Synthesis of Aromatic Aza-metallapentalenes from Metallabenzene via Sequential Ring Contraction/Annulation

The concept of aromaticity has long played an important role in chemistry and continues to fascinate both experimentalists and theoreticians. Among the archetypal aromatic compounds, heteroaromatics are particularly attractive. Recently, substitution of a transition-metal fragment for a carbon atom in the anti-aromatic hydrocarbon pentalene has led to the new heteroaromatic osmapentalenes. However, construction of the aza-homolog of osmapentalenes cannot be accomplished by a similar synthetic manipulation. Here, we report the synthesis of aza-osmapentalenes by sequential ring contraction/annulation reactions of osmabenzenes via osmapentafulvenes. Nuclear magnetic resonance spectra, X-ray crystallographic analysis, and DFT calculations all suggest that these aza-osmapentalenes exhibit aromatic character. Thus, the stepwise transformation of metallabenzenes to metallapentafulvenes and then aza-metallapentalenes provides an efficient and facile synthetic route to these bicyclic heteroaromatics.

ered as analogs of a pentalene dianion, and these molecules are actually aromatic 22 . The isolation of the metallapentalene prompted us to investigate whether an aza-metallapentalene exists and, if so, how the synthesis could be achieved. As shown in Fig. 1, the structure of an aza-metallapentalene resembles that of pentalene, which is generally considered to be an anti-aromatic molecule, prompting us to consider whether an aza-metallapentalene, which is structurally and electronically analogous to metallapentalene, is also aromatic. In this paper, we present the synthesis and characterisation of azametallapentalenes, which were obtained from metallabenzenes via sequential ring contraction/annulation reactions.

Results
Synthesis of the first aza-metallapentalene. The recently reported osmapentalene 19,20 was synthesised (derived) from the protonation of osmapentalyne (Fig. 2a). Experimentally, osmapentalyne 23 was observed to be thermally stable and easy to prepare via the reaction of osmium complexes with terminal alkynes (Fig. 2a). Inspired by this observation, we initially attempted to react osmium complexes with a number of nitriles such as benzonitrile or acetonitrile in the hope of obtaining the corresponding azaosmapentalynes. However, the expected reactions did not occur, only osmabenzenes or other osmacycles were detected, which have been previously reported 24,25 . Therefore, the preparation of an azametallapentalene via the synthesis of an aza-osmapentalyne followed by protonation is unfeasible.
Previously, we synthesised osmacyclopentadienes, each bearing an exocyclic C 5 N bond, by reacting an osmabenzene complex with amines 26 (Fig. 2b). We envisioned that an aza-osmapentalene might be directly obtained by introducing a new carbon atom, with the intention of inducing a further annulation reaction yielding an osmacyclopentadiene monocyclic system, as illustrated by the retrosynthetic analysis shown in Fig. 2b.
We first modified an osmabenzene complex by ligand substitution to generate the more stable osmabenzene 1-I. The structure of osma-benzene 1-I was verified by X-ray diffraction analysis ( Fig. 3b; the relevant experimental details are presented in Supplementary  Figures and Methods). Subsequent treatment of 1-I with aniline in the presence of sodium hexafluorophosphate afforded the expected osmacyclopentadiene 2-PF 6 in 95% yield (Fig. 3a). The complex 2-PF 6 was characterised by multinuclear nuclear magnetic resonance (NMR) spectroscopy, single-crystal X-ray diffraction analysis, and high-resolution mass spectroscopy. The structure of 2-PF 6 is shown in Fig. 3c. The X-ray diffraction analysis demonstrated that 2-PF 6 contains a coplanar metallacyclopentadiene ring with an exocyclic imino group. The structural parameters indicated that the metallacycle of 2-PF 6 could be represented by two resonance structures, 2-PF 6 and 2b-PF 6 ( Fig. 3a), with 2-PF 6 as the more dominant structure. The structural data (averaged bond lengths) also indicated that both the metallacycles in 1-I and 2-PF 6 have delocalised structures. In addition, the six atoms Os1 and C1-C5 in each complex are approximately coplanar, which is reflected by their small mean deviation (0.0129 Å for 1-I; 0.0331 Å for 2-PF 6 ) from the least-squares plane.
As shown in Fig. 3a, the resonance structure 2b-PF 6 (metallapentafulvene) contributes to the overall structure of the complex 2-PF 6 . Since pentafulvene molecules 27 are normally considered aromatic, we assume that the metallapentafulvene 2-PF 6 is also aromatic, as is indeed supported by our density functional theory (DFT) results (vide infra).
According to the retrosynthetic analysis shown in Fig. 2b, the complex 2-PF 6 could be used as the precursor to afford the desired aza-metallapentalene by introducing an additional carbon atom. To test this idea, reactions of 2-PF 6 with the propynols RCH(OH)C ; CH in the presence of Ag 2 O were carried out in the hope of obtaining the desired aza-osmapentalene complexes. Triethylamine was used in the reactions to ensure that an acetylide ligand could be easily delivered to the osmium metal centre (Warning: the reaction may proceed through the formation of silver acetylide, which decomposes violently on contact with moisture and water producing highly flammable and explosive acetylene gas, and causing fire and explosion   hazard). As shown in Fig. 4a, treatment of 2-PF 6 with phenylpropynol in the presence of Ag 2 O and triethylamine under reflux for approximately 3 h led to the formation of complex 3a-PF 6 , which was isolated in 81% yield. The complex 3a-PF 6 was characterised by NMR spectroscopy and HRMS, and its structure was determined by X-ray crystal structure analysis.
Complex 3a-PF 6 exhibits an essentially planar metal-bridged bicyclic structure (Fig. 4b). The mean deviation from the leastsquares plane through Os1, N1 and C1-C6 is 0.0383 Å . The Os1-C6 bond length (2.180(9) Å ) is appreciably longer than both the Os1-C1 bond length (2.040(8) Å ) and the Os1-C4 bond length (2.100(7) Å ). The three Os-C bond lengths of the metallabicycle all lie in the high end of the reported range for typical Os-C(vinyl) 28 bond lengths (1.859-2.359 Å ) (on the basis of a search of the Cambridge Structural Database, CSD version 5.35, conducted in November, 2013). In addition, the structural parameters indicate a considerable bond distance alternation within the two fused fivemembered rings. Therefore, the complex 3a-PF 6 is not the desired aza-osmapentalene, although the osmabicyclic framework is consistent with the suppositional aza-osmapentalene formula.
Thus, using the strategy outlined in Fig. 2b, we successfully synthesised the first aza-metallapentalene complex in three steps starting from the six-membered osmabenzene complex 1-I. Further investigations demonstrated that this synthetic strategy could be extended to prepare other aza-osmapentalene complexes with other substituents on the metallacycle. The vinyl-containing aza-osmapentalene 4b-(PF 6 ) 2 was obtained from the reaction of 2-PF 6 with propynol ( Fig. 4a) and was isolated as a brown solid. Similarly, the butenynylcontaining aza-osmapentalene 4c-(PF 6 ) 2 was obtained from the reaction of 2-PF 6 with penta-1,4-diyn-3-ol and was isolated as a dark-brown solid (Fig. 4a).
As previously mentioned, X-ray crystallography revealed a planar structure with no alternation in the carbon-carbon bond lengths of the osmabicyclic rings in the aza-osmapentalene 4a-(PF 6 ) 2 . Protons on the periphery of aromatic compounds are well known to have relatively large downfield NMR spectroscopic chemical shifts due to a diamagnetic ring current. Fig. 5a shows a comparison of the chemical shifts of three specific protons among four different metallacyclic complexes. Compared with the structurally similar complex 3a, complex 4a gave considerable downfield chemical shifts because of its aromatic character. The 1 H NMR chemical shifts as well as the X-ray diffraction data indicate that the metallacycle in the cation of 4-(PF 6 ) 2 has a delocalised structure, with contributions from the six resonance structures 4 to 4E, as shown in Fig. 5b. Consistent with their aromaticity, all of the new aza-osmapentalenes exhibit remarkably high thermal stability. Solid samples of 4-(PF 6 ) 2 were heated in air at 160uC for at least 5 h without noticeable decomposition. 4-(PF 6 ) 2 was stable even at 80uC in 1,2-dichloro-ethane solution for 5 h. These new aza-osmapentalenes each contain a phosphonium substituent, which is also believed to partially contribute to the high stability 29 .
DFT calculations of osmabenzene, osmacyclopentadiene and azaosmapentalene. To gain more insight into the electronic structure and aromaticity of the aza-osmapentalenes 4, we performed density functional theory (DFT, B3LYP/6-31111G(d,p)) calculations. The nucleus-independent chemical shift (NICS) 30  previously reported values for osmapentalenes 19,20 . The negative NICS values and the significant ASE values calculated for these model complexes indicate that aromaticity is closely associated with the metallacycle rings in complexes 1, 2 and 4a.
Additional evidence for the aromatic nature of aza-osmapentalene 4 was provided by examination of the molecular orbitals calculated for the further simplified model complex 4a0. p Electrons are normally responsible for aromatic behaviour. Therefore, our attention is paid to those occupied orbitals. As shown in Fig. 7a, the five occupied p molecular orbitals (MOs) calculated for 4a0 (HOMO-1, HOMO-2, HOMO-3, HOMO-7 and HOMO-18) reflect the p-delocalisation along the perimeter of the bicyclic system. HOMO-1, HOMO-3 and HOMO-18 are derived from the orbital interactions between the p p atomic orbitals of the C 6 NH 6 unit (perpendicular to the bicycle plane) and the Os 5d xz orbital, whereas HOMO-2 and HOMO-7 are formed by the orbital interactions between the p p atomic orbitals of the C 6 NH 6 unit and the Os 5d yz orbital. The approximately uniform contribution of p p orbitals from ring atoms in each of these occupied p molecular orbitals is typical for and consistent with a delocalised p molecular system. The four occupied p MOs calculated for the antiaromatic azapentalene are also presented in Fig. 7b for comparison.

Conclusion
We synthesised the first examples of aza-metallapentalenes using metallabenzenes as the starting material. The synthetic route to these aza-metallapentalenes is particularly interesting and remarkable. The bicyclic eight-membered ring complexes are derived from sequential ring contractions, followed by annulation reactions of the six-membered metallabenzenes via the five-membered metallapentafulvenes. Both experimental and theoretical studies suggest that the metallabenzenes, metallapentafulvenes and aza-metallapentalenes all exhibit aromatic character. The present results not only expand the aromaticity concept but also open a promising avenue for the construction of new bicyclic metallaaromatics containing main-group heteroatoms.
Synthesis of osmapentafulvene complex 2-PF 6 . Aniline (60 mL, 0.66 mmol) was added to a mixture of 1-I (300 mg, 0.21 mmol) and sodium hexafluorophosphate (71 mg, 0.42 mmol) in CH 2 Cl 2 (25 mL). The mixture was heated at reflux for approximately 10 h to afford a purple suspension. The solid suspension was removed by filtration, and the volume of the filtrate was reduced to approximately 2 mL under vacuum. The addition of ether (20 mL) to the solution then produced a purple solid that was collected by filtration, washed with diethyl ether (3 3 2 mL), and dried under vacuum. Yield: 282 mg, 95%. 1  Synthesis of osmabicyclic complex 3a-PF 6 . A mixture of (Ph)CH(OH)C ; CH (40 mg, 0.30 mmol) Ag 2 O (60 mg, 0.26 mmol), NEt 3 (1 mL) and 2-PF 6 (365 mg, 0.25 mmol) in dichloromethane (30 mL) was heated at reflux for 3 h to afford a blue suspension (Warning: the reaction may proceed through the formation of silver acetylide, which decomposes violently on contact with moisture and water producing highly flammable and explosive acetylene gas, and causing fire and explosion hazard). The solvent was removed under vacuum, and the residue was extracted with CH 2 Cl 2 (3 3 5 mL). The volume of the filtrate was filtered through a Celite pad to remove the silver salt and was subsequently reduced to approximately 2 mL under vacuum. Diethyl ether (20 mL) was added slowly with stirring to afford a blue solid, which was collected by filtration, washed with hexane (3 3 3 mL), and dried under vacuum. Yield: 296 mg, 81%. 1   Synthesis of aza-osmapentalene complex 4a-(PF 6 ) 2 . A mixture of 3a-PF 6 (200 mg, 0.14 mmol) and sodium hexafluorophosphate (28 mg, 0.17 mmol) in CH 2 Cl 2 (10 mL) was stirred at room temperature for 30 min to afford a red suspension. The solid suspension was removed by filtration, and the volume of the filtrate was reduced X-ray Crystallographic Analysis. Single crystals suitable for X-ray diffraction were grown from a dichloromethane solution layered with hexane. Diffraction data were collected on an Oxford Gemini S Ultra charge-coupled device (CCD) area detector (2-PF 6 , 3a-PF 6 , 4a-(PF 6 ) 2 and 4b-(PF 6 ) 2 ) or on a Bruker Apex CCD area detector (1-I) using graphite-monochromated Mo Ka radiation (l 5 0.71073 Å ) or Cu Ka radiation (l 5 1.54178 Å ). Semi-empirical or multi-scan absorption corrections (SADABS) were applied 40 . All structures were solved by the Patterson function, completed by subsequent difference Fourier map calculations, and refined by fullmatrix least-squares on F 2 with all the data using the SHELXTL program package 41 . All non-hydrogen atoms were refined anisotropically unless otherwise stated. Hydrogen atoms were placed at idealised positions and were refined using a riding model. See Supplementary Methods for detailed crystal data related to complexes 1-I, 2-PF 6 , 3a-PF 6 , 4a-(PF 6 ) 2 and 4b-(PF 6 ) 2 .
Computational details. All structures were optimised at the B3LYP level of DFT [42][43][44] . In addition, the frequency calculations were performed to confirm the characteristics of the calculated structures as minima. In the B3LYP calculations, the effective core potentials (ECPs) of Hay and Wadt with a double-f valence basis set (LanL2DZ) were used to describe the Os, P, and I atoms, whereas the standard 6-31111G(d,p) basis set was used for the C and H atoms 45