The metals of the s block of the periodic table are well known to be exceptional electron donors, and the vast majority of their molecular complexes therefore contain these metals in their fully oxidized form. Low-valent main-group compounds have recently become desirable synthetic targets owing to their interesting reactivities, sometimes on a par with those of transition-metal complexes. In this work, we used stabilizing cyclic (alkyl)(amino)carbene ligands to isolate and characterize the first neutral compounds that contain a zero-valent s-block metal, beryllium. These brightly coloured complexes display very short beryllium–carbon bond lengths and linear beryllium coordination geometries, indicative of strong multiple Be–C bonding. Structural, spectroscopic and theoretical results show that the complexes adopt a closed-shell singlet configuration with a Be(0) metal centre. The surprising stability of the molecule can be ascribed to an unusually strong three-centre two-electron π bond across the C–Be–C unit.
1-(2,6-diisopropylphenyl)-3,3,5,5-tetramethylpyrrolidine-2-ylidene beryllium dichloride
2-(2,6-diisopropylphenyl)-3,3-dimethyl-2-azaspiro[4,5]decan-1-ylidene beryllium dichloride
(1-(2,6-diisopropylphenyl)-3,3,5,5-tetramethylpyrrolidine-2-ylidene) carbon dioxide adduct
Molecular complexes of metal atoms in their zero oxidation state allow chemists to work with soluble and discrete forms of the metals in their elemental—and usually electron-rich—electronic state. Although zero-valent monoatomic transition-metal compounds, LnTM0 (L, neutral donor ligand (for example, carbonyl, phosphine, N-heterocyclic carbene); TM, transition-metal element), are well known and have proved their usefulness as both reagents and catalysts in countless applications throughout academia and the pharmaceutical, agricultural and petrochemical industries1,2, analogous main-group metal complexes remain a rarity owing to the difficulty of stabilizing such highly reactive species. Over the past decade, however, stable low-valent p-block compounds have moved from curiosities to highly sought-after synthetic targets and exhibit a rich chemistry that is traditionally the preserve of transition-metal compounds, as well as some unique bond-activating reactivity3,4,5. The use of strongly σ-donating and sterically shielding N-heterocyclic carbenes (NHCs) has even enabled the isolation of a number of neutral dinuclear group 13, 14 and 15 (L→E02←L) complexes (E, main group element) in their zero oxidation state6, such as the diboryne I, ditin II or diphosphorus complexes III presented in Fig. 1a7,8,9. Monoatomic LnE0 main-group species, however, remain limited to a couple of examples, such as compound IV, which features silicon or germanium centres (Fig. 1a), and heavily rely on the stabilizing effect of the more π-acidic cyclic (alkyl)(amino)carbene (CAAC) donors10,11.
Although there are a handful of reports of unusual ‘alkalide’ salts that contain alkali metals formally in their –1 oxidation state12, the only reports of neutral zero-valent s-block compounds to date have come from the computational community and pertain to hypothetical Be0 and Mg0 species13,14. In light of the ever-increasing popularity of the metal-bonded Mg+1 compounds, first reported by Jones, Stasch and co-workers in 200715 as selective reducing agents16, such species are highly desirable synthetic targets. Whereas Mg0 compounds have been predicted to be synthetically unviable13, the combined smaller size and greater electronegativity of beryllium, as well as its ability to form strong covalent bonds, should enable the synthesis of Be+1 or Be0 complexes under the right circumstances. Indeed, subvalent compounds of beryllium have been actively sought by a handful of researchers, thus far without success17,18. Computational work has identified NHC stabilization as providing perhaps the best chance to achieve this goal13,14; however, despite reports of a number of NHC–beryllium complexes19,20,21,22, none have shown the ability to stabilize the atom in an oxidation state other than +2.
The aforementioned computational studies on the stabilization of Be0 all relied on rather substantial π backdonation from the electron-rich beryllium centre to the π systems of the carbene ligands, enabled by the efficient overlap of the p orbitals on beryllium and the carbene carbon atoms of the stabilizing ligands13,14. With the limited π-accepting ability of NHCs23 and the successful use of CAACs for the stabilization of a variety of both low-valent transition-metal and p-block elements in mind24, we expected a similar stabilization of subvalent beryllium by making use of Be–CAAC π bonding.
Results and discussion
Beryllium–carbene adducts [Be(RL)Cl2] (1 with MeL = 1-(2,6-diisopropylphenyl)-3,3,5,5-tetramethylpyrrolidine-2-ylidene and 2 with CyL = 2-(2,6-diisopropylphenyl)-3,3-dimethyl-2-azaspiro[4,5]decan-1-ylidene (Fig. 1b)) were prepared by the addition of either the free ligand MeL to BeCl2 (1), analogous to the synthesis of other carbene–BeCl2 adducts21, or by in situ deprotonation of the iminium ligand precursor CyLHBF4 and the addition of BeCl2 (2). The adducts, characterized by their 9Be NMR chemical shifts at 12.9 ppm (1) and 12.8 ppm (2), were obtained in fair-to-good yields (1, 79%; 2, 43%). Single crystals of 1 and 2 were additionally analysed by single-crystal X-ray diffractometry (Supplementary Fig. 17). Cyclic voltammetry (CV) experiments carried out on 1 and 2 indicated the presence of a chemically irreversible reduction wave at –1.79 V (versus ferrocene/ferrocenium (Fc/Fc+)) for 2 (Supplementary Fig. 15), whereas two distinct reduction waves were observed at –1.79 and –2.96 V for 1 (Supplementary Fig. 14), which hints at the possibility of chemical reduction to BeI or Be0. Unfortunately, attempts at chemical reduction of pure samples of 1 and 2 with a variety of reducing agents only led to intractable mixtures of products.
When the chemical reduction of 1 was, instead, carried out in the presence of another equivalent of MeL a new broad 9Be NMR resonance was observed at 32 ppm, well downfield of the range of other known two-coordinate beryllium species25,26, but still more than 10 ppm upfield from the beryllium bis(diazaborolyl)complex reported by our group at 44 ppm (ref. 27). A computational model of the neutral [Be(MeL)2] complex, geometrically similar to the bis(CAAC) compounds reported for many transition metals and optimized at the M06/def2SVP level, was predicted to present the same 9Be NMR frequency at 32 ppm. The identity of this compound as [Be(MeL)2] (3) was confirmed by crystallographic analysis (Fig. 1d). Attempts to synthesize the analogous complex [Be(CyL)2] were complicated by the need to deprotonate the ligand precursor salt in situ; however, the reduction of 2 in the presence of MeL exclusively yielded the mixed-carbene complex [Be(MeL)(CyL)] (4) in 24% yield, as confirmed by X-ray crystallographic analysis (Fig. 1d) and the presence of two inequivalent 2,6-diisopropylphenyl group resonances in the 1H NMR spectrum in a 1:1 ratio. The absence of ligand redistribution during this synthesis confirms the strong covalent bonding between CAAC and beryllium, unlike bonding in the heavier group 2 congeners, which is primarily ionic and frequently leads to ligand redistribution during synthesis28. To the best of our knowledge 4 is thus the first heteroleptic bis(carbene) main-group complex; outside the main group, examples are limited to cationic group 11 complexes29. Both 3 and 4 were isolated as purple crystalline solids presenting ultraviolet–visible (UV-vis) absorption maxima at 575 nm (Supplementary Fig. 13). Furthermore, CV carried out on 3 indicated a facile irreversible oxidation at –1.57 V (versus Fc/Fc+) (Supplementary Fig. 16).
The central C–Be–C motif in 3 and 4 is linear, with the CAAC ligand frameworks coplanar (dihedral angles <2°) and arranged in a trans configuration. The Be–C bonds in 3 (1.664(2) Å and 1.659(5) Å) and 4 (1.659(4) Å and 1.657(4) Å) are substantially shorter than the Be–C bonds in 1 (1.779(3) Å) and 2 (1.791(2) Å) and about 0.04 Å shorter than those in the only other solid-state structure of a two-coordinate beryllium complex that presents a Be–C bond, [(2,6-Mes2C6H3)BeN(SiMe3)2] (Mes, 2,4,6-trimethylphenyl) (1.700(4) Å)25. Within the limited sample of crystallographically characterized two-coordinate Be2+ complexes25,26,27,30, the observation of shortened bonds and linear geometry around beryllium as well as computational investigations suggest a partial double-bond character. Although such shortening is also to be expected with a reduction in coordination number, the density functional theory (DFT)-optimized structure of a dicationic two-coordinate [Be(CAACModel)2]2+ complex, in which all the substituents on the ligands are replaced by hydrogen atoms, presents Be–C bonds (1.738 Å) only slightly shorter than those in 1 (Supplementary Fig. 19). The most probable explanation for the contraction of these bonds is therefore the formation of strong π interactions across the C–Be–C unit, which would also account for its linearity and the coplanarity of the CAAC ligands.
Surprisingly, complex 3 did not display any reactivity towards H2, boranes or borohydrides. It was also unreactive towards sterically demanding alcohols, but it decomposed in the presence of equimolar amounts of stronger Brønsted acids. A reaction with 2 equiv. of elemental selenium, however, led to the formation of the selenone, MeL = Se (5), as the only isolable CAAC-containing product (Fig. 1c). Furthermore, a reaction with CO2 yielded the zwitterionic species [(MeL)CO2] (6) as well as a black precipitate, which is, presumably, elemental beryllium (Fig. 1c). Both 5 and 6 have been reported as products of the reactions of neutral MeL with selenium31 and CyL with CO2 (ref. 32), respectively, which led us, at this point, to suggest that 3 may be described as a bis(CAAC) adduct of elemental beryllium.
To clarify the electronic structure of 3, DFT calculations were performed at various levels of theory. The results showed a large spectrum of S0 – S1 energy differences (0.1–5.3 kcal mol–1) (Supplementary Figs 21–23). As DFT is known, however, to be unreliable at modelling biradicals because it formulates the wavefunction as a single determinant33, the electronic structure of 3 was assessed by multireference CASSCF[4,4]/def2-TZVPP calculations. These calculations yielded configuration interaction coefficients of 0.89 and –0.46 for the singlet closed shell and the second excited singlet state, respectively. The coefficient of the singlet biradical state (8 × 10–7) is negligible, unlike that provided by CASSF[2,2] calculations for the biradicaloid bis(MeL) germylone complex IV (–0.28 (Fig. 1a))11. Also, the simulation of UV-vis data for the singlet biradical configuration yielded a main absorption at 438 nm (Supplementary Fig. 31 and Supplementary Table 3), whereas computation of the singlet closed-shell configuration provided an absorption at 552 nm that corresponds to the highest occupied molecular orbital to lowest unoccupied molecular orbital (HOMO → LUMO) excitation, in close agreement with the experimental value of 575 nm.
Figure 2a shows two relatively low-lying orbitals (HOMO–7 and HOMO–8) responsible for the σ bonding between Be and the MeL ligands, as well as a HOMO that corresponds to the π bond of the ground-state singlet closed shell of 3 at the BP86/def2-TZVPP level, the structure of which is in good agreement with crystallographic data (Supplementary Fig. 26). The nature of the C–Be–C bonding was analysed by energy decomposition analysis combined with natural orbitals for chemical valence (EDA-NOCV) using Be and [MeL···MeL] as interacting fragments (Supplementary Table 1). The results of these calculations, based on the rules proposed by Frenking and co-workers for EDA-NOCV bonding analysis34,35, suggest that the Be–C bonding in 3 is best described as a combination of donor–acceptor interactions [MeL⇄Be(0)⇄MeL] between ground-state singlet MeL ligands and Be(0) in a 1s22s02p2 electronic configuration, which results in a three-centre two-electron π-bond stretching over the C–Be–C core (Fig. 2c and Supplementary Table 1). Furthermore, the stabilization that arises from π backdonation from Be to the CAAC ligands (ΔE1π┴ = –148.6 kcal mol–1) was found to predominate significantly over that from the σ donation from CAAC to Be (ΔE2σ(+,–) + ΔE3σ(+,+) = –71.6 kcal mol–1) (Fig. 2b and Supplementary Table 1). Although unusual, this kind of bonding situation, in which stabilization through π backdonation exceeds stabilization through σ donation, has been calculated previously for the similarly linear donor–acceptor complex [OC⇄B⇄CO]– (ref. 36) and is in agreement with the strong π-acceptor properties of CAAC ligands. Furthermore, the strength of this π backdonation may explain the relatively unreactive nature of 3 towards hydrides and even alcohols.
In summary, complexes 3 and 4 represent not only the first subvalent beryllium complexes, but also the first isolated s-block singlet closed-shell donor–acceptor complexes in which the metal centre is in the zero oxidation state. This bonding situation is favoured by the strong σ-donor and π-acceptor properties of the CAAC ligands. As the chemistry of zero-valent main-group compounds is still in its infancy, the field remains wide open for new strategies for the stabilization of unusual bonding modes and oxidation states.
All the reactions were performed under an atmosphere of dry argon using standard Schlenk or glovebox techniques. BeCl2 was purchased from Sigma-Aldrich and used as received. All the solvents were purified by distillation using the appropriate drying agents, deoxygenated using three freeze–pump–thaw cycles and stored over molecular sieves under dry argon prior to use. The deuterated solvents used for NMR spectroscopy were purchased from Cambridge Isotope Laboratories, deoxygenated by freeze–pump–thaw cycles and dried under an argon atmosphere over molecular sieves. The starting materials MeL (ref. 31), CyLHBF4 (ref. 37) and KC8 (ref. 38) were prepared by established methods.
1H, 9Be and 13C NMR spectroscopy data were obtained at ambient temperature using either a Bruker DRX-400 (operating at 400 MHz for 1H, 56 MHz for 9Be and 100 MHz for 13C) or a Bruker Avance 500 NMR spectrometer (operating at 500 MHz for 1H and 125 MHz for 13C). 1H NMR spectra were referenced via residual proton resonances of C6D6 (1H, 7.16 ppm), THF-d8 (1H, 3.58, 1.72 ppm) or CD2Cl2 (1H, 5.23 ppm). 13C NMR spectra were referenced to C6D6 (13C, 128.06 ppm), THF-d8 (13C, 67.21, 25.31 ppm) or CD2Cl2 (13C, 53.84 ppm). 9Be NMR signals were referenced to Be(H2O)42+ (sulfate salt). Elemental analyses were performed on an Elementar vario MICRO cube elemental analyser. UV-vis spectra were recorded on a JASCO V-660 UV-vis spectrometer using standard quartz-glass cuvettes. Electron paramagnetic resonance (EPR) measurements at the X band (9.38 GHz) were carried out at room temperature and at 70 K using a Bruker ELEXSYS E580 EPR spectrometer equipped with an Oxford Instruments helium cryostat (ESR 900) and a MercuryiTC temperature controller.
CV experiments were performed using a Gamry Instruments Reference 600 potentiostat. A standard three-electrode-cell configuration was employed using a platinum-disk working electrode, a platinum-wire counter electrode and a silver wire separated by a Vycor tip, which served as the reference electrode. Formal redox potentials were referenced to the Fc/Fc+ redox couple. Tetra-n-butylammonium hexafluorophosphate was employed as the supporting electrolyte. Compensation for resistive losses (iR drop) was employed for all the measurements.
The crystal data of 1–4 were collected on a Bruker X8-APEX II diffractometer with a CCD (charge-coupled device) area detector and multilayer mirror-monochromated Mo Kα radiation. The structures were solved using direct methods, refined with the ShelX39 software package and expanded using Fourier techniques. All non-hydrogen atoms were refined anisotropically. Hydrogen atoms were assigned to idealized geometric positions and included in structure-factor calculations.
Calculations of 9Be NMR chemical shifts were carried out via the GIAO method available in the Gaussian 09 software package40. The isotropic shielding factors were calculated at the B3LYP41,42,43/6-311+G(2d,p)44 level and scaled against a shielding factor of 108.96, as previously described45. The input geometry for the 9Be NMR prediction of 3 was based on a geometry optimized at the M0646/def2-SVP47 level of theory. DFT calculations on 3 were performed at B3LYP, M05-2X48 and TPSS49 levels using the def2-TZVPP50 basis set. All the complexes were characterized as minima by the calculation of vibrational frequencies. The calculations at the CASSCF[4,4]51 level using the basis set def2-TZVPP were carried out over the optimized geometries of 3 at the B3LYP/def2-TZVPP level. The EDA-NOCV calculations52,53 were carried out at the BP86 level in conjunction with a triple-ζ-quality basis set TZ2P54 in which relativistic effects are considered with the ZORA approximation55 over the singlet closed-shell optimized geometry of 3 at the BP86/def2-TZVPP level. The rule suggested by Frenking and co-workers (the EDA-NOCV calculations that give the weakest ΔEorb term indicate the best choice of the bonding description)34,35 was applied because the electronic structure of the molecule requires the least alteration of the electronic-charge distribution. UV-vis spectra were calculated at the CAM-B3LYP level56 using the def2-SVP basis set, including the PCM57 solvent model.
Crystallographic data have been deposited with the Cambridge Crystallographic Data Centre as supplementary publication numbers 1439665 (1), 1439666 (2), 1439667 (3) and 1439668 (4). These data can be obtained free of charge from www.ccdc.cam.ac.uk/data_request/cif.
Financial support from the Julius-Maximilians-Universität Würzburg (H.B.) and the Alexander von Humboldt Foundation (postdoctoral fellowship to M.A.) is gratefully acknowledged. We also thank G. Frenking for helpful discussions regarding the computational analysis.