Synthesis, molecular and electronic structure of a stacked half-sandwich dititanium complex incorporating a cyclic π-faced bridging ligand

Abstract

A thermally robust triple-decker complex [bis(η⁵-pentamethylcyclopentadienyltitanium)-μ-(η⁴:η⁴-1,2,4,5-tetrakis(trimethylsilyl)cyclohexa-1-4-diene-3,6-diyl)] (3) was obtained in 4% yield by thermolysing Cp*TiMe₃ in the presence bis(trimethylsilyl)acetylene (BTMSA). The solid-state structure of centrosymmetric 3 features rather long all C–C bonds in the nearly planar bridging ligand (1.4720(14)–1.4896(15) Å) and a short distance of its least-square plane to the titanium atoms (1.7381(5) Å). Computational results revealed the bonding of the central ligand to be accomplished through back-bonding of its two C=C bonds and through the simultaneous generation of two σ-Ti–C(H) bonds. Based on CASSCF and CASPT2 results, the molecule acquires several electronic configurations simultaneously, which hinders its representation by one single Lewis structure. Apart from being coordinated to the central ligand, the metal atoms are involved in a direct Ti–Ti bonding by the formation of one σ- and two δ-bonds between them. The bond order of this Ti–Ti overlap shows only a slight decrease upon electronic excitation. The presence of ionic contribution to the bonding of the central ligand is manifested by the charge −1.4e summed on the carbon atoms of the bridging ring. Based on computational results, the spin multiplicity of the ground state is singlet, while the first low lying excitation state is triplet. This is in agreement with the absence of EPR signal in either toluene solution and glass, and with slight downfield shifts of broadened ¹H NMR signals of SiMe₃ and Cp* methyl groups observed with increasing temperature.

---

Introduction

In the course of the past 30 years a large number of titanocene complexes incorporating internal alkynes have been studied.¹ In particular, the bis(trimethylsilyl)acetylene (BTMSA) complexes of titanocene and decamethyltitanocene [Cp′₂Ti(η²-BTMSA)] (Cp′ = η⁵-C₅H₅ (Cp) or η⁵-C₅Me₅ (Cp*)) have been explored extensively, in attempting to mimic reactions of titanocene moieties with a variety of reagents. These studies exerted a large impact on the development of titanocene chemistry and its applicability to organic synthesis.² On the other hand, half-sandwich titanocene-internal acetylene complexes received little notice, limited only to two types of BTMSA complexes: [Cp′Ti(η²-BTMSA)(μ-Cl)]₂ (Cp′ = η⁵-C₅H_5−nMe_n (n = 3–5)) (Chart 1, A)^3a and [Cp′′Ti(μ-η²:η²-BTMSA)₂MgCp′′] (Cp′′ = η⁵-C₅H_5−nMe_n (n = 0–2)) (Chart 1, B).^3b,c

Chart 1 Depiction of half-sandwich titanocene-internal acetylene complexes.

In efforts to extend this chemistry area we payed attention to the eventual exploitation of pentamethylcyclopentadienyltitanium intermediates, which apparently take part in the thermolysis of pentamethylcyclopentadienyltitanium trimethyl [(η⁵-C₅Me₅)TiMe₃] (1).⁴ The thermolysis of 1 at temperatures 96–120 °C was reported by Mena et al.⁵ to yield a titanium-methylidyne tetramer [{Cp*Ti(≡CH)}₄] (2) based on the Ti₄C₄ cubane-like structure. Methane elimination during the reaction was proven to originate from titanium-methyl groups, hence the methylidene and methylidyne intermediates (1a–1c) depicted schematically in Scheme 1 were suggested to take part in the reaction pathway to compound 2.⁵

Scheme 1 Thermolysis of 1 as suggested by Mena et al.⁵

The present work was initiated by our attempts to obtain new cyclopentadienyl-titanium derivatives by trapping the above intermediates with some internal acetylenes. These attempts resulted in the isolation of [bis(η⁵-pentamethylcyclopentadienyltitanium)-μ-(η⁴:η⁴-1,2,4,5-tetrakis(trimethylsilyl)cyclohexa-1-4-diene-3,6-diyl)] (3) only for the case of bis(trimethylsilyl)acetylene (BTMSA). It has to be noted that apart from [(CpV)₂-μ-(η⁶:η⁶-benzene)] and [(CpV)₂-μ-(η⁶:η⁶-mesitylene)],⁶ compound 3 is the only structurally-defined triple-decker complex in the first row of transition metals (Ti–Ni). The syntheses, molecular and electronic structure of 3 are the subject of the present communication.

Results and discussion

Adding 2 molar equivalents of 1-phenylpropyne, 1-(trimethylsilyl)propyne or bis(trimethylsilyl)acetylene (BTMSA) to a thermolysed m-xylene solution of 1 at 110–125 °C suppressed completely the formation of 2, which otherwise formed in good yields as a brown finely crystalline powder.^7a The obtained yellowish-brown solutions were evaporated under high vacuum at maximum 100 °C leaving dark oily residues, which exhibited high solubility in hexane. Repeated crystallization from hexane at low temperature (−28 °C) afforded a crystalline titanium-containing product 3 only for the system with BTMSA. The formation of 3 from likely 1c should involve insertion of BTMSA into titanium–carbon bonds to give e.g., an 1,5-dititanacyclooctatetraene intermediate that can undergo ring contraction rearrangement (Scheme 2). A brief survey of the X-ray crystal structure of 3 (see below) led to a suggestion that 1,2,4,5-tetrakis(trimethylsilyl)benzene can act as an η⁶:η⁶-bridging ligand, however, this has been questioned by unusual geometric parameters and disproved by subsequent computational study (see below). Also, the synthesis of 3 attempted via reduction of Cp*TiCl₃ with magnesium in the presence of 1,2,4,5-tetrakis(trimethylsilyl)benzene in THF was unsuccessful. The only isolated product was a low-soluble, brown Ti(II) dimer [Cp*Ti(η²-BTMSA)(μ-Cl)]₂ (Chart 1, A) previously obtained by the same reductive method in the presence of BTMSA instead of using the arene. Its isolated yield of 11% indicates, that the arene retro-synthesis proceeds under magnesium-reduction yielding Ti(II). Other reaction products except some unreacted arene were not identified and the presence of 3 was not observed at all.

Scheme 2 A suggested formation of 3 from 1c.

Compound 3 was reproducibly obtained as green crystals in very low yield of 4% related to 1. The thermally robust compound melts at 280 °C (nitrogen atmosphere), even though undergoing simultaneous decomposition. The EI-MS spectra of 3 display the molecular ion m/z 732 as a base peak. Its fragmentation is rather poor, showing mainly the elimination of one [Cp*Ti] moiety (m/z 549) and the [Cp*Ti − 2H]⁺ (m/z 181) and [SiMe₃]⁺ fragment ions. Infrared spectra of 3 in KBr pellet are dominated by very intense absorption bands of the SiMe₃ substituents at 1247 cm⁻¹ and 835 cm⁻¹ leaving other absorption bands of the central bridging ligand and Cp* ligands of rather low intensity. The absorption band of aromatic C–H valence vibration reported for the free arene at 3076 cm⁻¹ (ref. 8) was not observed, likely being shifted to lower wavenumbers into the methyl groups ν(C–H) region. ¹H NMR spectra in toluene-d₈ showed two broad resonances at 2.3 ppm (ν_1/2 = 43 Hz) and 1.3 ppm (ν_1/2 = 54 Hz) which were tentatively assigned to C₅Me₅ and SiMe₃ groups on the basis of their integral intensities (10 : 12; i.e. two Cp* rings versus four SiMe₃ groups). Their chemical shifts were not far from the values for Cp* and SiMe₃ groups in related diamagnetic complexes e.g., Cp*₂Ti(η²-BTMSA) (C₆D₆): δ_H 1.718 ppm (Cp*) and 0.016 ppm (SiMe₃),⁹ and [Cp*Ti(η²-BTMSA)(μ-Cl)]₂ (C₆D₆): δ_H 2.078 ppm (Cp*) and 0.070 ppm (SiMe₃).^3a The signal of the SiMe₃ groups is also significantly shifted downfield compared to that in 1,2,4,5-tetrakis(trimethylsilyl)benzene (δ_H 0.39 ppm)⁸ and the corresponding dianion in [Li(dme)]₂[1,2,4,5-(SiMe₃)₄C₆H₂] (dme = dimethoxyethane) (δ_H 0.21 ppm).¹⁰ The ¹H NMR resonances acquired at temperatures in the range 25–65 °C (see ESI†) showed a slight downfield shift with increasing temperature (0.04 ppm and 0.07 ppm per 20 °C, respectively), whereas their half-widths remained with no change. These observations imply a partial mixing of thermally populated triplet state with a ground singlet state resembling the behaviour of the green dimeric titanocene, [(μ-η⁵:η⁵-C₅H₄C₅H₄)(μ-H)₂{Ti(η⁵-C₅H₅)}₂].¹¹ The two central ring hydrogen atoms were not detected at all, supposed to undergo a considerably larger broadening of their signal. The EPR spectra of the same toluene-d₈ solutions exhibited no signals both at 22 °C as well as in glass at −160 °C as the d¹–d¹ triplet state mentioned above should relax rapidly. Essential information on the molecular structure of 3 was obtained from its single crystal X-ray diffraction study and its electronic structure was examined by computational methods (see below).

Crystal structure of 3

The single crystal structure of 3 consists of two Cp*Ti moieties bridged by a π-faced ligand tentatively resembling 1,2,4,5-tetrakis(trimethylsilyl)benzene. The molecule possesses an exact centre of symmetry, incorporating only one half of its atoms in the asymmetric part of the unit cell (Fig. 1). The distances between the titanium atoms and the least-squares plane of their neighbouring cyclopentadienyl rings (2.1068(2) Å) and the least-squares plane of π-faced bridging ring (1.7381(2) Å) are equal to the corresponding centroid–titanium distances within standard uncertainties (Table 1).

Fig. 1 PLATON drawing of 3 at the 30% probability level with atom labelling scheme. Hydrogen atoms are omitted for clarity. Symmetrically dependent atoms were generated by applying the transformation: −x + 2, −y, −z + 1.

Table 1 Bond lengths (Å) and bond angles (deg) for 3^a

a Symmetry transformations used to generate equivalent atoms: −x + 2, −y, −z + 1.b Cg(1) is centroid and Pl(1) is least-squares plane of the C(1–5) cyclopentadienyl ring.c Cg(2) is the centroid and PL(2) is the least-squares plane of the C(11–13 and 11′–13′) bridging ring.d Dihedral angle between Pl(1) and PL(2).e Dihedral angle between Pl(2) and the Si(1), C(11), C(12), and Si(2) least-squares plane.
Bond lengths (Å)
Ti–Cg(1)^b	2.1076(6)	Ti–Cg(2)^c	1.7381(5)
Ti–PL(1)^b	2.1068(2)	Ti–PL(2)^c	1.7381(2)
Ti–C(1–5)	2.3946(11)–2.4500(11)	C–C(1–5)	1.4136(18)–1.4240(17)
Ti–C(13)	2.2229(11)	Ti–C(13′)	2.2621(11)
Ti–C(11)	2.2809(11)	Ti–C(12′)	2.2851(10)
Ti–C(12)	2.3155(11)	Ti–C(11′)	2.3184(10)
C(11)–C(12)	1.4896(15)	C(12)–C(13)	1.4720(14)
C(11)–C(13′)	1.4735(15)	Si(1)–C(11)	1.8889(11)
Si(2)–C(12)	1.8871(11)	Ti(1)–Ti(1′)	3.4762(4)
Bond angles (deg)
Cg(1)–Ti–Cg(2)^a^,^b	176.85(2)	C(11)–C(12)–C(13)	117.06(9)
C(12)–C(11)–C(13′)	117.12(9)	C(12)–C(13)–C(11′)	125.52(9)
Dihedral angles (deg)
φ^d	4.62(9)	ψ(1)^e	7.05(7)

---

The overall molecule distortion is characterised by the angle Cg(1)–Ti–Cg(2) (Cg(1) is the centre of gravity of the C(1–5) cyclopentadienyl ring and Cg(2) is the centre of gravity of the C(11–13 and 11′–13′) ring) 176.85(2)° and the dihedral angle subtended by the cyclopentadienyl and the bridging ring least-squares planes amounting to 4.62(9)°.

The least-squares plane of Cp* (C1–C5) shows ring carbon atom deviations in the range −0.0076(7)–0.0092(7) Å, while all methyl carbon atoms (C6–C10) are displaced from this plane opposite to the neighbouring titanium atom. The displacement magnitude falls in the range 0.2629(22)–0.3427(21) Å except for C(10) which is displaced only by 0.0855(21) Å. This can be accounted for the absence of steric congestion between C(10) and the arene carbon atom C(13) bearing the sterically non-demanding C–H bond. The bridging ring (C11–C13, C11′–C13′) is nearly planar; its carbon atoms deviating from the least-squares plane by ±0.0242(6) Å for C(13) and ±0.0230(6) Å for C(11) and C(12). The Si(1) and Si(2) atoms are, however, displaced from this plane by −0.2713(20) Å and 0.2606(20) Å, respectively. The bridging ligand ring angle at the C(H) carbon atoms C(12)–C(13)–C(11′) exceeds the angle at carbon atoms C(11) and C(12) which bear the SiMe₃ groups (125.52(9)° versus average 117.09(9)°). These angles are differing only slightly from those found for the free crystalline 1,2,4,5-tetrakis(trimethylsilyl)benzene,¹² being thus apparently caused by steric demands of bulky SiMe₃ substituents.¹³ In contrast to nearly unaffected bond angles, all bond lengths in the bridging ring (C–C_av. 1.478 Å) are substantially elongated (Table 2). Their lengths are markedly larger than in vanadium triple-decker complexes [(CpV)₂-μ-(η⁶:η⁶-C₆H₆)] and [(CpV)₂-μ-(η⁶:η⁶-1,3,5-C₆H₃Me₃)] with C–C_av. 1.443(5) and 1.439(8) Å, respectively,⁶ and their elongation with respect to those in free arene¹² by av. 0.072 Å throws doubts on eventual aromaticity of the bridging ring.

Table 2 Comparison of geometric parameters for 1,2,4,5-tetrakis(trimethylsilyl)benzene (H),¹² its radical anion (H˙⁻),²⁰ and dianion (H²⁻)¹⁰ with those for 3

	H	H˙^− ^a	H^2− ^b	3
a The cation is [K([18]crown-6)(THF)2]^+.b Cations [Li(1,2-dimethoxyethane)]^+ are centred over opposite faces of 1,2,4,5-tetrakis(trimethylsilyl)benzenide.
Bond lengths
C11–C12	1.414(2)	1.462(4)	1.558(10), 1.549(11)	1.4896(15)
C12–C13	1.404(2)	1.401(5), 1.405(5)	1.392(11)–1.416(11)	1.4720(14)
Si–Cring	1.870(3)–1.897(2)	1.845(4)–1.854(4)	1.807(8)–1.839(8)	1.8871(11)–1.8889(11)
Bond angles
C12–C11–C13′	117.1(2)	115.5(3)–116.4(3)	113.5(6)–115.3(6)	117.12(9)
C12–C13–C11′	125.9(2)–129.1(2)	127.5(4), 128.3(4)	128.9(7)–130.2(6)	125.52(9)

---

Comparison of geometry parameters of 3 with those of relevant compounds containing Ti–Cp* and Ti–arene bonds will be useful in further bonding considerations. Thus, the Ti–Cg(1) distance of 2.1076(6) Å is virtually identical with analogous Ti–Cg distances of 2.0995(9) Å and 2.0970(8) Å for [Cp*₂Ti(η²-MeC≡CPh)] complex, where the titanocene (Ti^II) binds the alkyne by back-bonding and where the Cp* – alkyne steric hindrance is low.¹⁴ A shorter Ti–Cg distance of 2.042(5) Å was found for the half-sandwich (Ti^II) complex [Cp*Ti(η²-BTMSA)(μ-Cl)]₂ (Chart 1, A), and 1.983(2) Å for the sandwich titanocene [Cp*₂Ti] (co-crystal of [Cp*₂Ti] with [Cp*₂TiCl]) as a reference.¹⁵

Dibenzenetitanium [Ti⁰(η⁶-C₆H₆)₂] (Chart 2, C) showed a virtually identical distance between its titanium atom and the coplanar benzene ring least-squares planes (1.736 Å)¹⁶ as compound 3. A group of (η⁶-arene)Ti(II)bis(tetrahaloaluminate) complexes afforded average Ti–C_arene distances 2.49–2.50 Å for arene = benzene (Chart 2, D),^17a mesitylene,^17b durene^17c or hexamethylbenzene,^17d,e whereas the same parameter is distinctly shorter for 3 (av. 2.2808(11) Å). Relevant to the discussion of arene–titanium bonding in 3 is also the crystal structure of 1,2-cyclohexane-tethered bis(amidinate)titanium(II)(η⁶-toluene) (Chart 2, E) which contained the puckered toluene ring, having the two halves of its ring intersected with a dihedral angle of 20°. The ring geometry suggests the presence of cyclohexadiene dianion with enhanced sp³ hybridisation on intersecting carbon atoms and their σ-bonding to titanium(IV) as a resonance structure contributing to a Ti(II) centre back-bonding the arene ring.¹⁸

Chart 2 Titanium compounds relevant to 3.

An example of a dititanium complex containing two σ-/π-bonded tripyrrole auxiliary ligands and a bridging π-faced toluene ligand (Chart 2, F) is also closely related to the present case. The two titanium moieties are different because one potassium cation [K(dimethoxyethane)₂]⁺ is coordinated to one of the two auxiliary ligands, making the corresponding titanium atom formally Ti(I). DFT calculations indicated that the two titanium d-electrons are back-donated to two benzene π*-orbitals and the Ti(I) atom binds the toluene ligand more strongly than the Ti(II) one. A considerable difference in titanium–toluene binding strength is reflected in Ti–toluene centroid distances 1.758 and 1.830 Å.¹⁹ An even shorter Ti–Cg(2) distance of 1.7381(2) Å for 3 is to be understood on the basis of DFT calculations (see below).

A tentative clue to the nature of bonding in 3 could be drawn also from comparison of its crystallographic geometric parameters with those for 1,2,4,5-tetrakis(trimethylsilyl)benzene (H),¹² its radical-anion (H˙⁻),²⁰ and dianion (H²⁻)¹⁰ (Table 2).

A considerable elongation of all C–C bonds in 3 with respect to those of 1,2,4,5-tetrakis(trimethylsilyl)benzene (H) (cf. 1.4720(14)–1.4896(15) Å versus 1.403(2)–1.414(2) Å) (Table 2) is suggestive of the back-bonding mechanism resulting in elongation of π-coordinated unsaturated bonds.¹⁴ On the other side, these species do not differ in their ring angle and Si–C_arene bond length values. The anionic H˙⁻ and H²⁻ species show remarkably enlarged differences in bond lengths between carbon atoms bearing the trimethylsilyl groups and the other ones, in larger differences in ring angles, and in a pronounced shortening of the arene carbon–Si atom bond lengths. All these indicate that ionic bonding should not dominate in 3, however, it has to be noted that ionic bonding of bridging arenes is quite common for f-elements. For instance, the lanthanocene and cerocene dinuclear benzene-bridged complexes [(Cp^x₂Ln^II)₂(μ-η⁶:η⁶-C₆H₆)]⁻ form ion pairs with a chelated potassium cation, whereas the negative charge of the lanthanide complex anion is concentrated on bridging benzene ligand – monoanion (C₆H₆)⁻ for Cp^x = C₅H₃(1,3-tBu₂) and dianion (C₆H₆)²⁻ for Cp^x = C₅H₄SiMe₃.²¹ A number of dilanthanocene²² and diuranocene²³ – π-faced arene complexes were prepared and their crystal structures were used as probes for bonding models involving in addition to mono- and di-anionic arene species even π-faced tetraanionic arenes (6C, 10π aromatic system).²⁴ The latter tetraanions should be characterised by the largest sum of C–C bond lengths in the tetraanionic arene ligand,²⁵ which value is virtually equal to that for 3 (∑(C–C) = 8.8702 Å). In spite of the above surprising fit of ∑(C–C) parameters one can hardly accept that two equivalent, formally Ti(I) atoms could generate [Cp*Ti]²⁺ ions to bind to the central [arene]⁴⁻ ion. Hence, computational methods have been employed to elucidate the nature of the Ti-π-faced bridging ligand bonding in 3.

Computational studies of 3

As noted previously, transition metal–benzene anion complexes incorporating various central metals tend to acquire different spin multiplicities with complexes having different metallic centres.²⁶ Computational studies of first-row transition metal triple-decker sandwiches with capped Cp ligands and bridging μ-(η⁶:η⁶-benzene)²⁷ and the similar inverted sandwich type dinuclear complexes containing bridging μ-(η⁶:η⁶-benzene or toluene) ligand and ketiminate anionic (“nacnac”)²⁸ ligands assumed a triplet spin multiplicity for the titanium complex (Chart 2, G).²⁹ Since compound 3 has its electronic arrangement similar to that of G (Cp* replacing nacnac) and since no experimental evidence for its triplet ground state was accessible, our initial computational attempts were aimed at determining the correct multiplicity of 3. This has been done by Density Functional Theory (DFT), and to avoid possible pitfalls with SCF ending possibly in false minima, the stability of wavefunctions has always been checked after achieving self-consistence. Because testing has reported any attempts to describe 3 as a singlet molecule using the spin-restricted approach to lead to a spin restricted → spin unrestricted instability, all our further DFT studies were based on spin-unrestricted computations.

The multiplicity of 3 was determined by comparing the total energies obtained for singlet, triplet and quintet multiplicity computed on the solid-state geometry. Since the lowest energy was obtained for the singlet multiplicity and since changes in multiplicity from singlet to quintet showed a steady increase in energy (Table 3), higher multiplicities were not considered. Using DFT results, the electronic properties and the bonding scheme of 3 were devised from the final model assuming the molecule a diradicaloid singlet.

Table 3 Energy differences for various spin states of 3 from the lowest lying singlet state^a

Multiplicity	ΔE (kJ mol^−1)
a Energies were obtained using the M06 functional and the 6-31+G(d,p) basis set on all atoms.
Singlet	0.00
Triplet	20.89
Quintet	128.59

---

According to DFT results, the coordination of the Cp* ligands is accomplished in the usual manner, however, the coordination of central bridging ligand to the two titanium atoms poses some peculiarities. There are two π systems localised between C(SiMe₃) atoms, and the coordination effected by their overlap with the metals bears analogy with the coordination of alkene ligands in mononuclear bent titanocene alkene/alkyne complexes. The bonding in the latter examples employs a π-donation to the metal and a d-electron back donation to the corresponding alkene/alkyne π*-orbital.¹⁴ The dinuclear 3 coordinates analogously through its two olefin moieties, even with one notable distinction. With mononuclear complexes, the π/π* orbital overlap is effected by ligand orbitals directed towards the empty space of the bent titanocene shell, while for dinuclear 3, the π/π* systems oriented perpendicular to the ligand plane are utilized. This is an appropriate orientation enabling both π/π* systems to concurrently overlap with both metals (Fig. 2).

Fig. 2 Schematic representation of π coordination and the σ coordination coupled with its back-bonding counterpart in [Cp*₂Ti(η²-BTMSA)] (left) and the central ligand in 3 (right). The carbon π/π* orbitals involved in π coordination to the metal(s) are drawn in red. The red lobes on carbon atoms are used to depict simultaneously the π and π* orbitals. Green spheres represent the metals; their particular orbitals utilized for coordination are not drawn. Similar to the alkyne drawn on the left, the coordination in mononuclear alkene complexes is practically the same – both types of complexes engage only their π and π* orbitals directed towards the metal. For mononuclear alkyne complexes, their perpendicular π system remains virtually unaffected.

In analogy with the case of mononuclear complexes, where the back donation to the alkyne ligand results in the carbon–carbon multiple bond elongation, the solid state structure of 3 displayed both its double bonds elongated to 1.4896(15) Å. The π/π* coordination of the central ligand is confined to the olefin carbon atoms, whereas its C–H carbon atoms utilize a σ-like overlap with the metals. Despite of both bonds being of completely different genesis, DFT results reported the canonical orbitals connected with the π/π* and the σ-coordination to be near-degenerate in energy. In addition, Mayer bond orders computed for the six Ti–C bonds were reported to be nearly equal (0.396 to the olefin atoms; 0.323 for the σ-coordinated ones), suggesting a remarkable similarity of the six carbon atoms. The sum of NBO charges on the six skeletal ligand atoms amounted to −1.41e that implies a remarkable contribution of ionic bonding in 3.

Combined with the planarity of the ligand, these results did not exclude the possibility of a benzene-like ligand embedded between two metal atoms. Should this be true, the ligand planarity would inherently come from its aromaticity. However, the aromatic character of a system having originally four π electrons remained still puzzling. On the other side the possibility of non-aromatic character of the ligand was suggested by the missing fully bonding combination of the six skeletal carbon atoms among the canonical orbitals, which contradicted the model of a benzene-like moiety. To resolve these inconclusive observations, the extent of aromaticity of the central ligand has been investigated by Nuclear Independent Chemical Shift (NICS).³⁰

The first NICS computed exactly at the centroid of the central ring (denoted further as NICS(0)) yielded NICS(0) = −46.1517 ppm and NICS(0)_zz, = −48.6315 ppm, which neither provided any directly useful information, since their absolute value exceeded the range expected for any aromatic or even antiaromatic organic system by a wide margin. Such low values indicated the presence of a diatropic ring current and as such would suggest significant ligand aromaticity. However, since NICS values and NICS(0) in particular have been questioned as unequivocal and reliable measure of ring aromaticity³¹ we have also computed the NICS(1) value (NICS taken 1.000 Å from the centroid and normal to the ring plane), which reached NICS(1) = −67.3144 ppm and NICS(1)_zz = −73.5728 ppm. Neither these values were useful in assessing ligand aromaticity, and their exceedingly high magnitude was explained by the contribution of some local currents contributed from the metals to the devised NICS values.

In order to eliminate the contribution from its two neighbouring metal atoms, the NICS computation was repeated for the central ligand only, which was treated as dianion during this step. Considering the dianion of the central ligand only, both NICS values have fallen in the expected range: NICS(0) = 20.3732 ppm, NICS(0)_zz = 12.0100 ppm and NICS(1) = 13.5098 ppm, NICS(1)_zz = 9.1035 ppm. Based on these values, the central ligand is an antiaromatic system. Its multiple bonds thus remain localized and the planarity of this particular ligand in 3 must be the consequence of its interaction with the metals.

To find the reason of the near-degeneracy in energy of the two canonical orbitals incorporating the σ-coordination of the C–H carbons and the π-coordination of the acetylenic atoms and also to provide an explanation for the similar Ti–C bond orders found for all the six carbon atoms, the mixing between the respective canonical orbitals has been examined. Since the central ligand is non-aromatic on its own, the respective mixing was expected to be mediated by the metals. This was verified by NBO second-order Perturbation Theory, which has reported a C=C bond delocalization occurring into the Ti–C σ-antibond with energy 51.30 kJ mol⁻¹. This significant delocalization energy thus confirmed the anticipated mixing between the two orbitals and has also underlined the importance of the metallic contribution in its occurrence. This mixing is well discernible also among the canonical orbitals in the particular orbital that incorporates the Ti–C σ-overlap. The presence of such delocalization explains the preference for a planar ligand skeleton – the mixing is most efficient when the carbon atoms are lying in plane (Fig. 3); such an arrangement however cannot occur without the involvement of the metals, which provide their d-orbitals (mostly d_x²−y²) for this purpose.

Fig. 3 σ-Coordination (left) and back donation to the π* ligand orbitals (right) in 3. The left orbital includes also contribution from the C=C π-system. The C–H bonds of the central ligand are lying in the paper plane. Both orbitals are drawn with 5% probability.

In search for the reason of high thermal stability of complex 3, neither the covalent nor the ionic interactions between the ligands and the metals provided a fully convincing explanation for its occurrence. Seeking for additional stabilizing contributions, the presence of a direct metal–metal interaction has been uncovered. This interaction is generated from the overlap between two originally nonbonding metallic d orbitals (Fig. 4), and although the separation between the respective centres disfavours a markedly beneficial overlap, the Ti–Ti Mayer bond order computed by DFT still acquired the value 0.333. This value suggested the importance of the direct Ti–Ti bond in the molecule, which could contribute to the overall stability of 3.

Fig. 4 Alpha canonical orbitals no. 198 (HOMO; left) and 199 (LUMO; right) of 3. Both orbitals are drawn with 4% probability and are dominantly of d-character (96%).

However, the presence of this particular metal–metal bond has raised suspicion that other d orbitals could also take part in its formation. Since these orbitals were for principal reasons treated as unoccupied by DFT, results obtained by DFT could possibly describe the metal–metal bonding underestimated in its magnitude. The probable involvement of other d orbitals was also supported by technical details during DFT studies, like the SCF procedure yielding notoriously unstable wavefunctions, and also by NBO, which reported a system apparently in excited state even on a stabilized ground-state wavefunction. These observations were interpreted as warning signs of treating a molecule that is mixing several electronic configurations simultaneously. To cope with such an electronic system, we have extended our computational studies by Complete Active Space SCF (CASSCF) followed by Second Order Perturbation Theory.

The active space in CASSCF was constructed from the canonical orbitals that incorporated a large contribution either from the metals, or from the central ligand, or from both. Any orbitals involved in coordinating the Cp* ligands only have been kept inactive. To account for possible additional low-lying excitation(s), computations were carried out as state-averaged CASSCF with three roots.

In agreement with DFT, CASSCF confirmed the coordination of the central ligand through concurrent ligand π-coordination and back donation to its π* orbitals and reported also the σ-coordination from the two C–H carbon atoms. However, it has uncovered some additional nuances in describing the metal–metal bond. Comparably with DFT, a direct Ti–Ti σ-interaction has been reported (cf. Fig. 4), however, state-averaged CASSCF yielded markedly non-integer occupation number (o.n.) not only for its bonding combination (o.n. 0.84) but also for its corresponding antibonding counterpart (o.n. 0.52). These results justified the necessity to apply a multiconfigurational approach in order to describe the Ti–Ti bond in 3.

To further increase the quality of results obtained by multiconfigurational computations, a CASPT2 (CASSCF followed by Second Order Perturbation Theory) computation has been carried out. Using the same CASSCF active space as described above, the successive CASPT2 approach yielded the first excitation energy 1.675 eV, in fairly good agreement with the experimental value 1.503 eV (λ_max = 825 nm). Bond analysis has revealed, that in addition to the σ-coordination described above, two δ-type orbitals (with their antibonding counterparts) are also present (Fig. 5) in generating a direct bond between the metals.

Fig. 5 CASSCF orbitals involved in the excitation of 3. Top row: Orbital 99 (left), orbital 100 (centre), orbital 101 (right). Bottom row: Orbital 267 (left), orbital 266 (centre), orbital 265 (right). The top row represents orbitals in bonding overlap (all a_g symmetry); the bottom row represents their antibonding combinations (all a_u symmetry). All orbitals are drawn with 5% probability.

Even though there are three metallic orbitals involved in the generation of the direct Ti–Ti bond, due to the large intermetallic separation (d(Ti–Ti) 3.4762(4) Å) in the molecule, the overlap between the δ-orbitals is impaired. The interaction between the metals can thus be considered as some intermediate between a regular chemical bond and antiferromagnetically coupled electrons separated to two centres.

The occupancy numbers of orbitals (Table 4) obtained by CASSCF suggested that all d-orbitals (and their combinations) are partially populated in ground state. Although upon excitation some major changes in occupancies of individual orbitals could be observed, the overall change in the Ti–Ti bond order remained still modest. This insignificant change of the Ti–Ti bond order was uncovered by Natural Bond Order analysis based on the LoProp computed density, yielding the values: root 1 = 1.864; root 2 = 1.719; root 3 = 1.710. On behalf of these results, the thermal stability of 3 can be explained by the presence of a direct Ti–Ti covalent bond, whose bond order decreased only slightly even upon excitation of the parent molecule.

Table 4 Occupancy numbers for the six most important orbitals involved during excitation of 3

CASSCF root	Orb. 99 (3dδg)	Orb. 100 (3dσg)	Orb. 101 (3dδg)	Orb. 265 (3dδu)	Orb. 266 (3dσu)	Orb. 267 (3dδu)
1 (ground state)	0.23	1.05	0.21	0.09	0.95	0.08
2	0.25	0.73	0.89	0.27	0.30	0.09
3	0.92	0.74	0.25	0.10	0.29	0.23

---

By using the multiconfigurational methods, we have rechecked the correct multiplicity of 3. In accordance with DFT results, the lowest total energy was reported for the singlet ground state, since the CASPT2 root 1 energy was 109.33 kJ mol⁻¹ higher for the analogous triplet state. However, the first excited triplet state energy was already falling below the corresponding value of the singlet state. Although not studied further in detail, the observed downfield shift of NMR signals of 3 (see above) can possibly be attributed to the presence of thermally populated triplet state molecules in solution. The intersystem crossing between the singlet and triplet states (and vice versa) is possibly achieved through spin–orbital coupling, supported by the presence of two metals in the complex. CASPT2 energies of the first three roots for both singlet and triplet states are given in ESI.†

Conclusions

Thermally robust triple-decker complex [bis(η⁵-pentamethylcyclopentadienyltitanium)-μ-(η⁴:η⁴-1,2,4,5-tetrakis(trimethylsilyl)cyclohexa-1-4-diene-3,6-diyl)] (3) forms reproducibly by thermolysis of Cp*TiMe₃ in the presence of BTMSA in yield of 4%. The X-ray single crystal diffraction of 3 revealed the centrosymmetric molecule of the above chemical formula, having the following important metric data for evaluation of bonding: d(Ti–Ti) 3.4762(4) Å, d(Ti–C_Cp*) 2.3946(11)–2.4500(11) Å, d(Ti–C_{cyclohexadiene}) 2.2229(11)–2.3184(10) Å, and d(C–C)_{cyclohexadiene} = 1.4720(14)–1.4896(15) Å. The exceedingly long C–C bonds in the coordinated planar cyclohexadiene ring (av. 1.478(4) Å versus av. 1.407(2) Å for the arene¹²) indicate a strong interaction of titanium d-electrons with antibonding π* orbitals of the cyclohexadiene double bonds. DFT computational studies revealed a concurrent σ- and π-coordination from the central ligand, the latter overlap complemented by back donation from the metal to the ligand π* acceptors. Apart from being bonded through the central bridging ligand, the two metals are also bonded by direct metal–metal bonds. Using a multiconfigurational approach, the Ti–Ti bonding was found to be achieved by the formation of one σ- and two δ-bonds. The overall Ti–Ti bond order decreases only insignificantly upon excitation of the molecule. CASPT2 suggests a singlet multiplicity for the molecule with some contribution of an energetically more favoured exited state of triplet multiplicity. This assessment is compatible with the absence of EPR signals for toluene solution and glass and only a slight downfield shifts (∼0.1 ppm) of ¹H NMR resonances upon increasing the measurement temperature from 25 °C to 65 °C.

Experimental

General considerations

All reactions leading to low-valent titanium compounds and their subsequent reactions were carried out on a vacuum line in sealed all-glass devices equipped with breakable seals. ¹H (300 MHz) NMR spectra were recorded on a Varian Mercury 300 spectrometer in toluene-d₈ solutions at 25 °C. Chemical shifts (δ/ppm) are given relative to the residual solvent signal (CD₂H: δ_H 2.08 ppm). EI-MS spectra were obtained on a VG-7070E mass spectrometer at 70 eV. Crystalline samples in sealed capillaries were opened and inserted into the direct inlet under argon. The spectra are represented by the peaks of relative abundance higher than 7% and by important peaks of lower intensity. Crystalline samples for EI-MS measurements and melting point determinations were placed in glass capillaries in a glovebox Labmaster 130 (mBraun) under purified nitrogen (concentrations of oxygen and water were lower than 2.0 ppm) and sealed with flame. KBr pellets were prepared in the glovebox and were measured in an air-protecting cuvette on a Nicolet Avatar FT IR spectrometer in the range 400–4000 cm⁻¹. UV-near IR spectra in the range 300–2000 nm were measured on a Varian Cary 17D spectrometer in all-sealed quartz cells (Hellma). Elemental analyses were carried out on a FLASH EA1112 CHN/O Automatic Elemental Analyzer (Thermo Scientific). Melting points were measured on a Koffler block in sealed glass capillaries under nitrogen, and are uncorrected.

Chemicals

The solvents hexane, and m-xylene were dried by refluxing over LiAlH₄ and stored as solutions of green dimeric titanocene [(μ-η⁵:η⁵-C₅H₄C₅H₄)(μ-H)₂{Ti(η⁵-C₅H₅)}₂] (ref. 32) on a vacuum line. Toluene-d₈ (99.5% D) (Sigma Aldrich) was degassed, distilled under vacuum on singly tucked-in permethyltitanocene [Ti(C₅Me₅)(C₅Me₄CH₂)],^7b and stored as its solution on a vacuum line. The pentamethylcyclopentadienyltitanium trimethyl [(η⁵-C₅Me₅)TiMe₃] was purchased from (Sigma Aldrich), opened under nitrogen in glovebox, and dissolved in degassed hexane to give 0.1 M solution. Internal alkynes 1-(trimethylsilyl)propyne, 1-phenylpropyne, and bis(trimethylsilyl)acetylene (BTMSA) (all Sigma Aldrich) were degassed and handled in vacuum. 1,2,4,5-Tetrakis(trimethylsilyl)benzene was prepared from 1,2,4,5-tetrachlorobenzene as reported.¹² The pure product was separated from a cyclohexadiene byproduct C₆H₂(SiMe₃)₆ by fractional vacuum sublimation.

Synthesis

Preparation of 3. A solution of 1 in hexane (0.1 M, 4.0 mL) was evaporated in vacuum, and the solid residue was dissolved in degassed m-xylene (4.0 mL). Bis(trimethylsilyl)acetylene (BTMSA) (0.2 mL, 0.9 mmol) was added, turning a slightly yellowish colour of the solution to yellow. Gradual heating in a flame-sealed glass ampule (volume 50 mL) turned the yellow color of the mixture to brown at 110 °C. After heating to 125 °C for total 19 h and cooling to ambient temperature the ampule was opened into vacuum and all volatiles were distilled off, finally at 100 °C. The residue was dissolved in hexane (2.0 mL) and the solution was cooled in a freezer (−28 °C) for two weeks. An oily brown mother liquor was slowly decanted from a little crystalline product in a refrigerator. The product was recrystallized from hexane to give green crystals of 3. Yield 6 mg (4%). This thermolytic synthesis was reproduced five times affording similar yields. No further solid products were obtained from the mother liquors by crystallisation at −7 to −20 °C with simultaneous slow solvent evaporation in a closed degassed device. Additional experiments using a higher excess of BTMSA (2 mmol) as well as the above reaction performed at 110 °C for 40 h did not give an improved yield.
Analytical data for compound 3. Mp: 280 °C dec. EI-MS (260 °C): m/z (relative abundance) 736 (10), 735 (23), 734 (47), 733 (70), 732 (M⁺; 100), 731 (34), 730 (26), 729 (9), 728 (9), 659 ([M − SiMe₃]⁺; 9), 551 (15), 550 (24), 549 ([M − Cp*Ti]⁺; 42), 548 (7), 476 ([M − Cp*Ti − SiMe₃]⁺; 9), 183 (13), 182 (28), 181 ([Cp*Ti − 2H]⁺; 71), 180 (33), 179 (21), 178 (21), 177 (11), 155 (12), 105 (8), 73 ([SiMe₃]⁺; 59). IR (KBr, cm⁻¹): 2982 (m), 2951 (s), 2900 (s), 2856 (m), 2722 (vw), 1486 (vw), 1437 (w,b), 1404 (w), 1376 (w), 1309 (vw), 1253 (s), 1243 (s), 1155 (vw), 1104 (w), 1049 (s), 1023 (w), 835 (vs,b), 750 (m), 678 (w,b), 642 (w), 637 (w), 495 (m), 467 (w). ¹H NMR (toluene-d₈): 1.3 (br s, ν_1/2 = 54 Hz, 36H, SiMe₃); 2.3 (br s, ν_1/2 = 43 Hz, 30H, C₅Me₅). UV-near IR (toluene-d₈, nm): 342 ≫ 490(sh) ∼ 630–825. Found (%): C, 62.31; H, 9.38. C₃₈H₆₈Si₄Ti₂ requires (%): C, 62.26; H, 9.35.

Thermolysis of 1 in the presence of 1-(trimethylsilyl)propyne or 1-phenylpropyne. Thermolyses of 1 in the presence of two-fold molar excess of the alkynes were conducted analogously to synthesis of 3 at 125 °C. After heating for 20 h all volatiles were distilled off in high vacuum at 100 °C. Non-volatile residues were dissolved in minimum hexane and numerous unsuccessful attempts were carried out to obtain crystalline product(s) at temperatures from −70 °C to −5 °C combined with slow solvent evaporation. ¹H and ¹³C NMR spectra of the residue samples revealed a tangle of signals indicating mixtures of products. EPR spectra of the same NMR tube solutions indicated the presence of apparently Ti(III) complexes in rather non-significant concentrations.

Attempted synthesis of 3 by magnesium-reduction method. 1,2,4,5-Tetrakis(trimethylsilyl)benzene (0.37 g, 1.0 mmol) and magnesium (0.24 g, 10 mmol) were degassed, and mixed with a solution of Cp*TiCl₃ (0.58 g, 2.0 mmol) in THF (20 mL).

The stirred mixture changed the initial red color rapidly to green and then to brown. After 2 h a dark brown solution was evaporated in vacuum, and a dark residue was repeatedly extracted with hexane. The concentrated extract was cooled to −28 °C precipitating brown crystalline solid. The latter was fractionally crystallised to give [C₆H₂-1,2,4,5-(SiMe₃)₄] (0.22 g, 60%) and brown compound [Cp*Ti(η²-BTMSA)(μ-Cl)]₂ (Chart 1, A) (0.09 g, 11%). The both compounds were identified by ¹H NMR and IR (KBr) spectra, the latter also by X-ray single crystal diffraction. The presence of 3 was not detected in either of the fractions.

X-ray crystallography

A single crystal of 3 was mounted into Lindemann glass capillary in a Labmaster 130 glovebox (mBraun) under purified nitrogen. Diffraction data were collected on a Nonius Kappa diffractometer equipped with a Bruker APEX II area detector (Mo Kα radiation, λ = 0.71073 Å). The data reduction and solution of the phase problem was carried out by the APEX2 program package.³³ The structure model was refined by full-matrix least-squares on F² using SHELXL-97.³⁴ All non-hydrogen atoms were refined anisotropically. All hydrogen atoms were refined isotropically in their theoretical positions (riding model). Molecular graphics was done with the PLATON program.³⁵ Relevant crystallographic data are gathered in ESI.†

Computational details

All computational studies were conducted on the Bose cluster at the J. Heyrovský Institute of Physical Chemistry of the Czech Academy of Sciences, v.v.i. The molecular geometry used for the computational studies was the same as obtained from the diffraction experiment. Density Functional Theory computations were carried out by Gaussian 09, Revision D.01 (ref. 36) using the M06 functional,³⁷ the 6-31+G(d,p) basis set for all atoms. The point group C_i was assumed for the molecule. Numerical integration in DFT was carried out on an ultrafine grid; the SCF convergence criterion was tightened to 10⁻⁹. Wavefunctions were always verified for stability post SCF. NBO studies were performed by a standalone version of NBO 5.G³⁸ on a wavefunction obtained with the 6-31G(d,p) basis set.

CASSCF and CASPT2 computations were carried out by MOLCAS 8.0 (ref. 39) using the Dolg pseudopotential as built in MOLCAS for all atoms. The State-Averaged CASSCF employed three roots all with equal weights. The active space included a total of 13 orbitals and 10 electrons and was constructed from canonical orbitals incorporating a large contribution either from the metals, from the central ligand or from both. Orbitals involved in coordinating only the Cp* ligands were kept inactive. This selection of orbitals allowed for a smooth convergence in CASSCF with no large orbital rotations reported during convergence. Upon successful termination of CASSCF, orbitals in the active space were checked for possible rotations out from the active space. To account also for dynamic correlation energy, the CASSCF step was followed by Multi-State CASPT2 computation.

Attempts to perform geometry optimization prior to devising any properties from the wavefunction were also attempted. The DFT optimization has however removed the planarity of the central ring (although not too significantly), which was interpreted as the result of using only one electron configuration for the description of a multiconfigurational molecule. The attempt to optimize the geometry using CASSCF proved to be more successful and after computing the required gradients the program has reported a structure at stationary point with no requirement for further geometry optimization.

Acknowledgements

This research was supported by Grant Agency of the Czech Republic (Project No. P207/12/2368). R.Gy. is grateful to the Slovak Grant Agency VEGA (Project No. 1/0336/13).

Notes and references

1. (a) S. A. Cohen and J. E. Bercaw, Organometallics, 1985, 4, 1006–1014; (b) B. Demerseman, P. L. Coupanec and P. U. Dixneuf, J. Organomet. Chem., 1985, 297, C35; (c) V. B. Shur, V. V. Burlakov and M. E. Vol'pin, J. Organomet. Chem., 1988, 347, 77–83; (d) V. Burlakov, U. Rosenthal, P. V. Petrovskii, V. B. Shur and M. E. Vol'pin, Metalloorg. Khim., 1988, 1, 953–954.
2. (a) U. Rosenthal, V. V. Burlakov, P. Arndt, W. Baumann and A. Spannenberg, Organometallics, 2003, 22, 884–900; (b) U. Rosenthal, V. V. Burlakov, P. Arndt, W. Baumann, A. Spannenberg and V. B. Shur, Eur. J. Inorg. Chem., 2004, 4739–4749; (c) U. Rosenthal and V. V. Burlakov, in Titanium and Zirconium in Organic Synthesis, ed. I. Marek, Wiley-VCH, Weinheim, Germany, 2002, pp. 355–389.
3. (a) J. Hiller, U. Thewalt, J. Podlaha, V. Hanuš and K. Mach, Collect. Czech. Chem. Commun., 1997, 62, 1551–1561; (b) V. Varga, K. Mach, G. Schmid and U. Thewalt, J. Organomet. Chem., 1993, 454, C1–C4; (c) V. Varga, K. Mach, G. Schmid and U. Thewalt, J. Organomet. Chem., 1994, 475, 127–137.
4. M. Mena, P. Royo, R. Serrano, M. A. Pellinghelli and A. Tiripicchio, Organometallics, 1989, 8, 476–482.
5. R. Andrés, P. Gómez-Sal, E. de Jesús, A. Martín, M. Mena and C. Yélamos, Angew. Chem., Int. Ed. Engl., 1997, 36, 115–117.
6. A. W. Duff, K. Jonas, R. Goddard, H.-J. Kraus and C. Krüger, J. Am. Chem. Soc., 1983, 105, 5479–5480.
7. (a) Compound 2 was prepared and characterised including the X-ray crystal structure in accordance with ref. 5; some additional data (EI-MS, IR and electronic absorption spectra) are listed in ESI.†; (b) J. Pinkas, I. Císařová, R. Gyepes, M. Horáček, J. Kubišta, J. Čejka, S. Gómez-Ruiz, E. Hey-Hawkins and K. Mach, Organometallics, 2008, 27, 5532–5547.
8. K. Mach, F. Tureček, H. Antropiusová and V. Hanuš, Organometallics, 1986, 5, 1215–1219.
9. (a) V. V. Burlakov, A. V. Polyakov, A. I. Yanovsky, Yu. T. Struchkov, V. B. Shur, M. E. Vol'pin, U. Rosenthal and H. Görls, J. Organomet. Chem., 1994, 476, 197–206; (b) V. Varga, K. Mach, M. Polášek, P. Sedmera, J. Hiller, U. Thewalt and S. I. Troyanov, J. Organomet. Chem., 1996, 506, 241–251.
10. A. Sekiguchi, K. Ebata, C. Kabuto and H. Sakurai, J. Am. Chem. Soc., 1991, 113, 7081–7082.
11. D. A. Lemenovskii, I. F. Urazowski, Yu. K. Grishin and V. A. Roznyatovsky, J. Organomet. Chem., 1985, 290, 301–305.
12. H. Bock, M. Ansari, N. Nagel and Z. Havlas, J. Organomet. Chem., 1995, 499, 63–71.
13. Similar differences in benzene ring angles (123.1(9)° and 116.8(6)°) were observed for 1,3,5-tris(trimethylsilyl)-2,4,6-triphenylbenzene where the carbon atoms bearing phenyl groups were less sterically demanding due to orientation of the phenyl groups planes nearly perpendicularly to the central benzene ring, perpendicularly to the central benzene ring. R. Klein, G. Schmid, U. Thewalt, P. Sedmera, V. Hanuš and K. Mach, J. Organomet. Chem., 1994, 466, 125–131.
14. J. Pinkas, R. Gyepes, I. Císařová, J. Kubišta, M. Horáček and K. Mach, Dalton Trans., 2015, 44, 7276–7291.
15. J. Pinkas, I. Císařová, R. Gyepes, J. Kubišta, M. Horáček and K. Mach, Organometallics, 2012, 31, 5478–5493.
16. G. G. Tairova, E. F. Kvashina, O. N. Krasochka, G. A. Kichigina, Yu. A. Shvetsov, E. M. Lisetsky, L. O. Atovmyan and Yu. G. Borodko, New J. Chem., 1981, 5, 603–604.
17. (a) U. Thewalt and F. Stollmaier, J. Organomet. Chem., 1982, 228, 149–152; (b) V. Kupfer, K. Mach and U. Thewalt, Collect. Czech. Chem. Commun., 2000, 65, 192–202; (c) S. I. Troyanov and K. Mach, J. Organomet. Chem., 1990, 389, 41–46; (d) U. Thewalt and F. Österle, J. Organomet. Chem., 1979, 172, 317–324; (e) S. I. Troyanov, J. Poláček, H. Antropiusová and K. Mach, J. Organomet. Chem., 1992, 430, 317–325.
18. J. R. Hagadorn and J. Arnold, Angew. Chem., Int. Ed., 1998, 37, 1729–1731.
19. G. B. Nikiforov, P. Crewdson, S. Gambarotta, I. Korobkov and P. H. M. Budzelaar, Organometallics, 2007, 26, 48–55.
20. P. B. Hitchcock, M. F. Lappert and A. V. Protchenko, J. Organomet. Chem., 2011, 696, 2161–2164.
21. C. M. Kotyk, M. E. Fieser, C. T. Palumbo, J. W. Ziller, L. E. Darago, J. R. Long, F. Furche and W. J. Evans, Chem. Sci., 2015, 6, 7267–7273.
22. (a) M. C. Cassani, D. J. Duncalf and M. F. Lappert, J. Am. Chem. Soc., 1998, 120, 12958–12959; (b) Y. K. Gun'ko, P. B. Hitchcock and M. F. Lappert, Organometallics, 2000, 19, 2832–2834.
23. (a) P. L. Diaconescu, P. L. Arnold, T. A. Baker, D. J. Mindiola and C. C. Cummins, J. Am. Chem. Soc., 2000, 122, 6108–6109; (b) W. J. Evans, S. A. Kozimor, J. W. Ziller and N. Kaltsoyannis, J. Am. Chem. Soc., 2004, 126, 14533–14547; (c) W. J. Evans, C. A. Traina and J. W. Ziller, J. Am. Chem. Soc., 2009, 131, 17473–17481.
24. (a) W. Huang, F. Dulong, T. Wu, S. I. Khan, J. T. Miller, T. Cantat and P. L. Diaconescu, Nat. Commun., 2013, 4, 1448–1454; (b) D. Patel, F. Tuna, E. J. L. McInnes, J. McMaster, W. Lewis, A. J. Blake and S. T. Liddle, Dalton Trans., 2013, 42, 5224–5227.
25. A. Falceto, D. Casanova, P. Alemany and S. Alvarez, Chem.–Eur. J., 2014, 20, 14674–14689.
26. H. F. Li, X.-Y. Kuang and H.-Q. Wang, Dalton Trans., 2011, 40, 4578–4589.
27. H. Liu, Q. Li, Y. Xie, R. B. King and H. F. Schaefer III, J. Phys. Chem. A, 2010, 114, 8156–8164.
28. Ketiminate complexes of this type were prepared for Cr: (a) W. H. Monillas, G. P. A. Yap and K. H. Theopold, Angew. Chem., Int. Ed., 2007, 46, 6692–6694 ; and V: ; (b) Y.-C. Tsai, P.-Y. Wang, K.-M. Lin, S.-A. Chen and J.-M. Chen, Chem. Commun., 2008, 205–207.
29. Y. I. Kurokawa, Y. Nakao and S. Sakaki, J. Phys. Chem. A, 2010, 114, 1191–1199.
30. (a) M. Bühl and C. v. Wullen, Chem. Phys. Lett., 1995, 247, 63–68; (b) P. v. R. Schleyer, C. Maerker, A. Dransfeld, H. Jiao and N. J. R. v. E. Hommes, J. Am. Chem. Soc., 1996, 118, 6317–6318.
31. C. Foroutan-Nejad, Theor. Chem. Acc., 2015, 134, 8.
32. H. Antropiusová, A. Dosedlová, V. Hanuš and K. Mach, Transition Met. Chem., 1981, 6, 90–93.
33. Bruker, APEX2, SAINT, Bruker AXS Inc., Madison, Wisconsin, USA, 2010.
34. G. M. Sheldrick, SHELXL97, Program for Crystal Structure Refinement from Diffraction Data, University of Göttingen, Göttingen, 1997.
35. A. L. Spek, PLATON, A Multipurpose Crystallographic Tool, Utrecht University, Utrecht, 2007.
36. Gaussian 09, Revision D.01, Gaussian, Inc., Wallingford CT, 2010.
37. Y. Zhao and D. G. Truhlar, Theor. Chem. Acc., 2008, 120, 215–241.
38. E. D. Glendening, J. K. Badenhoop, A. E. Reed, J. E. Carpenter, J. A. Bohmann, C. M. Morales and F. Weinhold, NBO 5.0, Theoretical Chemistry Institute, University of Wisconsin, Madison, USA, 2001.
39. F. Aquilante, J. Autschbach, R. K. Carlson, L. F. Chibotaru, M. G. Delcey, L. De Vico, I. F. Galvan, N. Ferré, L. M. Frutos, L. Gagliardi, M. Garavelli, A. Giussani, C. E. Hoyer, G. Li Manni, H. Lischka, D. Ma, P. Å. Malmqvist, T. Müller, A. Nenov, M. Olivucci, T. B. Pedersen, D. Peng, F. Plasser, B. Pritchard, M. Reiher, I. Rivalta, I. Schapiro, J. Segarra-Marti, M. Stenrup, D. G. Truhlar, L. Ungur, A. Valentini, S. Vancoillie, V. Veryazov, V. P. Vysotskiy, O. Weingart, F. Zapata and R. Lindh, MOLCAS 8, Journal of Computational Chemistry, 2016, vol. 37, pp. 506–541.

---

Footnote

† Electronic supplementary information (ESI) available: CIF file for the structure 3. ¹H NMR spectra of solution of 3 in toluene-d₈ measured at variable temperatures, experimental data for 1,2,4,5-tetra(trimethylsilyl)benzene and crystallographic data and data collection, structure refinement details and parameters for 3. CCDC 1446324 (3). For ESI and crystallographic data in CIF or other electronic format see DOI: 10.1039/c6ra14940e

---