1-Titanacyclobuta-2,3-diene – an elusive four-membered cyclic allene

The synthesis and characterisation of a 1-titanacyclobuta-2,3-diene complex, an organometallic analog of elusive 1,2-cyclobutadiene, is presented.


Reaction of 2 with acetophenone to E/Z isomere mixture of (4-phenylpent-3en-1-yne-1,3-diyl)bis(trimethylsilane) (5E and 5Z)
pentane, r.t. Compound 2 (0.20 mmol, 0.100 g) was dissolved in pentane (2 mL) and acetophenone (0.20 mmol, 24 mg) was added. After 16 hours of stirring at room temperature an NMR sample (E/Z ratio 0.8 : 1) was taken and then the solvent of the reaction mixture was removed in vacuo. Next, the orange residue was suspended in pentane (5 mL) and Silica Gel was added. After removing the solvent, the E and Z isomers (5E and 5Z) were separated by column chromatography (hexane/ethyl acetate 20 : 1).   Compound 2 (0.20 mmol, 0.100 g) was dissolved in pentane (2 mL) and benzaldehyde (0.20 mmol, 22 mg) was added. After 16 hours of stirring at room temperature the solvent of the reaction mixture was removed in vacuo. Next, the orange residue was suspended in pentane (5 mL) and was filtered over Silica Gel. After removing the solvent, a mixture of E/Z isomers (6, E/Z ratio 0.8 : 1) was separated as colourless liquid. Yield: 27 mg, 50 %,. Compound 2 (0.20 mmol, 0.100 g) was dissolved in pentane (8 mL) at ambient temperature and then cooled to -78 °C. To this solution, neat acetone (0.2 mmol, 0.012 g) was added at this temperature. The temperature was slowly raised to -15°C, where a colour gradient from red to petrol was obtained within 4 hours. This turbid reaction mixture was dried in vacuo at -20 °C for 4 h, the residue was extracted/filtered with pentane (2 mL) at -20 °C. This filtrate was concentrated to approximately 1 mL and was slow cooled to -78 °C. The resulting dark petrol coloured residue was identified as 7 by low temperature NMR and IR spectroscopy. This complex is only stable at temperatures below -10 °C.  Figure S20, Figure S21). Compound 2 (0.05 mmol, 0.025 g) was dissolved in [D 8 ]toluene (0.7 mL) at ambient temperature and then cooled to -78 °C. To this solution, neat benzaldehyde (0.05 mmol, 0.005 g) was added at this temperature. The sample was brought to -10 °C in the probe of the NMR spectrometer and the reaction sequence was monitored at that temperature for 6 hours while recording a series of NMR spectra. The conversion proved slow enough to characterise the intermediate 8. 1 Table S1: Crystallographic details of 2 and 3.

3
Chem. Formula   Figure S7: Molecular structure of compound 3. Thermal ellipsoids correspond to 30% probability. Hydrogen atoms are omitted for clarity.  Compound  Compound

Compound 3
[a] = two best fitting descriptions are presented: Values taken from literature [10] (C sp3 -C sp3 1.54; C sp3 -C sp2 1.50; C sp2 -C sp 1.42; C sp -C sp 1.38; C sp2 = C sp2 1.34; C C sp2 = C sp 1.31; C sp = C sp 1.28 S17 Compound          . For this spectrum we carried out a low temperature NMR experiment, were the complex 2 was dissolved in [D 8 ]toluene at ambient temperature; the resulting red solution was then cooled to -50 °C and an excess of acetone was added at this temperature. The sample was positioned in the cooled NMR spectrometer and the reaction was monitored via 1 H NMR spectra. This spectrum was recorded after approximately 3 hours reaction time.

Assignment of the most important vibrations
In this chapter the experimental IR and Raman spectra (black) with their respective calculated uncorrected vibration spectra (red) are presented. The calculated spectra were taken from the frequency analyses with BP86/LANL2DZ/TZVP level of theory.  Figure S31: General carbon atom assignment for compounds 3-6.    Figure S40: Experimental IR spectrum of isomer mixture of 6 (blue), calculated spectra for E-isomer (green) and Z-isomer (red).  Figure S41: Representation of a selection of the IR spectra of 6 with the most noticeable differences of the E/Z-isomers which can be assigned to the CH in plane vibrations of the phenyl substituents which are mixed with CC stretching vibrations. Red line represents the calculated spectrum of the Z isomer, green for the E isomer and the blue spectrum represents the experimental spectrum, which clearly shows the resulting product as a mixture of E and Z isomer as confirmed by NMR spectroscopy.  Figure S42: Experimental Raman spectra of isomer mixture of 6 (blue), calculated spectra for E-isomer (green) and Z-isomer (red).   Figure S46: IR spectra of the intermediate species 7 immediately measured (red) and after 3 minutes exposure to air and ambient temperature (black). The black spectrum features a new characteristic vibration at 2115 cm -1 which might be assigned to the C1≡C2 stretch vibration of 4 which is formed due to the ambient temperature measurement.

Computational Details
All calculations were carried out with the Gaussian 09 package of molecular orbital programs. 6 In a first step we carried out an optimisation test with real-size molecule 2, in this study we compared the Methods BP86, 12 B3LYP 12,13 and PBE1PBE 14 as well as the basis sets def2-TZVP, 15 {TZVP(C, H, Si); 16 LANL2DZ(Ti) 17 } and aug-cc-pvdz. 18 The main result is that pure density functional (DF) BP86 in combination with the LANL2DZ basis set and corresponding effective core potential (ECP) at Ti and the TZVP basis set on all other atoms (notation BP86/LANL2DZ/TZVP) is clearly the best combination for the metallacyclic systems, both in terms of performance and HF energy (see Table S9). Therefore, if not further mentioned the energies and discussed results were performed with this procedure. Vibrational frequencies were also computed, to include zero-point vibrational energies in thermodynamic parameters and to characterise all structures as minima on the potential energy surface. In addition, we used these results to assign the experimental IR and RAMAN spectra and to superimpose the experimental and calculated vibration spectra (see above). NBO analyses were performed using NBO 6.0. 19 QT-AIM and ELF calculations were performed using MultiWfn 3.5. 20

Thermochemistry
For basic thermochemistry, molecular structures were optimised using the pure density functional (DF) BP86 in combination with the LANL2DZ basis set and corresponding ECP at Ti and the TZVP basis set on all other atoms (notation BP86/LANL2DZ/TZVP). All optimised structures were confirmed as minima by frequency analyses.  Figure S47: Calculated Gibbs free energies of isodesmic titanocene reactions. [t] Energy given for the more stable triplet state.

MO and DFT studies of rac-(ebthi)TiC 3 (SiMe 3 ) 2 (2)
To obtain a better understanding of the bonding situation in titana-cyclobutadiene 2, several singlepoint calculations were performed: firstly, the Kohn-Sham (KS) wave function was recalculated using the pure DF BP86 in conjunction with the def2-TZVP basis on all atoms; secondly, a hybrid DF was employed (B3LYP 12a,13 /def2-TZVP); and lastly, the canonical MOs were calculated at the HF/def2-TZVP level of theory. All (KS) wave functions were tested with respect to RHF/UHF or RKS/UKS instabilities, in order to analyse the biradical character of Ti complex 2. While the KS wave function based on the pure DF (BP86) showed no instabilities, the hybrid DF (B3LYP) and HF solution exhibited a low-lying, "broken-symmetry" open-shell singlet state. This kind of behaviour is often observed if the biradical character is not too large, 22 since part of the non-dynamic correlation is treated by the exchange-correlation functional of the (pure) density functional. Mixing in exact exchange reduces the amount of non-dynamic correlation treated by the DF and thus the "broken-symmetry" solution becomes more stable. In consequence, structures that were optimised using the BP86 functional are expected to show good agreement with experimental structures (as verified by comparison with structural data from single-crystal X-ray diffraction, cf. Table S12). The electronic energy, however, should be considered as a rough approximation due to incorrect treatment of the non-dynamic correlation.

Biradical character
The "broken-symmetry" solution is not a true eigenfunction of the S 2 operator. In fact, it may be considered as a 50:50 mixture of the singlet and triplet state, if the overlap between the singly occupied orbitals and spin polarisation are small. 23,24,25 The actual singlet wave function can then be expressed in terms of a linear combination of two "broken-symmetry" wave functions where , are the singly occupied orbitals and the overline indicates β spin. Therefore, the open-+ shell singlet must be described by a multi-reference wave function.
In the "broken-symmetry" picture, the singly occupied orbitals and are, in principle, localised Hence, the multi-reference wave function expressed in terms of the canonical MOs is given by where the expansion coefficients are the square roots of the relative weight of each determinant. This type of multi-determinant open-shell singlet wave function can be obtained by the Complete Active Space (CAS) SCF method [21][22][23][24][25][26][27][28][29] and gives a qualitatively correct description of the electronic structure of a biradical. The biradical character can be evaluated as where a value of β = 1 indicates a "perfect" biradical with two electrons in two degenerate orbitals. 24,26 Smaller values indicate an increasing energy gap between HOMO and LUMO, and β → 0 indicates a closed-shell species. Consequently, the smallest active space to properly describe a biradical is a CAS(2,2) calculation (i.e. two electrons in two orbitals). In case of compound 2, we chose to include eight electrons in nine orbitals in the active space (comprising the formal π orbitals at the ligand and d-orbitals at Ti, vide infra), as these orbitals are energetically relatively closely spaced. The calculations show that the largest contributions to the multi-determinant wave function are the two determinants placing two electrons either in the formal HOMO (ϕ 4 ) or LUMO (ϕ 5 , Figure S49; β = 28 %).  Figure S49: Schematic depiction of the active orbitals of a CAS (8,9) calculation. Only contributions to the wave function with relative weights > 1 % are shown. The orbital localisation scheme indicates that one of the radical centres is localised at Ti, while the other is delocalised across the C 3 backbone.
Hence, compound 2 can be regarded as a biradical. The singlet state is calculated to be the ground state (ΔE S-T = −39.0 kJ/mol); i.e. the radical centres are antiferromagnetically coupled. The calculated exchange coupling constant 27 is The radical centres are localised at Ti and on the C 3 backbone of the ligand ( Figure S49, right). Therefore, the electronic structure can be understood as a complex between a formal Ti(III) fragment and an organic radical, whose "free" electrons are antiferromagnetically coupled. (This, by the way, is also indicated by the BS-B3LYP calculations; however, these results will not be discussed further as BS calculations predict unphysical spin polarisation.) Therefore, complex 2 should be EPR silent in its ground state. SiMe 3 Figure S50: Left: Schematic MO diagram of the formal π-type orbitals of the ligand system. There is a 4e3c bond in the z plane (blue) and a 3e3c bond in the x plane (red). Right: Lewis resonance scheme. The electrons in p z (p x ) orbitals are indicated in blue (red). Each π-bonding system is independently delocalized across the C 3 unit.

Lewis resonance scheme
Analysis of the ligand-centred orbitals shows that there are two formal π bonding systems. One of them is in-plane with the TiC 3 ring system and acts as σ donor (ϕ 1 , ϕ 3 , ϕ 6 , ϕ 8 ); the other is perpendicular to the ring and contains the delocalised radical centre (ϕ 2 , ϕ 4 , ϕ 5 , ϕ 7 ). The ligand could be considered as a propadienylide anion, i.e. the one-electron reduced congener of propynylidene, 28 which is corroborated by the fact that the ligand-centred orbitals in the complex nicely correspond to the MOs of the isolated ligand system ( Figure S50). Note that the electrons in both the formal π x and π z bonding systems are delocalized across the C 3 unit and that each of these π-bonding systems can be interpreted independently of the other, resulting in a variety of different Lewis resonance structures. Therefore, the leading resonance structures of complex 2 are proposed as depicted in Scheme S2.

NBO analysis
NBO analyses 19 of the BP86/def2-TZVP and CAS (8,9)/def2-TZVP densities led to similar results. The NBO routine found a double bond between both C1 and C2 as well as C2 and C3, in agreement with the Lewis structures in Scheme S2. It is worthy to note that both π-type NBOs are only occupied by approx. 1.6 electrons, indicating that the double bonds are delocalised. Furthermore, there are formally two Ti-C σ-bonds (Ti1-C1 and Ti1-C3) which are occupied by 1.5 electrons each. This can be attributed to both the delocalisation of the Ti-C bond (vide supra) as well as the biradical character, which is not well represented in the NBO picture. The calculated natural charge of the C 3 (SiMe 3 ) 2 ligand amounts to −0.39 e (CAS) or −0.64 e (BP86), which is in the expected range of a formally anionic ligand.

QT-AIM analysis
QT-AIM analysis 29 revealed two Ti-C "bond" paths (Ti1-C1 and Ti1-C3), in agreement with the Lewis resonance scheme (Scheme S2). Despite the short interatomic distance between Ti1 and C2, there is no strong bonding interaction between those atoms; on the contrary, a ring critical point is found near the centre of the TiC 3 ring system (i.e. there is a minimum in electron density within the ring plane). Moreover, the Laplacian of the electron density ∇ 2 r indicates that the Ti-C bonds are strongly polarised towards the C atoms, in agreement with their description as dative bonds. ( Figure S51). The densities obtained from CAS (8,9) and BP86 calculations are quite similar, indicating that the pure DFT method is suitable to approximately describe the electron density despite its single-determinant character ( Figure S52) Figure S52: Same as Figure S51, but density taken from BP86/def2-TZVP calculation.

Electron Localisation Function
The results from QT-AIM analysis are corroborated by ELF analysis ( Figure S53). There is no localised electron density in the valence region of C2 directed towards Ti1, whereas the bonding electrons between C1/C3 and Ti are localised in approx. the same region of space as indicated by the Laplacian of the electron density. It is worthy to note that there is no localised electron density around C2 S54 pointing away from Ti1 either, i.e. there is no lone pair of electrons at the central carbon atom. Consequently, the electronic structure of the C 3 scaffold is different from that of structurally related bent allenes, such as so-called "carbodicarbenes" ( Figure S54). 30 Figure S53: ELF plot of Ti complex 2 in the TiC 3 ring plane.

CAS computations of Cp 2 TiC 3 (SiMe 3 ) 2 (2Cp) and Cp* 2 TiC 3 (SiMe 3 ) (2Cp*)
CAS (8,9)/def2-TZVP computations were carried out in an analogous manner for the closely related Ti complexes Cp 2 TiC 3 (SiMe 3 ) 2 (2Cp) and Cp* 2 TiC 3 (SiMe 3 ) (2Cp*). A summary of the results is shown in Table S13. It should be pointed out that the singlet-triplet gap and therefore the biradical character greatly depend on the pyramidalisation of the carbon atoms C1 and C3 of the TiC 3 ring system. Since the coordination environment around C1/C3 is nearly planar in compound 2Cp* (most likely due to steric reasons), it displays the highest biradical character. This trend is agreement with previous computations. 31