Synthesis and characterisation of light lanthanide bis-phospholyl borohydride complexes†

Organometallic lanthanide (Ln) chemistry is dominated by complexes that contain substituted cyclopentadienyl (Cp) ligands. Closely related phospholyls have received less attention, and although they have proven utility in stabilising low oxidation state Ln complexes the trivalent Ln chemistry of these ligands is limited in comparison. Herein, we synthesise two families of heteroleptic Ln complexes, [Ln(Htp)2(μBH4)]2 (Htp = 2,5-di-tert-butylphospholyl; 1-Ln; Ln = La, Ce, Nd, Sm), and [[Ln(Htp)2(μ-BH4)2K(S)]n (2-Ln, Ln = La, Ce, S = 2 DME, n = 2; 3-Ce, Ln = Ce, S = Et2O and THF, n = ∞) via the reactions of parent [Ln (BH4)3(THF)3.5] with K(Htp), to investigate differences between Ln complexes with substituted phospholyl ligands and analogous Cp complexes. Complexes 1–3-Ln were characterised as appropriate by single crystal XRD, SQUID magnetometry, elemental analysis, multinuclear NMR, ATR-IR and UV-Vis-NIR spectroscopy. Ab initio calculations reveal that small changes in the Ln coordination spheres of these complexes can have relatively large influences on crystal field splitting.


Introduction
Although molecular lanthanide (Ln) chemistry lags behind that of the d-block, the remarkable physical properties and technological importance of Ln elements is now driving more rapid developments. 1 Since the first examples of Ln cyclopentadienyl (Cp) complexes, [Ln(Cp) 3 ], were isolated in 1954, 2 Cp ligands and their derivatives (Cp R ) have dominated Ln organometallic chemistry. 3,4 In the interim, phospholyl ligands have proven to be useful alternatives to Cp R ligands in Ln chemistry and tend to exhibit η 5 -binding modes with these metals. 5 Recent studies have shown that phospholyls are especially effective at stabilising divalent Ln complexes due to them being less electron-donating than their corresponding Cp R ligands. 6 The presence of a phosphorus lone pair can also promote the η 1 -binding mode and affect reactivity profiles, and 100% abundant 31 P nuclei can provide a useful spectroscopic handle. 5 However, despite these advantageous properties, trivalent Ln phospholyl chemistry is relatively underdeveloped.

Synthesis
The heteroleptic complexes 1-Ln (Ln = La, Ce, Nd, Sm) were prepared from the parent [Ln(BH 4 ) 3 (THF) 3.5 ] 16 and two equivalents of K(Htp) 17 by modification of the synthesis of [Tm(Htp) 2 (μ-I)] 2 , using TmI 3 and K(Htp) (Scheme 1). 9 Di-n-butyl ether was selected as the reaction solvent as its high boiling point (140.8°C) allowed the reaction mixture to be heated significantly to increase the solubility of K(Htp). The crystalline yields for 1-La, 1-Ce, 1-Nd and 1-Sm were 31%, 41%, 15% and 7%, respectively, indicating that these salt metathesis reactions tend to become more sluggish for smaller Ln 3+ cations. We were able to monitor the formation of diamagnetic 1-La by 31 P NMR spectroscopy; the reaction appeared to proceed relatively cleanly but sluggishly, with only 1-La, K(htp) and a byproduct (Htp) 2 present in appreciable quantities, thus we postulate that the low yields of 1-Ln were due to the loss of material during recrystallization processes. The especially low isolated yields of 1-Nd and 1-Sm were attributed to the ready formation of (Htp) 2 , which was observed in the 31 P NMR spectra of all reaction mixtures but appeared to form in greater quantities for the smaller Ln. The formation of (Htp) 2 in salt metathesis reactions has previously been seen in the synthesis of [Ga(Htp)] from the reaction of GaBr with one equivalent of Li(Htp). 18 The related complexes 2-Ln and 3-Ce were synthesised by analogous procedures using DME or diethyl ether, respectively, in the reactions of [Ln(BH 4 ) 3 (THF) 3.5 ] with two equivalents of K(Htp). The crystalline yields for 2-La, 2-Ce and 3-Ce were 16%, 46% and 31%, respectively; for 2-Ln the reaction mixtures were heated but in the case of 3-Ce the reaction was performed at room temperature. Although we are unable to conclusively determine if the differences between the degree of oligomerisation in 2-Ce and 3-Ce are due to the reaction temperature as the boiling point of diethyl ether (34.6°C) is far lower than that of DME (85°C), it is evident that changing the solvent to diethyl ether has allowed the salt metathesis reaction to proceed at a lower temperature, which is important to note for future synthetic attempts.
In common with observations for the synthesis of 1-Ln above, the yield of 2-La was lower than 2-Ce, indicating that reaction vectors and crystallisation processes are highly sensitive to Ln 3+ cation size, and that (Htp) 2 formation is likely an issue when reaction mixtures are heated for an extended period of time. We did not adapt these methods to attempt to prepare Nd and Sm analogues for 2-Ln. Variations of reaction stoichiometries to three equivalents of K(Htp) and changes in temperature and reaction times to those outlined above did not provide homoleptic complexes. Complexes 1-Ln, 2-Ln or 3-Ce were isolated in similar crystalline yields to those stated above upon the variation of any of these parameters, though the amount of (Htp) 2 formed in the reaction mixtures of 1-Nd and 1-Sm appeared to increase when these were heated for prolonged periods at elevated temperatures and monitored by 31 P NMR spectroscopy.
Elemental analysis results consistently gave low carbon values, likely due to carbide formation from incomplete combustion, but all other analytical data were consistent with their bulk purity (see below); for 3-Ce elemental analysis values are in agreement with 1 H NMR data where complete desolvation Scheme 1 Synthesis of 1-Ln, 2-Ln and 3-Ce. occurred when the sample was exposed to vacuum for 1 hour (1 × 10 −2 mbar). The ATR-IR spectra of most complexes clearly exhibit absorptions from 2500 to 2100 cm −1 (see ESI Fig. S47-S55 †); these are attributed to B-H vibrations by comparison to those reported for [Ln(Cp tt ) 2 (μ-BH 4 )] 2 (Ln = La, Ce, Sm). 16b,19 As expected the ATR-IR spectra of 2-La and 2-Ce are nearly superimposable, but 1-Ln fall into two distinct pairs, with analogous spectra for 1-La and 1-Sm, and for 1-Ce and 1-Nd; this observation is curious given that the single crystal XRD data indicate that 1-Ln are all structurally analogous in the solid state (see below).

NMR spectroscopy
1 H NMR spectra were recorded from −350 to +350 ppm for 1-Ln, 2-Ln and 3-Ce, though paramagnetic shifts were relatively small for Ce, Nd and Sm analogues ( Table 1). Two signals were observed in all spectra in a ratio of 36 : 4; these correspond to the t Bu groups and the Htp ring protons, respectively. Due to restricted rotation of the Htp rings in 2-La, we observed two t Bu group resonances at 298 K; VT 1 H NMR spectra in C 7 D 8 from 218-318 K showed that these two signals coalesced at 318 K, and exhibited greater separation at lower temperatures with an approximate rotational energy barrier of 31 (7)  The paramagnetism of 1-Ce, 1-Nd, 1-Sm, 2-Ce and 3-Ce precluded assignment of their 13 C{ 1 H} NMR spectra, however, for diamagnetic 1-La and 2-La these could be interpreted: for 1-La the expected two t Bu group resonances were observed at 34.50 and 37.02 ppm and the two Htp ring carbon environments were located at 125.01 and 178.08 ppm. The signals for carbon atoms on the Htp ring and t Bu groups are doublets from coupling with 31 P ( 1 J PC = 59.7 Hz; 2 J PC = 15.7 Hz; 3 J PC = 6.9 Hz); similar coupling constants were previously seen for [Pb(Dtp) 2 ] ( 1 J PC = 45.8 Hz; 2 J PC = 16.2 and 3. 4 Hz) and [Pb(Dsp) 2 ] ( 1 J PC = 66.0 Hz; 2 J PC = 3.0 Hz; 3 J PC = 6.0 Hz). 12 The Htp ring was only observed by 31 P{ 1 H} NMR spectroscopy for diamagnetic 1-La (δ P : 105.65 ppm) and 2-La (δ P : 96.49 ppm), however, for paramagnetic 1-Ln, 2-Ce and 3-Ce only diamagnetic impurities and (Htp) 2 were observed at ca. −62 ppm and −24 ppm, respectively.

Crystallography
The solid-state structures of 1-Ln, 2-Ln and 3-Ce were determined by single crystal XRD (1-Ce, 2-Ce and 3-Ce are depicted in Fig. 1, selected bond distances and angles are compiled in Table 2; see ESI † for additional crystallographic data). Due to the poor data quality for 1-Nd metrical parameters are not included, however the data is of sufficient quality to provide connectivity. All Ln 3+ cations in 1-Ln, 2-Ln and 3-Ce are capped with two η 5 -Htp ligands and have two equatorial BH 4 − anions. For the dinuclear 1-Ln series the BH 4 − anions bridge   two Ln 3+ cations, whereas for 2-Ln and 3-Ce these bridge one Ln 3+ and one K + cation.

Magnetism
We were unable to collect reliable EPR spectroscopic data for paramagnetic 1-Ln, 2-Ln and 3-Ce as powders or as frozen toluene solutions. Multiple measurements were attempted at both X and K-band frequencies to probe the electronic structure and the presence of exchange interactions. The sample stability and weak signal, including issues with grinding polycrystalline samples, the presence of inequivalent Ln 3+ sites and multiple hyperfine splittings (by 31 P, 10 B and 11 B nuclei), proved problematic for obtaining reliable and reproducible spectra and as such discussion of these data is omitted. However, variable temperature DC magnetic measurements were recorded for all paramagnetic complexes as solids suspended in eicosane by SQUID magnetometry with 0.1 T applied magnetic field, or in solution at 298 K by the Evans method 21 (see ESI † for full details). Relatively small discrepancies between solid and solution χ M T values for all samples are attributed to weighing errors and the estimation of diamagnetic corrections. The magnetic susceptibility-temperature product (χ M T ) for solid 1-Ce at 300 K was 1.32 cm 3 mol −1 K, which is similar to the solution value of 1.47 cm 3 mol −1 K but lower than the expected value for a dinuclear Ce 3+ complex of 1.60 cm 3 mol −1 K (S = 1/2, L = 3, 2 F 5/2 ). 1 For solid 1-Nd the 300 K χ M T value of 2.68 cm 3 mol −1 K is higher than the solution value of 2.02 cm 3 mol −1 K but is lower than the calculated value of 3.26 cm 3 mol −1 K for two non-interacting Nd 3+ ions (S = 3/2, L = 6, 4 I 9/2 ). In contrast, the χ M T product for 1-Sm at Table 2 Selected bond distances and angles of 1-Ln, 2-Ln and 3-Ce. Data for 1-Nd excluded due to their low quality

Dalton Transactions Paper
This journal is © The Royal Society of Chemistry 2020 300 K (0.61 cm 3 mol −1 K) is consistent with the solution value (0.71 cm 3 mol −1 K), but both are higher than the expected value of a dinuclear Sm 3+ complex (0.18 cm 3 mol −1 K; S = 5/2, L = 5, 6 H 5/2 ). Experimentally obtained χ M T for Sm 3+ complexes are consistently higher than free-ion values due to the mixing of low lying J multiplets. 1 The χ M T values at 300 K for solid 2-Ce (1.16 cm 3 mol −1 K) and 3-Ce (0.63 cm 3 mol −1 K) are similar to solution moments (1.55 and 0.68 cm 3 mol −1 K, respectively). A dinuclear formulation was used to calculate the moment of 2-Ce and a single Ce 3+ centre was used in the calculation for polymeric 3-Ce; as a result the moment of 2-Ce is similar to 1-Ce, and that of 3-Ce is approximately half that of 1-Ce. Again these values are lower than the predicted free-ion χ M T values (0.80 cm 3 mol −1 K for a single Ce 3+ ion; S = 1/2, L = 3, 2 F 5/2 ), consistent with the results obtained for 1-Ln. The low temperature magnetisation measurements of 1-Ce, 1-Nd and 1-Sm measured between 0-7 T at temperatures of 2 K and 4 K did not reach magnetic saturation, which can be ascribed to the large magnetic anisotropy of the system and/or to the presence of low-lying excited states. 22 In contrast, the isothermal magnetisation curves for 2-Ce and 3-Ce at 2 K exhibit near-saturation at 7 T.

CASSCF-SO calculations
In order to probe the variation in the crystal field (CF) environments of 1-Ln, 2-Ln and 3-Ce, complete active-space self-consistent field spin-orbit (CASSCF-SO) calculations were performed with Molcas 8.0 23 using the X-ray crystal coordinates. For polymeric 3-Ce, a molecular fragment containing two K + ions and a single Ce 3+ ion was used, completing the ligand coordination sphere around each K + ; see ESI † for full details. The CF splitting of the ground J Russell-Saunders terms are presented in Tables S5-S12. † The theoretically predicted magnetic properties were compared to the experimentally determined susceptibility and magnetisation curves. For 1-Ce, 1-Sm and 3-Ce, ab initio calculations are in good agreement with experimental data, whilst for all other paramagnetic complexes the experimentally obtained values are all lower than those calculated, as is also seen by Curie law comparison of the room temperature χ M T. The strong CASSCF agreement for 1-Sm, which was not predicted by the Curie Law, arises from temperature independent paramagnetism owing to large CF splitting and a weakly magnetic ground doublet state. 1 For the 1-Ln series, the two inequivalent Ln 3+ ion sites result in unique CF splitting of the ground free ion J multiplets (Tables S5-S10 †). These sites can be distinguished by the orientation of the P atoms in the Htp rings with respect to the bridging BH 4 − moieties: Ln(1) refers to the smaller Htp cent ⋯Ln⋯Htp cent angle where the P atoms are positioned relatively close to BH 4 − , whilst Ln(2) refers to the CF environment with the larger Htp cent ⋯Ln⋯Htp cent angle where the P atoms are far away from BH 4 − . Consistently, Ln(2) has a larger overall CF splitting of the m J sublevels, which follows the trend with linearity of the Htp cent ⋯Ln⋯Htp cent angle. Complex 1-Ce reveals the most dramatic contrast between the two sites, with Ce(2) having ca. 50% larger CF splitting (892 cm −1 ) than Ce(1) (626 cm −1 ); this contrast is also observed for Nd and Sm but to a much smaller extent (<20% difference in CF splitting). As there is almost no variation in the Ln⋯Htp cent distances of the two sites, the CF splitting behaviour is likely a result of the coordination angle of the Htp ligands and influenced by the skewed electron density resulting from the position of the P atom within the Htp ring. There is ca. 0.1 Å difference between the Ln⋯B distances of each site from the bridging BH 4 − moieties across all variants. Here, the longer Ln⋯B interaction is always Ln (2), with the larger Htp cent ⋯Ln⋯Htp cent angle. For the Ce Htp complexes (Tables S5, S6, S11 and S12 †), the coordination environment of Ce(1) from 1-Ce shows a similar bonding motif of the Htp ligands with 2-Ce and 3-Ce (   (Tables S13 and S14 †). 25 This system has two unique Ce 3+ coordination environments, albeit with minimal differences in metrical parameters (Cp tt cent ⋯Ln⋯Cp tt cent angles of 119.2 and 119.6°, Ce⋯Cp tt cent distances of 2.53 and 2.54 Å, and Ce⋯B distances of 2.93(2) Å). Interestingly, the coordination environment of 4-Ce is similar to Ce(1) from 1-Ce (with slightly shorter Ce⋯Cp tt cent distances due to the less sterically demanding Cp tt ligand), though the C 1 -positions on both {Ln(Cp tt ) 2 } + moieties in 4-Ce are always the closest ring carbon atoms to the BH 4 − units, which contrasts the alternating coordination observed in 1-Ce. This is likely a result of minor steric differences between the Cp tt and Htp ligands. The CF splitting of the ground J multiplet is similar in these two systems (626 cm −1 for Ce(1) in 1-Ce vs. 642 and 677 cm −1 for 4-Ce; Tables S5, S13 and S14 †), therefore, the inclusion of the P heteroatom appears to have minimal effect on the overall CF splitting and ground state stabilisation. However, since the Ce⋯Htp bond lengths are longer, Htp must have a more significant influence on the CF than Cp tt in order to arrive at the same magnitude of CF splitting. Whether this is due to P substitution causing an overall increase in ring electron density, or a result of localisation of the electron charge, is unclear.  2 indicated that Htp has a relatively large influence on the CF compared to Cp tt , which we attribute to either charge localisation or increased ring electron density in Htp due to P substitution. We envisage that such subtle differences in phospholyl and Cp R Ln 3+ chemistry could be of importance in the construction of geometrically precise f-element complexes, and more specifically for CF engineering in Ln SMM design.

Materials and methods
All manipulations were carried out using standard Schlenk line and glove box techniques under dry argon. Solvents were passed through columns containing alumina or were dried by refluxing over K, and were stored over K mirrors or 4 Å molecular sieves (THF) and degassed before use. For NMR spectroscopy, C 6 D 6 and C 7 D 8 were dried by refluxing over K. NMR solvents were degassed by three freeze-pump-thaw cycles, and vacuum-transferred before use. [Ln(BH 4 ) 3 (THF) 4 ] (Ln = La, Ce, Nd, Sm) 16 and K(Htp) 17 were prepared according to literature methods and KBH 4 was used as received. 1 H (400 and 500 MHz), 13 C{ 1 H} (100 and 125 MHz), 31 P{ 1 H} (162 and 202 MHz), and 11 B{ 1 H} (128 and 160 MHz) NMR spectra were obtained on Avance III 400 or 500 MHz spectrometers at 298 K. EPR spectroscopy measurements were performed at X-band using a Bruker super-high-Q resonator, and at K-band with a standard cavity, both attached to a Bruker EMX bridge, on powder samples in quartz EPR tubes that were flame-sealed under vacuum. UV-vis-NIR spectroscopy was performed on samples in Youngs tap-appended 10 mm path length quartz cuvettes on an Agilent Technologies Cary Series UV-vis-NIR spectrophotometer at 175-3300 nm. ATR-Fourier transform infrared (ATR-IR) spectra were recorded as microcrystalline powders using a Bruker Tensor 27 spectrometer. Elemental analyses were performed by Mrs Anne Davies and Mr Martin Jennings at The University of Manchester School of Chemistry Microanalysis Service, Manchester, UK. Elemental analysis results for 1-Ln and 2-Ln consistently gave low carbon values, which we assign to carbide formation as other analytical techniques were indicative of bulk purity. General synthetic procedures for 1-Ln, 2-Ln and 3-Ce are given below; full details are in the ESI. † General procedure for synthesis of 1-Ln Di-n-butyl ether (20 mL) was added to a pre-cooled (−78°C) Rotaflow tap-appended ampoule containing [Ln(BH 4 ) 3 (THF) 3.5 ] (2 mmol) and K(Htp) (4 mmol). The reaction mixture was refluxed for 16 hours, allowed to settle and filtered. The solution was concentrated to ca. 2 mL and stored at −25°C overnight to afford 1-Ln (Ln = La, Ce, Nd, Sm).