consumables, a DTG studentship for H. M. N. and a Doctoral Prize Fellowship for C. A. P. G.), and the University of Manchester for funding this work. We acknowledge the EPSRC UK National Electron Paramagnetic Resonance Service for access to the EPR facility and the SQUID magnetometer, and the University of Manchester (Fellowship for N. F. C.) for access to the Computational Shared Facility. We thank the peer reviewers for constructive comments. We also thank Mr Adam Brookfield and Dr Amga Baldansuren for assistance with EPR spectroscopy and Dr Fabrizio Ortu for assistance with single crystal XRD. Research data files supporting this publication are available from Mendeley Data at doi: 10.17632/fh245wm5nj.1.
Notes and references

The Rare Earth Elements: Fundamentals and Applications, ed. D. A. Atwood, Wiley, Chichester, 2012 Search PubMed.

D. Gatteschi and C. Benelli, Introduction to Molecular Magnetism, WileyVCH, Berlin, 2015 Search PubMed.

Lanthanides and Actinides in Molecular Magnetism, ed. R. A. Layfield and M. Murugesu, WileyVCH, Berlin, 2015 Search PubMed.
 D. N. Woodruff, R. E. P. Winpenny and R. A. Layfield, Chem. Rev., 2013, 113, 5110 CrossRef CAS PubMed.
 J. D. Rinehart and J. R. Long, Chem. Sci., 2011, 2, 2078 RSC.

Luminescence of Lanthanide Ions in Coordination Compounds and Nanomaterials, ed. A. de BettencourtDias, John Wiley and Sons Ltd., London, 2014 Search PubMed.
 J.C. G. Bünzli, J. Coord. Chem., 2014, 67, 3706 CrossRef.

F. Ortu and D. P. Mills, Chapter 306  Low Coordinate Rare Earth and Actinide Complexes, in Handbook on the Physics and Chemistry of Rare Earths, ed. J.C. G. Bünzli and V. K. Pecharsky, Elsevier B. V., Amsterdam, 2019,55, p. 1 Search PubMed.
 J.C. G. Bünzli and C. Piguet, Chem. Soc. Rev., 2005, 34, 1048 RSC.
 M. Shibasaki and N. Yoshikawa, Chem. Rev., 2002, 102, 2187 CrossRef CAS PubMed.
 Y.S. Meng, L. Xu, J. Xiong, Q. Yuan, T. Liu, B.W. Wang and S. Gao, Angew. Chem., Int. Ed., 2018, 57, 4673 CrossRef CAS PubMed.
 N. F. Chilton, Inorg. Chem., 2015, 54, 2097 CrossRef CAS PubMed.
 L. Ungur and L. F. Chibotaru, Phys. Chem. Chem. Phys., 2011, 13, 20086 RSC.
 N. F. Chilton, C. A. P. Goodwin, D. P. Mills and R. E. P. Winpenny, Chem. Commun., 2015, 51, 101 RSC.
 C. A. P. Goodwin, F. Ortu, D. Reta, N. F. Chilton and D. P. Mills, Nature, 2017, 548, 439 CrossRef CAS.
 C. A. P. Goodwin, D. Reta, F. Ortu, N. F. Chilton and D. P. Mills, J. Am. Chem. Soc., 2017, 139, 18714 CrossRef CAS.
 C. A. P. Goodwin, D. Reta, F. Ortu, J. Liu, N. F. Chilton and D. P. Mills, Chem. Commun., 2018, 54, 9182 RSC.
 J. Liu, D. Reta, J. Cleghorn, Y. X. Yeoh, F. Ortu, C. A. P. Goodwin, N. F. Chilton and D. P. Mills, Chem.–Eur. J., 2019, 25, 7749 CAS.
 K. R. McClain, C. A. Gould, K. Chakarawet, S. J. Teat, T. J. Groshens, J. R. Long and B. G. Harvey, Chem. Sci., 2018, 9, 8492 RSC.
 F.S. Guo, B. M. Day, Y.C. Chen, M.L. Tong, A. Mansikkamäki and R. A. Layfield, Science, 2018, 362, 1400 CrossRef CAS PubMed.

(a) J. B. Lambert, L. Lin and S. Keinan, Org. Biomol. Chem., 2003, 1, 2559 RSC;
(b) E. S. Stoyanov, I. V. Stoyanova and C. A. Reed, Chem.–Eur. J., 2008, 14, 7880 CrossRef CAS PubMed.
 C. A. Gould, K. R. McClain, J. Yu, T. J. Groshens, F. Furche, J. R. Long and B. G. Harvey, J. Am. Chem. Soc., 2019, 141, 12967 CrossRef CAS PubMed.
 D. C. Bradley, J. S. Ghotra and F. A. Hart, J. Chem. Soc., Chem. Commun., 1972, 349 RSC.
 D. C. Bradley, J. S. Ghotra and F. A. Hart, J. Chem. Soc., Dalton Trans., 1973, 1021 RSC.

D. P. Mills and S. T. Liddle, Ligand Design in Modern Lanthanide Chemistry, in Ligand Design in Metal Chemistry: Reactivity and Catalysis, ed. M. Stradiotto and R. J. Lundgren, John Wiley & Sons Ltd., Hoboken, 2016, pp. 330–363 Search PubMed.

C. A. P. Goodwin and D. P. Mills, Silylamides: towards a halfcentury of stabilising remarkable felement chemistry, in Specialist Periodical Reports, Organometallic Chemistry: Volume 41, ed. I. Fairlamb, J. M. Lynam, N. J. Patmore and P. Elliott, Royal Society of Chemistry, Cambridge, 2017, pp. 123–156 Search PubMed.
 M. Niemeyer, Eur. J. Inorg. Chem., 2001, 1969 CrossRef CAS.
 A. Cofone and M. A. Niemeyer, Z. Anorg. Allg. Chem., 2006, 632, 1930 CrossRef CAS.
 I. V. Basalov, O. S. Yurova, A. V. Cherkasov, G. K. Fukin and A. A. Trifonov, Inorg. Chem., 2016, 55, 1236 CrossRef CAS PubMed.
 C. N. de BruinDickason, A. J. Boutland, D. Dange, G. B. Deacon and C. Jones, Dalton Trans., 2018, 47, 9512 RSC.
 P. Zhao, Q. Zhu, J. C. Fettinger and P. P. Power, Inorg. Chem., 2018, 57, 14044 CrossRef CAS.
 C. Eaborn, P. B. Hitchcock, K. Izod and J. D. Smith, J. Am. Chem. Soc., 1994, 116, 12071 CrossRef CAS.
 C. Eaborn, P. B. Hitchcock, K. Izod, Z.R. Lu and J. D. Smith, Organometallics, 1996, 15, 4783 CrossRef CAS.
 G. Qi, Y. Nitto, A. Saiki, T. Tomohiro, Y. Nakayama and H. Yasuda, Tetrahedron, 2003, 59, 10409 CrossRef CAS.
 C. A. P. Goodwin, N. F. Chilton, G. F. Vettese, E. Moreno Pineda, I. F. Crowe, J. W. Ziller, R. E. P. Winpenny, W. J. Evans and D. P. Mills, Inorg. Chem., 2016, 55, 10057 CrossRef CAS.
 C. A. P. Goodwin, B. L. L. Réant, J. G. C. Kragskow, I. M. DiMucci, K. M. Lancaster, D. P. Mills and S. Sproules, Dalton Trans., 2018, 47, 10613 RSC.
 J. Sievers, Z. Phys. B: Condens. Matter Quanta, 1982, 45, 289 CrossRef CAS.
 A. M. Bienfait, B. M. Wolf, K. W. Törnroos and R. Anwander, Inorg. Chem., 2018, 57, 5204 CrossRef CAS PubMed.
 L. J. Nugent, R. D. Baybarz, J. L. Burnett and J. L. Ryan, J. Phys. Chem., 1973, 77, 1528 CrossRef CAS.
 J. B. Lambert, S. Zhang and S. M. Ciro, Organometallics, 1994, 13, 2430 CrossRef CAS.

I. Bertini, C. Luchinat, G. Parigi and E. Ravera, NMR of Paramagnetic Molecules: Applications to Metallobiomolecules and Models, Elsevier B. V., Amsterdam, 2nd edn, 2015, pp. 255–276 Search PubMed.
 A. J. Pell, G. Pintacuda and C. P. Grey, Prog. Nucl. Magn. Reson. Spectrosc., 2019, 111, 1 CrossRef CAS.

(a) J. L. Stewart and R. A. Andersen, Polyhedron, 1998, 17, 953 CrossRef CAS;
(b) S. M. Mansell, B. F. Perandones and P. L. Arnold, J. Organomet. Chem., 2010, 695, 2814 CrossRef CAS.
 D. J. Liptrot and P. P. Power, Nat. Rev. Chem., 2017, 1, 0004 CrossRef.
 E. K. Brady, D. L. Clark, J. C. Gordon, P. J. Hay, D. W. Keogh, R. Poli, B. L. Scott and J. G. Watkin, Inorg. Chem., 2003, 42, 6682 CrossRef CAS PubMed.
 L. Perrin, L. Maron, O. Eisenstein and M. F. Lappert, New J. Chem., 2003, 27, 121 RSC.
 D. L. Clark, J. C. Gordon, P. J. Hay, R. L. Martin and R. Poli, Organometallics, 2002, 21, 5000 CrossRef CAS.
 L. Maron and O. Eisenstein, New J. Chem., 2001, 25, 255 RSC.

IUPAC, Compendium of Chemical Terminology, ed. A. D. McNaught and A. Wilkinson. Blackwell Scientific Publications, Oxford, 2nd edn, (“the gold book”), 1997 Search PubMed.
 R. E. Da Re, C. J. Kuehl, M. G. Brown, R. C. Rocha, E. D. Bauer, K. D. John, D. E. Morris, A. P. Shreve and J. L. Sarrao, Inorg. Chem., 2003, 42, 5551 CrossRef CAS PubMed.
 R. G. Denning, J. Harmer, J. C. Green and M. Irwin, J. Am. Chem. Soc., 2011, 133, 20644 CrossRef CAS PubMed.
 S. K. Sur, J. Magn. Reson., 1989, 82, 169 CAS.

A. Abragam and B. Bleaney, Electron Paramagnetic Resonance of Transition Ions, Clarendon Press, Oxford, 1970 Search PubMed.

J. H. Van Vleck, The Theory of Electric and Magnetic Susceptibility, Oxford University Press, Oxford, 1932 Search PubMed.
 M. Xémard, M. Cordier, F. Molton, C. Duboc, B. Le Guennic, O. Maury, O. Cador and G. Nocton, Inorg. Chem., 2019, 58, 2872 CrossRef PubMed.
 H. M. Nicholas, C. A. P. Goodwin, J. G. C. Kragskow, S. J. Lockyer and D. P. Mills, Molecules, 2018, 23, 1138 CrossRef PubMed , and references cited therein.
Footnote 
† Electronic supplementary information (ESI) available. CCDC 1880942–1880946. For ESI and crystallographic data in CIF or other electronic format see DOI: 10.1039/c9sc03431e 

This journal is © The Royal Society of Chemistry 2019 