Electronic structure comparisons of isostructural early d- and f-block metal(iii) bis(cyclopentadienyl) silanide complexes

We report the synthesis of the U(iii) bis(cyclopentadienyl) hypersilanide complex [U(Cp′′)2{Si(SiMe3)3}] (Cp′′ = {C5H3(SiMe3)2-1,3}), together with isostructural lanthanide and group 4 M(iii) homologues, in order to meaningfully compare metal-silicon bonding between early d- and f-block metals. All complexes were characterised by a combination of NMR, EPR, UV-vis-NIR and ATR-IR spectroscopies, single crystal X-ray diffraction, SQUID magnetometry, elemental analysis and ab initio calculations. We find that for the [M(Cp′′)2{Si(SiMe3)3}] (M = Ti, Zr, La, Ce, Nd, U) series the unique anisotropy axis is conserved tangential to ; this is governed by the hypersilanide ligand for the d-block complexes to give easy plane anisotropy, whereas the easy axis is fixed by the two Cp′′ ligands in f-block congeners. This divergence is attributed to hypersilanide acting as a strong σ-donor and weak π-acceptor with the d-block metals, whilst f-block metals show predominantly electrostatic bonding with weaker π-components. We make qualitative comparisons on the strength of covalency to derive the ordering Zr > Ti ≫ U > Nd ≈ Ce ≈ La in these complexes, using a combination of analytical techniques. The greater covalency of 5f3 U(iii) vs. 4f3 Nd(iii) is found by comparison of their EPR and electronic absorption spectra and magnetic measurements, with calculations indicating that uranium 5f orbitals have weak π-bonding interactions with both the silanide and Cp′′ ligands, in addition to weak δ-antibonding with Cp′′.


Introduction
1][12][13] Given that the physicochemical properties of the f-elements have been exploited in numerous technologies, 14,15 it follows that a deeper understanding of f-block silicon chemistry could lead to new applications that complement d-block silicon analogues.

View Article Online View Journal | View Issue
[An(Cp ′ ) 3 {Si(SiMe 3 ) 3 }] (An = Th, U) by some of us; 28 Porchia, 29 Tilley, 30 and Marks 31 have all reported examples of An silanide complexes that were not characterised in the solid state.
It has recently been demonstrated that the extent of covalency in f-block M-Si bonds can be established by a combination of 29 Si NMR spectroscopy and density functional theory (DFT) calculations. 32However, this approach is currently limited to diamagnetic complexes and the vast majority of f-block complexes are paramagnetic; conversely, pulsed EPR spectroscopy has been applied to quantify An-C bond covalency in 5f 3 U(III) and 6d 1 Th(III) substituted Cp complexes. 33Although no U(III) silanide complex has been structurally authenticated to date, we posited that a substituted Cp-supported system could provide the necessary kinetic stabilisation.Ti(III) bis-Cp silanide complexes have been extensively studied by uid solution continuous wave (CW) EPR spectroscopy, including mononuclear complexes with bidentate silanides, 34,35 and monodentate silanides supported by a tethered donor atom or neutral co-ligand, 34,[36][37][38] as well as dinuclear Ti(III) complexes; 39 however, powder and frozen solution spectra are rare. 39The only EPR spectra of nd 1 Zr(III) and Hf(III) bis-Cp silanides reported to date are of [K (18-crown-6)][M(Cp) 2 {[Si(SiMe 3 ) 2 SiMe 2 ] 2 }]. 35e reasoned that a series of early d-and f-block M(III) complexes containing M-Si bonds could be achieved by using two substituted Cp ligands and one bulky silanide.We decided to adapt our previous strategy where we prepared An(IV) silanide complexes with three Cp ′ and one hypersilanide ligand, {Si(SiMe 3 ) 3 }; 28 we were encouraged to continue using hypersilanide as this has provided the largest number of f-block silanide complexes to date, 6 and to increase the size of the Cp ′ ring to Cp ′′ ({C 5 H 3 (SiMe 3 ) 2 -1,3}) in an effort to maintain kinetic stabilization of the M-Si bonds when the number of coordinated ligands is reduced.The approach of using multiple silyl groups increases the number of signals to assign in 29 Si NMR spectra, but this was preferred to the use of related alkylsubstituted silanide ligands that we have only previously found applicable to Ln(II) systems. 32Here we report the synthesis of an isostructural family of M(III) complexes, [M(Cp ′′ ) 2 {Si(SiMe 3 ) 3 }] (M = Ti, Zr, La, Ce, Nd, U), providing an opportunity to directly compare the electronic structures of early d-and f-block silanide bonds.This is predominantly achieved using a combination of CW EPR spectroscopy and complete active space self-consistent eld-spin orbit (CASSCF-SO) calculations, complemented by supporting characterisation data including single crystal X-ray diffraction, elemental analysis, SQUID magnetometry, and NMR, UV-vis-NIR and ATR-IR spectroscopies.e could not prepare Hf(III) and Th(III) homologues of 3-M using these procedures; although the M(IV) precursors [M(Cp ′′ ) 2 Cl 2 ] (M = Hf, Th, 52 see ESI ‡ for full Experimental details) can be prepared in appreciable yields.Lappert previously reported that the reduction of [Th(Cp ′′ ) 2 Cl 2 ] with Na/K alloy in THF gave [Th(Cp ′′ ) 3 ] by ligand scrambling; 54 we found that some decomposition occurred when [M(Cp ′′ ) 2 Cl 2 ] (M = Hf, Th) were treated with KC 8 in THF, and though no products could be identied from the Hf reaction, we identied crystals of [Th(Cp ′′ ) 3 ] by single crystal XRD. 54

NMR spectroscopy
Multinuclear NMR spectroscopy was performed on 1-Ti, 2-M and 3-M (see ESI Fig. S1-S42 ‡ for annotated NMR spectra); we focus here on the spectra of 3-M.With the exception of diamagnetic 3-La, the collection of reliable NMR spectra for 3-M was challenging due to paramagnetism.C 6 D 6 solutions of 3-Ti and 3-Zr also showed decomposition at ambient temperatures (t1 2 ca.S25 and S27 ‡) that were assigned to the organosilane HSi(SiMe 3 ) 3 by comparison with an authentic sample. 62An additional resonance observed in the 29 Si NMR spectrum of 3-Ti at −21.83 ppm was attributed to silicone grease, whilst in 3-Zr a signal at d Si = −9.83ppm could not be condently assigned.Previous reports of complexes containing Ti(III)-Si and Zr(III)-Si bonds showed that solution decomposition processes are commonly observed by NMR spectroscopy. 24,35,60The rest of the 3-M series were stable in C 6 D 6 solution at ambient temperature for a sufficient duration for multinuclear NMR spectra to be acquired (ca. 1 h; experiments had to be performed at relatively fast acquisition times to obtain data that are representative of freshly prepared solutions, e.g.1D 29 Si INEPT 128 scans).
For diamagnetic 3-La the 1 H NMR spectrum showed the four expected resonances: two inequivalent Cp-H signals for the 4,5and 2-Cp-H positions at 7.17 and 7.46 ppm, respectively, and two resonances in a ratio of 4 : 3 for the chemically inequivalent trimethylsilyl environments, d H : 0.25 (Cp-SiCH 3 ) and 0.55 ppm (Si(SiCH 3 ) 3 ); these correlated with ve resonances in the 13 C NMR spectrum.The 29 Si NMR spectra of 3-La revealed three resonances; those at −6.57 and −10.32 ppm were respectively assigned as Si(SiCH 3 ) 3 and Cp-SiCH 3 via a 1 H- 29 Si HMBC experiment, whilst a weak signal at −130.25 ppm correlated with the 1 H resonances of the hypersilanide ligand; the latter signal can only be tentatively assigned as the quaternary metalbound silicon atom due to quadrupolar broadening from coupling to 99.9% abundant I = 7/2 139 La nuclei.
The  29 Si NMR spectrum of 3-Nd (22.18 ppm), whilst no 29 Si NMR signals could be seen for 3-Ti, 3-Zr, 3-Ce or 3-U.The experimental parameters of the 1 H- 29 Si HMBC experiment prohibited correlation with 1 H NMR resonances, therefore we cannot condently assign the 29 Si NMR resonance observed for 3-Nd; however, this is unlikely to be due to the metal-bound silicon atom, as this resonance is not observed in the 1D 29 Si NMR spectra of diamagnetic 3-La.To the best of our knowledge there have not been any previous reports of 29 Si NMR chemical shis for paramagnetic M(III)-Si complexes in the literature for the metals studied here. 63

Magnetism
Solutions of 1-Ti, 2-Zr and 3-M (M = Ti, Zr, Ce, Nd, U) in C 6 D 6 were prepared at 0 °C and the effective magnetic moments (m eff ) and molar magnetic susceptibilities (c M ) were measured by the Evans method immediately upon warming to 300 K (Table 2, ESI Fig. S43-S52 ‡). 64Solution magnetic susceptibilities for 1-Ti and 3-M are in good agreement with the corresponding data obtained from powdered samples examined by variabletemperature SQUID magnetometry, and CASSCF calculations (Table 2,     systems, as expected for mononuclear nd 1 complexes.Previous studies of 2-Zr have reported that this complex is essentially diamagnetic based on NMR chemical shis, 65 but also that it exhibits an EPR spectrum in solution. 45We nd near-zero magnetic susceptibility for a powder sample of 2-Zr, implying that the two S = 1/2 centres are strongly antiferromagnetically coupled in the solid state (Fig. 2).A solution of 2-Zr in C 6 D 6 at 300 K was found to be paramagnetic by the Evans method, but less than expected for two uncoupled S = 1/2 (0.26 compared to 0.75 cm 3 K mol −1 ), indicating weaker antiferromagnetic coupling than in the solid state, some minor sample decomposition, and/or the presence of some monomeric [Zr(Cp ′′ ) 2 Cl] (1-Zr) in solution as proposed by Antiñolo et al. 65 SQUID magnetometry and EPR spectra of 2-Ce, 2-Nd and 2-U showed extensive exchange interactions between the metal ions; these data are challenging to model, 66 and will be communicated in a separate publication as they are outside the main focus of this study.Complexes 3-Ce and 3-Nd are more strongly magnetic than 3-Ti and 3-Zr due to the presence of orbital angular momentum; the smooth decrease in c M T with reducing temperature arises from crystal eld splitting of the lowest total angular momentum multiplet.Assuming well-isolated ground Kramers doublets for 3-Ce and 3-Nd, the 2 K magnetisation data suggest j±5/2i and j±9/2i ground states, respectively (ESI Table S5, Fig. S97b and S98b; ‡ in an axial crystal eld, M sat = 1 2 g J m J , where g J is the Landé g-factor 67 ).For 3-U the m eff (c M T) of 3.32 m B (1.38 cm 3 K mol −1 ) at 300 K is characteristic of a 4 I 9/2 U(III) ion; 27 this value smoothly decreases to 3.05 m B (1.16 cm 3 K mol −1 ) at 50 K, then rapidly decreases to 0.98 m B (0.12 cm 3 K mol −1 ) at 1.8 K.The sharp decrease below 50 K occurs due to slow thermalisation of the sample on cooling (see ESI for details ‡).The M sat value at 2 K and 7 T is diagnostic of a j±9/2i ground state (ESI Fig. S99b ‡).Alternating current susceptibility measurements on 3-U show out-of-phase signals owing to slow relaxation of magnetisation below 5 K (ESI Fig. S102-S105 ‡), which modelling suggests arises due to Raman and quantum tunnelling of magnetisation (QTM) processes (ESI Fig. S106 and Tables S6, S7 ‡).Slow relaxation is common for U(III) complexes, [68][69][70][71][72][73][74][75] however, as there is no effective barrier observed for the reversal of magnetisation and closed hysteresis loops around zero eld (ESI Fig. S101 ‡), we do not refer to 3-U as a single-molecule magnet, following several literature denitions. 76,77-vis-NIR spectroscopy Solutions of 1-Ti, 2-M and 3-M were prepared at 0 °C in toluene (2 mM concentration for all complexes) and warmed to room temperature to immediately record UV-vis-NIR spectra (Fig. 3; see ESI Fig. S65-S79 ‡ for individual spectra).Some spectra, most notably 2-Nd and 2-U, contain several jagged features; this is attributed to a combination of the most intense absorption maxima being close to the detector limit at the concentrations used, and the spectral resolution of 1 nm.Intense charge transfer (CT) absorptions tailing in from the UV region are found for 3-Zr ( y max = 22 400 cm −1 ; 3 = 890 M −1 cm −1 ), 3-La ( y max = 22 200 cm −1 ; 3 = 1880 M −1 cm −1 ), 3-Ce ( y max = 23 500 cm −1 ; 3 = 1200 M −1 cm −1 ), 3-Nd ( y max = 24 000 cm −1 ; 3 = 1570 M −1 cm −1 ) and 3-U ( y max = 24 200 cm −1 ; 3 = 2160 M −1 cm −1 ) (Fig. 3b).These transitions were not observed for 1-Ti or 2-M (Fig. 3a), and therefore can be assigned to changes in CT upon replacing a halide with hypersilanide.Apart from this CT band, the spectra of 2-La, 3-Zr and 3-La are otherwise essentially featureless.

EPR spectroscopy
EPR spectra have been recorded at two frequencies where possible and simulations have modelled both spectra simultaneously in EasySpin; 89 full details of simulations are included in the ESI (Tables S8-S12 and Fig. S107-S121 ‡).EPR spectra of 2-Ce, 2-Nd and 2-U show signicant exchange interactions between the metal ions, and these will be studied in a separate publication.
Titanium and zirconium, nd 1 .A uid solution spectrum of 1-Ti gives g iso = 1.9620 (ESI Fig. S107 ‡), in agreement with the literature value in toluene (g iso = 1.961) 90 and conrming that 1-Ti is stable in solution (g ave from frozen solution is 1.9624, see below).In contrast, 3-Ti and 3-Zr decompose over several hours at room temperature in uid aromatic solutions, consistent with nd 1 complexes being kinetically labile, particularly when  coordinatively unsaturated. 91Reproducible frozen solution spectra for 2-Zr, 3-Ti and 3-Zr could be obtained for these complexes by dissolving powders in a solvent mixture (9 : 1 toluene : hexane) that was pre-cooled to ca. 250 K, then frozen at 77 K and measured immediately.The g-values obtained for monomeric nd 1 complexes (1-Ti, 3-Ti and 3-Zr) with this method are generally in excellent agreement with the powder spectra (Table 4).
The powder EPR spectrum of 2-Zr is isotropic with g = 1.9825 and a half-eld transition, indicating that these transitions arise from a triplet state, reecting the dimeric structure in the solid state (ESI Fig. S112-S114 ‡).There is one previous report of an EPR spectrum of 2-Zr with g iso = 1.9506 and A iso = 60 MHz in uid toluene solution at 300 K. 45 Measurement of a frozen solution sample of 2-Zr (see Section 12 of the ESI ‡) gives clearly anisotropic spectra (g 1 = 1.9961, g 2 = 1.9834, g 3 = 1.8618; the g 3 resonance is at 360 and 1307 mT at X-and Q-band frequencies, respectively, far outside the spectral range of the powder spectrum of 2-Zr, ESI Fig. S115 cf.S112 ‡) and the half-eld transition is absent (ESI Fig. S114 ‡).These data suggest that the dimer breaks apart in solution to form monomeric 1-Zr (S = 1/2), in accord with solution magnetic susceptibility and UV-vis-NIR data (see above); there may be some 2-Zr still present in the frozen solution; however, at 50 K the signal (arising from Boltzmann population of the triplet state) is dwarfed by the S = 1/2 spectrum.Thus, we henceforth discuss these frozen solution results as representing 1-Zr.
Ti(III) and Zr(III) are S = 1/2 and are expected to have anisotropic g-values (g 1 , g 2 , g 3 ) close to the free electron value (g e z 2.0023), with deviations from g e (Dg i = g i − g e ) reecting second order spin-orbit coupling with low energy crystal eld states. 92X-band EPR spectra of frozen solutions of 1/3-Ti clearly show three g-features ( 47 Ti/ 49 Ti hyperne coupling not resolved), while spectra of 1/3-Zr show a similar rhombic pattern superimposed with hyperne coupling to the 91 Zr nuclear spin (11% abundant I = 5/2; Table 4, Fig. 4, ESI Fig. S115 and S117 ‡).Superhyperne coupling to a-29 Si and b-29 Si have been observed before, 35 but are not resolved here.5][36][37][38][39] The EPR spectra of 1/3-Ti/Zr are more comparable to pseudo-trigonal planar bent metallocenes [M(Cp R ) 2 X]. 88The pattern of 91 Zr hyperne constants (Table 4; one large and two small) are typical for 4d z 2 1 ground states with z aligned along g 1 , 92 which also suggests an electronic structure tending to pseudo-trigonal planar environments.Powder spectra of ground samples of 1-Ti, 3-Ti, 2-Zr and 3-Zr are considerably broadened relative to frozen solution spectra (likely owing to unresolved intermolecular dipolar interactions), resulting in overlap of the g 1 and g 2 features (see ESI Fig. S108, S110, S112 and S116 ‡).
Cerium 4f 1 , neodymium 4f 3 and uranium 5f 3 .The powder EPR spectra of 3-Ce at 7 K are characteristic of a rhombic S eff = 1/2 with three g-features characterised at X-band and one at Q-band (Fig. 5a, ESI Fig. S118a and b ‡).Simultaneous modelling of the spectra give g 1 , g 2 and g 3 of 3.884, 0.888 and 0.493, respectively.The large g 1 suggests majority j±5/2i component of the ground state, consistent with the M sat value.The frozen solution X-band EPR spectrum of 3-Ce in toluene : hexane (9 : 1) has g 1 of 3.907 and transverse g-values less than 0.4 (ESI Fig. S118c ‡), indicating a geometry change in solution to form a more axial ground state.
The X-band powder EPR spectrum of 3-Nd at 7 K revealed only a single g 1 feature of the ground Kramers doublet with resolved 143 Nd and 145 Nd hyperne coupling within the accessible eld range (Fig. 5b and ESI Fig. S119a-c ‡).The g 1 value of 5.490 is less than the value of 6.55 expected for a pure j±9/2i, but is larger than the maximum g value for a pure j±7/2i state, implying a mixed j±9/2i ground doublet, consistent with the M sat value.Whilst g 2 and g 3 were not visible in the powder spectrum, a frozen solution X-band spectrum showed part of a g 2 feature at the highest elds, estimated as g 2 = 0.360 (Fig. 5d).
The powder X-band EPR spectrum of 3-U at 7 K exhibits a sharp feature at g = 6.055 (Fig. 5c), assigned as the g 1 feature of an axial S eff = 1/2 ground state with g 2 , g 3 < 0.4.This indicates a majority j±9/2i ground state, consistent with the M sat value.Two broader peaks were observed at 67 and 175 mT (g = 9.94 and 3.82), however, the Q-band EPR spectrum was too weak to establish whether these peaks behaved as true g-features.The extra features cannot be explained by dipolar interactions (ESI Fig. S121 ‡), and we have not been able to assign them.Solution X-band spectra of 3-U at 5 K showed two sharp g 1 features at 5.949 and 5.536, indicating two very similar axial species present in solution (ESI Fig. S120c and d ‡).

CASSCF calculations
We have investigated the electronic structures of 1-Ti/Zr, 2-Zr and 3-M by CASSCF-SO calculations (performed in Open-Molcas, 93,94 see ESI for details ‡).For discussion of the data we adopt the coordinate system of Petersen and Dahl, with x along the M-Si/Cl axis, y tangential to Cp ′′ -M-Cp ′′ and z perpendicular to the plane dened by the Cp ′′ centroids and coordinating Si/Cl atom. 95,96itanium and zirconium, nd 1 .For 1-Ti, 3-Ti and 3-Zr we used the single crystal XRD structures, while for 1-Zr we used a DFT-optimised geometry (ESI Table S13 ‡).State-averaged (SA) CAS(7,8)SCF calculations were performed averaging over ve doublets, with an active space of the ve 3d or 4d orbitals and three almost doubly occupied M-L bonding orbitals with considerable (17-25%) d xy , d yz , and d x 2 −y 2 character (ESI Fig. S122-S125 ‡).Calculations for 2-Zr used the crystal structure with one Zr(III) centre replaced with diamagnetic Y(III) (2-Zr ′ ), averaging over 5 doublets in an active space of ve 4d orbitals, three 4p orbitals and four almost doubly occupied M-L bonding orbitals with 15-19% d character (ESI Fig. S126 ‡).The active space differs for 2-Zr ′ because the symmetry match of the Zr 4d xz orbital and chloride s-orbitals results in stronger bonding and anti-bonding interactions compared to 1/3-Ti/Zr, requiring a pair of orbitals to be included in the active space.Furthermore, 4p orbitals were included as they displaced the M-L bonding orbitals in orbital optimisation if not included initially.
Complexes 1/3-Ti/Zr showed similar results, with the nd 1 electrons located in their respective nd z 2 orbitals.A d z 2 1 ground state is standard for bent metallocenes [M(Cp R ) 2 X] and is consistent with our analysis of the 91 Zr hyperne coupling (see above). 88The excited states place d-d transitions for 1-Ti and 3-Ti between 16 000 and 21 000 cm −1 (Fig. 6, ESI Tables S14 and  S15 ‡), in reasonable agreement with experiment (Fig. 3).For 3-Zr, the d-d transitions are calculated at higher energies (23 224, 25 402 and 30 058 cm −1 , ESI Table S18 ‡) and are thus obscured by the CT band in agreement with experiment (Fig. 3).
The dimeric structure of 2-Zr results in the lowest energy d-d transition calculated for 2-Zr ′ shiing from below 5000 cm −1 for 1-Ti, 3-Ti and 3-Zr to 14 031 cm −1 , with the other d-d transitions at 18 101, 20 609 and 22 362 cm −1 (Fig. 6 and ESI Table S17 ‡); the former band corresponds well with the low energy shoulder at 13 100 cm −1 (3 = 300 M −1 cm −1 ; Fig. 3), while the latter three agree well with the broad absorption at 17 900 cm −1 (3 = 1060 M −1 cm −1 ; Fig. 3).There are no predicted d-d transitions in this region for 1-Zr, which has a low-energy band predicted at 7010 cm −1 and all other d-d transitions would be hidden beneath the CT band tailing from the UV region (ESI Table S16 ‡).These calculations agree with experimental data (see above) that both 1-Zr and 2-Zr are present in solution.
The pattern of calculated g-values for 1/3-Ti/Zr wellreproduces the experimental data (Table 3, ESI Tables S14-S16 and S18 ‡), while the absolute shis in g 2 and g 3 are overestimated (ESI Table S11 ‡).The CASSCF calculations also give insight into the orientations of the g-values: for all 1/3-Ti/Zr, the largest value g 1 is along the z direction (pseudo-three-fold), the intermediate value g 2 is along x (M-Cl/Si bond), and the smallest value g 3 is along y (tangential to Cp ′′ -M-Cp ′′ ; Fig. 7 and ESI Fig. S134 ‡).While orbital energies do not strictly exist in a SA-CASSCF calculation, in this case the ve states are each dominated by a single conguration with the unpaired electron located in one of the ve d-orbitals, so approximate d-orbital energies can be determined by assigning the state energies to the energy of the singly occupied natural orbital for that state (Fig. 6, ESI Fig. S127, S128, S130-S132 and ESI Tables S14-S18 ‡).
We nd that the d z 2 orbital remains lowest in energy in all cases, followed by the d xz orbital, which varies signicantly in energy between complexes.The d xy and d yz orbitals lie far higher in energy (16 000-26 000 cm −1 ) and are relatively close in energy to each other, whilst the d x 2 −y 2 orbital is highest in energy as chloride and hypersilanide are both strong s-donor ligands.This is consistent with the Lauher-Hoffmann bonding model of a naked bent metallocene, 88,97 with an additional monodentate ligand.With these orbital energies we can rationalise the observed g-values.Spin-orbit coupling along z cannot mix in any excited state into a ground state with the electron in the d z 2 orbital, 92 and so Dg z is approximately zero which agrees well with experiment (ESI Table S11 ‡; small deviations away from g e are ascribed to mixing of d x 2 −y 2 into the ground state 95 ).Spinorbit coupling along x (y) can mix in a state with an unpaired electron in d yz (d xz ) into the ground state to shi g x (g y ), where Dg x (Dg y ) is inversely proportional to the energy of the excited orbital (ESI eqn (S4) and (S5) ‡).As d xz is much lower in energy than d yz , Dg y is larger than Dg x , and as the d shell is less than half-lled all Dg are negative; hence g 3 is along y and g 2 is along x; this explains the obtained ab initio orientations (Fig. 7 and ESI Fig. S134 ‡).
The value of Dg 3 reects the energy of the d xz orbital and therefore the p-bonding character of the X ligand; this has been used previously to construct a p-donor spectrochemical series for [Ti(Cp*) 2 X]. 88Between 1-Ti/Zr and 3-Ti/Zr Dg 3 doubles, reecting the d xz orbital being 40% lower in energy for 3-Ti/Zr; this is because chloride is a p-donor and d xz is formally p-antibonding in 1-Ti/Zr (the nominal d xz orbitals have 1.9% and 3.1% Cl 2p z character for 1-Ti and 1-Zr, respectively; ESI Fig. S127 and S130 ‡).In contrast, hypersilanide is a weak p-acceptor, and such interactions can be seen with a very low isosurface value (ESI Fig. S129 and S133 ‡).Upon moving from Ti to Zr there is more effective overlap of the 4d and ligand orbitals, leading to a larger crystal eld splitting (Fig. 6), and a decrease in the metal contribution to the singly occupied natural orbitals (ESI Tables S14-S16 and S18 ‡).Spin-orbit coupling also increases moving from Ti to Zr, which acts to increase jDgj, whilst increased d-orbital splitting and ligand-metal mixing oppose this; as all Dg become more negative upon going from 1/3-Ti to 1/3-Zr, spin-orbit coupling is the dominant effect.
Cerium 4f 1 .For 3-Ce, a CAS(1,7)SCF calculation averaged over seven spin doublets was performed for an active space containing seven 4f orbitals.The resulting spin-free states were mixed with spin-orbit calculation to obtain the states of the 2 F 5/2 ground term (ESI Table S19 ‡).The ground state for 3-Ce is predicted to be 90% j±5/2i and 8% j±1/2i, consistent with the slightly reduced M sat and g 1 values from those expected for a pure j±5/2i ground doublet; the calculated g-values are in good agreement with experiment (Table 3).Differing from 1/3-Ti/Zr, the easy axis (g 1 ) is oriented tangential to the Cp ′′ -M-Cp ′′ direction (y), with the intermediate axis (g 2 ) along the Ce-Si bond (x) and the hard axis (g 3 ) along z (Fig. 8a).The magnetic anisotropy of 4f ions is oen dictated by pure electrostatic considerations, 98,99 and so the dominance of a j±5/2i ground state (which has an oblate spheroidal 4f electron density) with its easy axis along y indicates the pair of sandwich-like Cp ′′ ligands are stronger inuences than the hypersilanide.This is similar to the situation in [CeCp ttt 2 Cl] (Cp ttt = {C 5 H 2 tBu 3 -1,2,4}), which has a ground state of 97% j±5/2i and 3% jH1/2i, compared to [CeCp ttt 2 {(C 6 F 5 -k 1 -F)B(C 6 F 5 ) 3 }], which has a 100% j±5/2i ground state. 78The larger extent of j±1/2i mixing in the ground state of 3-Ce compared to [CeCp ttt 2 Cl] (reected also in the experimental g 1 -values of these two complexes: 3.884 vs. 4.19, 78 respectively) argues for a stronger crystal eld of hypersilanide over chloride; note that is the opposite of the discussion above for 1-Ti/Zr vs. 3-Ti/Zr as the bonding interactions with 4f orbitals are irrelevant and the more diffuse chloride ligand provides a weaker electrostatic eld.
Neodymium and uranium, nf 3 .The electronic structures of 3-Nd and 3-U were investigated by CAS(3,7)SCF calculations using an active space consisting of the seven nf orbitals and averaging over 35 spin quartets and 112 spin doublets (entire 5f 3 conguration), and then mixing with spin-orbit coupling (Fig. 8b, c, ESI Tables S20 and S21 ‡).The crystal eld splitting of the 4 I 9/2 ground term for 3-Nd gives a mixed ground doublet with 76% j±9/2i + 15% j±5/2i + 5% j±1/2i, whose g-values match reasonably well with experiment (Table 3): the mixed ground state composition explains the lower g 1 and M sat values than predicted for a pure j±9/2i state.Despite the large mixing, the pattern of g-values appears more axial with smaller and similar g 2 and g 3 values.Like for 3-Ce, the easy axis (g 1 ) is aligned tangentially to the Cp ′′ -M-Cp ′′ direction (y), with g 2 and g 3 in the perpendicular xz plane (Fig. 8b).Also likewise to 3-Ce, the dominant j±9/2i ground state and the easy axis orientation are indicative of the Cp ′′ ligands dominating the electrostatic potential as j±9/2i also has an oblate spheroidal 4f electron density. 98,99ASSCF calculations on 3-U averaging over all 5f 3 states reproduce well the experimental susceptibility and g 1 value, and suggest a 95% j±9/2i ground state.Like 3-Ce and 3-Nd, the oblate spheroidal electron density of j±9/2i in 3-U is orientated with g 1 tangential to Cp ′′ -M-Cp ′′ (Fig. 8c). 98,99However, this CASSCF calculation overestimates M sat ; averaging instead over only the 4 I 9/2 ground term (13 spin quartets) gives a more accurate reproduction of M sat but underestimates g 1 and the magnetic susceptibility (ESI Fig. S99 and ESI Table S22 ‡), suggesting a slightly more mixed ground state of 89% j±9/2i + 6% j±5/2i.The true ground state composition of 3-U is likely between these values, but it is certainly less mixed than that of 3-Nd.
The CAS(3,7)SCF averaged orbitals for 3-U (35 quartets and 112 doublets) showed small 6d orbital contributions (∼5%), so we extended the active space to also include two low-lying 6d orbitals (6d z 2 and 6d xz , ESI Fig. S136, ‡ in accordance with 3-Ti and 3-Zr, see above).Averaging over all f 3 and f 2 d 1 congurations in this active space, the f 2 d 1 states lie at ∼7000 cm −1 (ESI Fig. S138 ‡), suggesting that spin-and Laporteallowed f / d transitions contribute to the broad band from 6000 to 17 000 cm −1 in the UV-vis-NIR spectrum (Fig. 3).Examining the averaged molecular orbitals with a low isosurface value showed that the 5f orbitals participate in weak dantibonding and p-bonding with Cp ′′ , and weak p-bonding with low lying vacant orbitals on Si (ESI Fig. S137 ‡).

Discussion
From the single crystal X-ray diffraction and DFT-optimised structural data it can be seen that the M-Si bonds are 0.26-0.30Å longer than the corresponding M-Cl bonds in 1/3-Ti/Zr aer correcting for the difference in single-bond covalent radii (Table 1). 55The steric bulk of the hypersilanide ligand imposes the longer M-Si bonds, which leads to weaker M-Si bonding interactions through reduced orbital overlap.The bonding in the {M(Cp ′′ ) 2 } + fragments are expected to be similar in 1-M vs. 3-M.Whilst the orientation of Cp ′′ rings changes, the key M À Cp 00 cent distances and Cp 00 cent À M À Cp 00 cent angles for Zr(III) are virtually unchanged, and for Ti(III) there are only modest respective increases of these metrics of 0.014 Å and 2.6°.
The 3d 1 and 4d 1 systems 3-Ti and 3-Zr exhibit nd z 2 1 ground state occupancies, with spin density perpendicular to the Cp 00 cent , Cp 00 cent , Si plane.The nd z 2 1 ground state is non-bonding and is favoured by strong anti-bonding interactions of the remaining nd-orbitals with the two Cp ′′ ligands as well as with the hypersilanide ligand, which acts as a s-donor.The hypersilanide ligand also acts as a weak p-acceptor, giving rise to a low energy nd xz 1 excited state in both cases (<10 000 cm −1 ).
The stabilisation of nd z 2 and nd xz orbitals is echoed in 3-U, where low energy f 3 / f 2 d 1 transitions are seen above 6000 cm −1 .However, for 3-Ce and 3-Nd the f n / f n−1 d 1 transitions are not proximate to the ground states, as expected.Due to the orbitally non-degenerate ground state, the anisotropy of the g-tensors in 3-Ti and 3-Zr are dominated by second-order spin-orbit coupling with the nd xz 1 excited state.
This results in a large shi of g y away from g e , where y is tangential to Cp ′′ -M-Cp ′′ .In f-block complexes with near-complete f-orbital degeneracy, anisotropy in the ground state is dictated by electrostatics, and we nd that the bis-Cp ′′ crystal eld dominates over the hypersilanide contribution.This favours oblate f-electron density in the ground state, giving ground states dominated by j±5/2i, j±9/2i and j±9/2i, for 3-Ce, 3-Nd and 3-U, respectively, with the magnetic easy axis tangential to Cp ′′ -M-Cp ′′ .These data indicate that the M-Si bonds in 3-Ti and 3-Zr are both more covalent than the M-Si bonds in 3-Ln and 3-U, whilst the greater covalency of the M-Si bonds in 3-Zr vs. 3-Ti is also evident from the respective magnitudes of 4d vs. 3d crystal eld splitting.The presence of slow magnetic relaxation for 3-U and the lack of such behaviour for 3-Nd is an outlier compared to literature examples, where analogous f 3 compounds tend to show more similar behaviour. 75,100It is likely that faster magnetic relaxation occurs for 3-Nd due to the lower purity of the ground state, which itself is likely to arise due to mixing with low energy spin-orbit states at 65 and 176 cm −1 : the more-pure ground state in 3-U on the other hand, is well-separated from the lowest excited states at 330 and 524 cm −1 , which likely arises from a larger crystal eld effect due to 5f vs. 4f orbitals.This is in accord with the UV-vis-NIR spectrum of 3-U, where the f / d transitions are low energy for U(III), 53 and are a hallmark of polarised covalent metal-ligand bonding. 15The greater involvement of 5f vs. 4f orbitals in M-Si bonds was also seen in ab initio calculations.It follows that the M-Si bond in 3-U has greater covalency than that of 3-Nd, with the M-Si bonds of 3-La, 3-Ce and 3-Nd assumed to show similar predominantly electrostatic character due to their valence 4f orbitals.

Conclusions
We have reported the synthesis and characterisation of a series of isostructural early d-and f-block M(III) bis(cyclopentadienyl) hypersilanide complexes, providing the rst structurally authenticated examples of U(III) and Nd(III) silanides.By using a combination of CW EPR spectroscopy and CASSCF calculations we have shown that the d-block complexes herein have 3/4d z 2 1 ground states aligned perpendicular to the coordination plane, with the hypersilanide ligand acting as a strong s-donor and weak p-acceptor to impart axial anisotropy through p-bonding with low-lying nd xz orbitals; as expected, the orbital splitting is greater for 4d z 2 1 Zr(III) vs. 3d z 2 1 Ti(III).In contrast, the early f-block Ln/U(III) silanide 4/5f n complexes exhibit predominantly ionic bonding, with the dominant crystal eld imparted by the Cp ′′ ligands favouring oblate spheroidal f-electron densities and magnetic easy axes tangential to Cp ′′ -M-Cp ′′ .
The uranium silanide complex was found to exhibit increased covalency over Ln congeners, with calculations showing weak p-bonding between the 5f orbitals and both the silanide and Cp ′′ ligands, and weak d-antibonding between 5f orbitals and Cp ′′ .The greater crystal eld imposed for 5f 3 U(III) vs. 4f 3 Nd(III) gave a purer ground state due to the energies of low-lying excited states being raised to the extent that they can no longer mix, switching on slow magnetic relaxation in the former complex below 5 K.The U(III) congener additionally displayed low energy 5f 3 / 5f 2 6d 1 electronic transitions from 6000 to 17 000 cm −1 , signifying that the nd z 2 and nd xz orbitals have been stabilised in a similar manner to Ti(III) and Zr(III) homologues.Together, the combination of data acquired herein show the qualitative ordering of the extent of covalency to be Zr > Ti [ U > Nd z Ce z La, and reveal clear differences between the compositions of early d-block, Ln and An M-Si bonds.

Fig. 1
Fig. 1 Molecular structures of: (a) 3-Ti determined at 100 K and (b) 3-U determined at 150 K, with selective atom labelling.Displacement ellipsoids set at 30% probability level and hydrogen atoms removed for clarity.

Fig. 5
Fig. 5 CW X-band EPR spectra of (a) 3-Ce powder at 8 K, (b) 3-Nd powder at 7 K, (c) 3-U powder at 7 K, (d) 15 mM frozen solution of 3-Nd in 9 : 1 toluene : hexane at 7 K.Two perpendicular orientations of powder spectra are shown in black and blue.Simulations using parameters from ESI Table S12 ‡ are shown in red.Asterisk (*) denotes feature intrinsic to cavity.

Fig. 8
Fig. 8 CASSCF-calculated magnetic axes (blue: g 1 , easy; green: g 2 , intermediate; red: g 3 , hard) for complexes 3-Ce (a), 3-Nd (b) and 3-U (c).Length of the magnetic axes reflects the size of the g-values.Metal, silicon and carbon shown as metallic green, orange and grey respectively.Hydrogen atoms are omitted for clarity.
2 h); d Si resonances at −116.11 and −11.97 ppm grew in intensity during data collection (see ESI Fig.

Table 2
Magnetic moment, m eff

Table 3
CW EPR g-values of studied complexes as solid powders or in toluene : hexane (9 : 1) frozen solutions (FS).Band frequency: X = 9.37-9.47GHz, K = 24 GHz, Q = 34 GHz a Anisotropy in g-values not resolved.b Two species observed in solution, relative abundance given in brackets.