2.1 Synthesis of the allenylidene complexes 5–7

Very few iron allenylidene complexes were reported in the [(η⁵-C₅R₅)(η²-dppe)Fe(=C=C=CR¹R²)][X] series. To our knowledge, the preparation of [(η⁵-C₅H₅)(η²-dppe)Fe(=C=C=CPh₂)][BF₄] by photo-activation upon irradiation at 280 nm of [(η⁵-C₅H₅)(η²-dppe)Fe(CO)][BF₄] in the presence of the propargylic alcohol H-C≡C–C(Ph)₂OH constitutes a very rare example for such a complex [18]. To access to our synthetic targets, we followed the classical thermal procedure. This efficient method was first employed for the preparation of a ruthenium complex [19], but has found widespread application with other transition elements [20].

Complex (η⁵-C₅Me₅)(η²-dppe)FeCl (10) reacts with 1 equiv of 1,1-diphenyl-prop-2-yn-1-ol in methanol at 20 °C in the presence of KPF₆ for 12 h to give the purple allenylidene complex [(η⁵-C₅Me₅)(η²-dppe)Fe(=C=C=C(Ph)Ph)][PF₆] (5a, 95%). The analytically pure complex was characterized by IR and NMR. The IR spectrum of 5a shows typical ν_C=C=C and ν_(PF6) at 1896 and 839 cm^–1, respectively. The ³¹P{¹H} NMR of 5a in CDCl₃ displays a singlet for the dppe ligand and a septuplet characteristic of the PF₆ anion. The allenylidene derivatives are often formed from their hydroxyvinylidene precursors by spontaneous loss of H₂O, but sometimes this step requires assistance: for example, an acidic catalysis (Fig. 5). A salt effect can also facilitate the dehydration step. Indeed, in order to prepare the diphenylallenylidene iron complex as a tretraphenyl borate salt, we worked in the same conditions, replacing KPF₆ by NaBPh₄. Surprisingly, 5b was isolated as a mixture with the hydroxyvinylidene [(η⁵-C₅Me₅)(η²-dppe)Fe(=C=C(H)C(OH)(Ph)₂)][BPh₄] (14b) in the 1:1 ratio. The vinylidene intermediate was identified in the mixture by its IR characteristic vibration modes ν_C=C and ν_OH at 1616 and 3560 cm^–1, respectively. The ¹H NMR spectrum of complex 14b shows characteristic signals at δ 1.49 (s) and 4.82 (t, J_HP = 5 Hz), corresponding to the resonances of the protons of the C₅Me₅ ligand and the β carbon of the vinylidene fragment. A resonance at 88.6 ppm was also observed in the ³¹P NMR spectrum. This signal presents the same intensity than that of 5b, allowing the determination of the ratio 5b/14b in the crude product. The salt effect on the dehydration step could be explained by the ability of the hexafluorophosphate to give hydrogen bonds with acidic protons. For example, a hydrogen bond was recently observed by X-ray between a PF₆^– anion and an acidic hydrogen atom in an organometallic compound [21]. In the case of the mixture of 5b and 14b, treatment of the crude product with Amberlyst 15 in methanol provides the allenylidene 5b as a pure sample. The preparation of 5b was also carried out in the presence of Amberlyst 15. In this case, the vinylidene intermediate was not observed and 5b was directly obtained in 91% yield. The allenylidene complexes 6 and 7 were prepared according to the same procedure and they were isolated as purple powders in 90% yield. Subsequent crystallization gave analytically pure compounds which were characterized by IR, ¹H-, ¹³C-, ³¹P-NMR spectroscopies.

2.2 Crystal structures of 3, 4·C₄H₁₀O, and 5a·C₄H₁₀O

The crystal structures of 3, 4·C₄H₁₀O and 5a·C₄H₁₀O were determined as outlined in Table 1 and the ‘Experimental’ section. The molecular structures are shown in Figs. 6–8. Key distances and angles are given in Table 2. General features, such as the formally octahedral geometry at the iron centres, accord with past structures in this series [1, 11]. The Fe=C bond distances in 3 (1.763(7) Å) and 4·C₄H₁₀O (1.760(5) Å) are very similar and consistent with a bond order of about two. However, it can be observed that the Fe=C bond length in 3 and 4·C₄H₁₀O are shorter than the Fe=C bond in the related iron carbene 1 (1.787(9) Å) [11] or in the allenylidene 5a·C₄H₁₀O (1.781(9) Å). It is also noteworthy that theoretical calculations have shown that the metal–carbon bond in metallavinylidene derivatives possesses a partial triple bond character [22]. The C_α–C_β bond distances in 3 (1.224(9) Å) and 4·C₄H₁₀O (1.295(5) Å) are relatively short for a C=C double bond [23, 24]. The Fe–C_α–C_β angle slightly deviates from linearity as it was previously noted in the closely related vinylidene [(C₅H₅)(dppm)Fe(=C=C(Me)Ph)]I [24] and also in many other mononuclear complexes bearing a vinylidene ligand [23]. Moreover, the positions of the hydrogen atoms on C_β were found by analysis of the X-ray data allowing the observation of a partial pyramidalization of the β carbon. In contrast, this carbon presents a planar geometry in the complex 4·C₄H₁₀O. Note that very few complexes of primary vinylidene with transition metal were X-ray characterized, and for the first row of the d-elements there is only one precedent [25].

Although calculations have predicted that the plane of the vinylidene ligand should be perpendicular to the molecular plane in complexes of the type M(CCR₂)(L₂)(C₅R₅) [26], the barrier to rotation was computed to be only ca 15 kJ mol^–1, so that this preference is often overridden by steric or packing effects [23]. This is the case in complex 3, where the values of the torsion angles defined by Cp^*_cent–Fe–C₃₈–H_A and Cp^*_cent–Fe–C₃₈–H_B are 91.1 and 125.6°, respectively. A similar behaviour is also noted in the case of complex 4·C₄H₁₀O, for which the torsion angles Cp^*_cent–Fe–C₃₈–C₃₉ and Cp^*_cent–Fe–C₃₈–H are equal to 74.3 and 104.5°, respectively.

Structural parameters have been reported for a large number of mononuclear allenylidene complexes with many transition metals [20, 27]; however, only one precedent is known in the iron series [28]. As mentioned above, the Fe=C bond distance in the allenylidene 5a·C₄H₁₀O is 0.02 Å longer than in the vinylidene 3, but it compares very well with the value determined for the corresponding alkylidene derivative 1 [11]. As usual, in the three-carbon chain, the C_α–C_β distances (1.257(7) Å) is shorter than the C_β–C_γ distance (1.361(7) Å) [20]. These observed distances are not far from the values reported for similar diphenyl allenylidene complexes of ruthenium and osmium and indicate a contribution of the canonical form [M]-(C≡CC⁺Ph₂) [20, 29]. Such an electronic contribution is also supported by EHMO calculations, which show that there is an alternation of electron density along unsaturated carbon chain. These studies indicate that C_α and odd-numbered carbon atoms are electrophilic centres, while C_β and even-numbered atoms are electron rich [30].

The C₃ fragment is nearly linear, the angles at C_α and C_β being close to 180°. The plane defined by the carbon atom and the two ipso carbon atoms of the phenyl rings is almost perpendicular to the Cp^* plane, since the torsion angles Cp^*_cent–Fe–C₃₉–C₄₀ and Cp^*_cent–Fe–C₃₉–C₄₆ are equal to 6.4 and 170.5°, respectively. The weak deviation from the orthogonal orientation theoretically expected for these two planes was also noted in several cases [20] and can be explained by the small barrier to rotation for allenylidene ligands, which was computed to be 11.3 kJ mol^–1 for [(C₅H₅)(CO)₂Fe (=C=C=CH₂)]⁺ [22, 26].

2.3 Investigation of the iron–carbon bonding by Mössbauer spectroscopy

In the carbon-chain ligands (L =:C(=C)_nRR’), the factors governing the metal–ligand interaction are electronic in character, as it was clearly established by studies on the relative orientation of the CR₂ plane with respect to the symmetry plane of the CpML₂ fragment and the determination of the rotation barrier for the unsaturated carbon fragment [22, 26]. The important orbitals in the description of the iron–carbon bonding are the empty π orbital, and the lone pair on the carbon fragment, which is invariably the HOMO. The characteristic feature of these ligands is the presence of one (carbene) or more π bonds between the metal and the terminal carbon, which presents a carbene character. The strength of the Fe–C π bond is the resultant of the balance between attractive and repulsive interactions that vary with the number of carbon atoms and the nature of the terminal R groups [22]. The usual spectroscopic properties like IR, ¹H and ¹³C NMR, or UV–Visible do not provide clear information on the metal-carbon bonding. The more reliable data on the metal carbon-rich ligand interaction have been obtained by theoretical calculations and X-ray determination [20].

Mössbauer spectroscopy can shed some light on the strength of the π-bonding to cumulenylidene ligands. Indeed, the isomeric shift of the spectrum (δ) provides a rough indication of the electronic density at the iron nucleus and the width of the quadrupole doublet (QS) depends on several parameters including the anisotropic electron distribution in the valence shell of the Mössbauer atom [31]. Therefore, one can expect that the variation of the QS value mainly depends on the Fe=C bond distance. The zero-field Mössbauer spectra of the complexes 1–8 were run at 80 K and the characteristic parameters were computed. Careful examination of the data collected in Table 3 provides interesting information: in this homogeneous series of complexes, the δ and QS parameters roughly increase according to the same sequence (3 < 4 < 5a < 7 < 6 < 5b < 1 < 2a). This observation suggests that the positive charge on the iron nucleus decreases with the Fe=C bond order or, in other words, with the increase of the distance between the iron and the α carbon atom. Such a result is fully consistent with the valence bond (VB) analysis of the electronic structure of metallacumulenylidenes, as depicted in Fig. 9. When the π acidity of the carbon-rich ligand decreases, the weight of the VB structure B become less negligible; concomitantly, the electronic density on the metal increases and the Fe=C bond order decreases. In addition, in the case of R groups that contain an heteroatom, the predominance of the VB structure C would occur, especially in the case of metallacumulene possessing an odd number of carbon atoms [20, 32, 33]. Interestingly, a significant linear correlation (R = 0.81) between the Mössbauer parameters δ and QS clearly shows the dependence of the electronic density on the iron centre and the Fe=C bond distance for these complexes (Fig. 10). This relationship indicates a trend, but must be considered with caution, even in a homogenous series. Indeed, the methoxyallenylidene 8 lies far above the correlation depicted in Fig. 10, suggesting that the Fe=C bond in this complex is much weaker than one can deduce from the electronic density on the iron nucleus.

An alternative means to investigate the effects of the structural variation of the ligand L on the metal–carbon bonding in this series of iron cumulenylidene derivatives is to correlate the QS values determined by Mössbauer spectroscopy with the Fe=C bond distances obtained by X-ray analyses. Fig. 11 displays a good linear dependence of these two terms (R = 0.99). However, considering that X-ray data are available for only the five complexes 1, 2a, 3, 4, and 5a, this observation has to be taken with caution. Nevertheless, it seems that for homogeneous series of metallacumulenes, the variation of the QS terms provides an indication on the strength of the π-bonding between the metal and the ligand. It is noteworthy that, in the limit of the accuracy of the X-ray determination, only 4 deviates from this linear correlation. Such a behaviour, which results from a shorter Fe=C bond distance that can be predicted from the QS value, could be related to the partial triple bond character of the Fe=C bond in the vinylidene derivatives, as pointed out by theoretical calculations [22].

Support on this conclusion can also be obtained by considering the iron alkyl complex (η⁵-C₅Me₅)(η²-dppe)Fe(CH₃) (9) for which there is no π-bonding between the metal and the alkyl fragment. The QS value determined for this complex, which possesses a Fe-C bond with a pure σ-character, can be regarded as the limit value expected when the π-bonding interaction does not exist. The Fe–C bond distance was not determined by X-ray analysis for complex 9, but it was computed by DFT method (2.075 Å) [34]. The correlation in Fig. 11 was complemented by the data obtained for compound 9. Interestingly, these values confirm the trend observed for the cumulene derivatives and the fit remained unchanged (R = 0.99). The empiric dependence between the Fe=C distances and the QS value is given by equation (1).

The Mössbauer spectroscopy constitutes a convenient mean to obtain a powerful characterization of the double bond between the iron centre and the carbon-chain ligands. It is important to note that for the given (η⁵-C₅Me₅)(η²-dppe)Fe metal fragment, the Mössbauer parameters are very sensitive to the structure of the cumulene ligand L. However, the great sensitivity of the Mössbauer spectroscopy to the nature of the counter anion must be also emphasized. Indeed, comparison of the data obtained for the couples of compounds 2a/2b and 5a/5b clearly shows that the effect induced on the Mössbauer response of the iron nucleus by the replacement of the PF₆^– counter anion by CF₃OSO₂^– or BPh₄^– can be much larger than the effect of the structural modifications on the cumulene fragment. Indeed, the quadrupole splitting (QS) results from the interaction between the nuclear quadrupole moment and the electric field gradient (EFG) at the nucleus. Distant ions which surround the Mössbauer atom clearly contribute to the total EFG, and consequently, anions with different volumes and symmetry could give rise to different QS values [31].

On the other hand, the relationship between the QS and δ parameters depicted in Fig. 10 shows that both the couples of complexes 2a, 2b and 5a, 5b fit very well with the other data of this empiric correlation. In addition, the absorption corresponding to the ν_C=C=C stretch is shifted to lower wavenumbers in 5b (1888 cm^–1) relative to 5a (1896 cm^–1), in accord with a weaker Fe=C bond in 5b than in 5a. The nature of the cation-anion interaction is not currently clear, but nevertheless these results suggest that the anion could contribute either to strengthen or to weaken the Fe=C double bonds in these cationic metallacumulenes. Examination of the X-ray crystal structures of 1, 2a, 3, 4, and 5a does not reveal any short distances between the fluorine atoms of the hexafluorophosphate anion and hydrogen atoms of the cationic moiety as recently evidenced in the organometallic compound [(C₅H₅)Fe(p-MeC₆H₄)–NHN=CH–C₆H₄–CH=CH–(C₅H₄)Fe(C₅H₅)]⁺PF₆^– [21]. In the absence of structural determination with anions different from the hexafluorophosphate, it is assumed that the influence of the anions is probably due to either Coulombic interactions or packing effects. According to the empiric relationship represented by equation (1), the replacement of the hexafluorophosphate by triflate or tetraphenyl borate in 2a or 5a could induce a lengthening of the Fe=C bond of ca 0.2 Å. In the absence of experimental structural data, this hypothesis still remains an open question.

4.1 General data

All manipulations were carried out under inert atmosphere. Solvents or reagents were used as follow: Et₂O, THF, and n-pentane, distilled from Na/benzophenone; CH₂Cl₂, distilled from CaH₂ and purged with argon; complexes [(η⁵-C₅H₅)₂Fe⁺][PF₆^–] [36], and (η⁵-C₅Me₅)Fe(η²-dppe)Cl (10) [37] were prepared by previously published procedures. High-field NMR spectra experiments were performed on a multinuclear Bruker 300 MHz or 200 MHz instrument (AM300WB and 200DPX). Chemical shifts are given in parts per million relative to tetramethylsilane (TMS) for ¹H- and ¹³C-NMR spectra, H₃PO₄ for ³¹P-NMR spectra. Transmittance–FTIR spectra were recorded using a Bruker IFS28 spectrometer (400–4000 cm^–1). Mössbauer spectra were recorded with a 2.5 × 10^–2 Ci (9.25 × 10⁸ Bq) ⁵⁷Co source using a symmetric triangular sweep mode [38]. LSI–MS analyses were performed at the ‘Centre régional de mesures physiques de l’Ouest’ (CRMPO, Rennes) on a high-resolution MS/MS ZabSpec TOF Micromass spectrometer (8 kV). Elemental analyses were performed at the Centre for Microanalyses of the CNRS at Lyon-Solaise, France.

4.2 [(η⁵-C₅Me₅)(η²-dppe)Fe(=C=C=C(Ph)Ph)][PF₆] (5a)

In a Schlenk tube, the green complex 10 (0.300 g, 0.48 mmol), KPF₆ (0.1 g, 0.54 mmol), methanol (30 ml) and 1,1-diphenylprop-2-yn-1-ol ( 0.110 g, 0.54 mmol) were introduced under argon. The mixture was stirred at 20 °C for 12 h. Evaporation of the solvent, extraction of the solid residue with methylene chloride, removal of the solvent and washing with n-pentane yielded the desired complex as a pure purple powder (0.420 g, 95%) after drying in vacuo. Crystals were grown by slow diffusion of diethyl ether in a dichloromethane solution of 5a. Anal. calcd for C₅₁H₄₉FeP₃F₆: C, 66.24; H, 5.34. Found: C, 66.75; H, 5.37. FT–IR (ν, KBr/Nujol, cm^–1) 1896 (m, =C=C=C). NMR ¹H (δ, CDCl₃, 25 °C) 7.84–6.81 (30H, H_Ar/dppe); 3.01, 2.55 (2 m, 4H, CH_2dppe); 1.46 (s, 15H, C₅(CH₃)₅). NMR ³¹P (δ, CDCl₃, 25 °C) 92.1 (s, dppe); –143 (sept, PF₆). NMR ¹³C (δ, CDCl_3, 25 °C) 285.0 (t, ²JCP = 37 Hz, Fe=$\underset{¯}{C}$=C=C); 224.2 (s, Fe=C=$\underset{¯}{C}$=C); 144.6 (s, Fe=C=C=$\underset{¯}{C}$); 135.1–128.3 (m, Ph); 101.8 (s, $\underset{¯}{C}$₅(CH₃)₅); 32.2, 31.3 (2m, –$\underset{¯}{C}$H_2dppe); 10.5 (q, ¹J_CH = 127 Hz, C₅($\underset{¯}{C}$H₃)₅).

4.3 [(η⁵-C₅Me₅)(η⁵-dppe)Fe(=C=C=C(Ph)Ph)][BPh₄] (5b)

In a Schlenk tube, the green complex (10) (0.500 g, 0.8 mmol) NaBPh₄ (0.301 g, 0.88 mmol), Amberlyst 15 (0.500 g), methanol (30 ml) and 1,1-diphenylprop-2-yn-1-ol (0.183 g, 0.88 mmol) were introduced under argon. The mixture was stirred at 20 °C for 12 h. After evaporation of the solvent, MgSO₄ was added before extraction of the solid residue with methylene chloride. Removal of the solvent and washing with n-pentane yielded the desired complex as a pure purple powder (0.800g, 91%). FT-IR (ν, KBr/Nujol, cm^–1) 1888 (m, =C=C=C). NMR ¹H (δ, CDCl₃, 25 °C) 7.31–6.87 (50H, H_Ar/dppe); 2.89, 2.57 (2 m, 4H, CH_2dppe); 1.42 (s, 15H, C₅(CH₃)₅). NMR ³¹P (δ, CDCl₃, 25 °C) 92.3 (s, dppe); –143 (sept, PF₆). NMR ¹³C (δ, CDCl_3, 25 °C) 284.8 (t, ²J_CP = 37 Hz, Fe=$\underset{¯}{C}$=C=C); 223.9 (s, Fe=C=$\underset{¯}{C}$=C); 144.5 (s, Fe=C=C=$\underset{¯}{C}$); 166.1–163.2, 136.8–121.9 (m, Ph); 101.6 (s, $\underset{¯}{C}$₅(CH₃)₅); 32.1, 31.2 (2m, –$\underset{¯}{C}$H_2dppe); 10.4 (q, ¹J_CH = 127 Hz, C₅($\underset{¯}{C}$H₃)₅).

4.4 [(η⁵-C₅Me₅)(η⁵-dppe)Fe[=C=C=C(Me)Ph][PF₆] (6)

In a Schlenk tube, the green complex (10) (0.500 g, 0.8 mmol), KPF₆ (0.162 g, 0.88 mmol), Amberlyst 15 (0.500 g), methanol (30 ml) and 1,1-diphenylprop-2-yn-1-ol (0.0 g, 0.88 mmol) were introduced under argon. The mixture was stirred at 20 °C for 12 h. After evaporation of the solvent, MgSO₄ was added before extraction of the solid residue with methylene chloride. Removal of the solvent and washing with n-pentane yielded the desired complex as a pure purple powder (0.800 g, 91%). FT–IR (ν, KBr/Nujol, cm^–1) 1913 (m, =C=C=C). NMR ¹H (δ, CDCl₃, 25 °C) 7.60–7.25 (25H, H_Ar/dppe); 3.09, 2.63 (2 m, 4H, CH_2dppe); 1.70 (s, 3H, Me), 1.50 (s, 15H, C₅(CH₃)₅). NMR ³¹P (δ, CDCl₃, 25 °C) 92.1 (s, dppe); –143 (sept, PF₆). NMR ¹³C (δ, CDCl_3, 25 °C) 289.9 (t, ²J_CP = 38 Hz, Fe=$\underset{¯}{C}$=C=C); 218.2 (s, Fe=C=$\underset{¯}{C}$=C); 143.2 (s, Fe=C=C=$\underset{¯}{C}$); 148.5–126.6 (m, Ph); 101.0 (s, $\underset{¯}{C}$₅(CH₃)₅); 32.2–31.4 (m, –$\underset{¯}{C}$H_2dppe); 30.3 (q, ¹J_CH = 130 Hz, Me); 10.5 (q, ¹J_CH = 128 Hz, C₅($\underset{¯}{C}$H₃)₅).

4.5 [(η⁵-C₅Me₅)(η⁵-dppe)Fe(=C=C=C(Me)Et)][PF₆] (7)

In a Schlenk tube, the green complex (10) (0.500 g, 0.8 mmol) NaBPh₄ (0.301 g, 0.88 mmol), Amberlyst 15 (0.500 g), methanol (30 ml) and 3-methylpent-1-yn-3-ol (0.086 g, 0.88 mmol) were introduced under argon. The mixture was stirred at 20 °C for 12 h. After evaporation of the solvent, MgSO₄ was added before extraction of the solid residue with methylene chloride. Removal of the solvent and washing with n-pentane yielded the desired complex as a pure purple powder (0.590 g, 91%). FT-IR (ν, KBr/Nujol, cm^–1) 1931 (m, =C=C=C). NMR ¹H (δ, CDCl₃, 25 °C) 7.59–7.14 (20H, H_dppe); 3.01, 2.55 (2 m, 4H, CH_2dppe); 1.79 (q, ³J_HH = 7 Hz, 2H, C$\underset{¯}{H}$₂–CH₃); 1.47 (s, 3H, Me); 1.41 (s, 15H, C₅(CH₃)₅); 0.86 (t, ³J_HH = 7 Hz, 3H, CH₂C$\underset{¯}{H}$₃). NMR ³¹P (δ, CDCl₃, 25 °C) 92.7 (s, dppe); –143 (sept, PF₆). NMR ¹³C (δ, CDCl_3, 25 °C) 301.9 (t, ²J_CP = 37 Hz, Fe=$\underset{¯}{C}$=C=C); 208.4 (s, Fe=C=$\underset{¯}{C}$=C); 162.3 (s, Fe=C=C=$\underset{¯}{C}$); 135.0–128.6 (m, Ph); 100.0 (s, $\underset{¯}{C}$₅(CH₃)₅); 40.1 (t, ¹J_CH = 126 Hz, $\underset{¯}{C}$H₂CH₃); 32.8 (q, ¹J_CH = 126 Hz, CH₃); 31.2 (2m, –$\underset{¯}{C}$H_2dppe); 10.2 (q, ¹J_CH = 128 Hz, C₅($\underset{¯}{C}$H₃)₅). It was observed by ¹H and ³¹P NMR spectroscopy that 7 slowly decomposes in (CD₃)₂CO, CD₂Cl₂ and CDCl₃.

4.6 Crystallography

Data were collected on crystals of 3, 4·C₄H₁₀O and 5a·C₄H₁₀O as summarized in Table 1 [39, 40]. Cell constant and orientation matrix were obtained from a least-squares refinement using 25 high-θ reflections. After Lorentz and polarization corrections [41] and absorption corrections (φ scans), the structure was solved with SIR-97 [42], which revealed the non-hydrogen atoms and the solvate molecules. After anisotropic refinements, a Fourier difference map revealed many hydrogen atoms. Atomic scattering factors were taken from the literature [43]. ORTEP views were generated with PLATON-98 [44]. All calculations were performed on a Pentium NT Server computer.

4.7 Supplementary material

The supplementary material has been sent in electronic format to the Cambridge Crystallographic Data Centre, 12 Union Road, Cambridge CB2 1EZ, UK as cif file No. CCDC 197018–97020, and can be obtained by contacting the CCDC.