Trimerisation of carbon suboxide at a di-titanium centre to form a pyrone ring system

Bis(pentalene)dititanium Ti2(μ:η5,η5-Pn†)2 trimerises carbon suboxide (O 
<svg xmlns="http://www.w3.org/2000/svg" version="1.0" width="16.000000pt" height="16.000000pt" viewBox="0 0 16.000000 16.000000" preserveAspectRatio="xMidYMid meet"><metadata>
Created by potrace 1.16, written by Peter Selinger 2001-2019
</metadata><g transform="translate(1.000000,15.000000) scale(0.005147,-0.005147)" fill="currentColor" stroke="none"><path d="M0 1440 l0 -80 1360 0 1360 0 0 80 0 80 -1360 0 -1360 0 0 -80z M0 960 l0 -80 1360 0 1360 0 0 80 0 80 -1360 0 -1360 0 0 -80z"/></g></svg>
 C 
<svg xmlns="http://www.w3.org/2000/svg" version="1.0" width="16.000000pt" height="16.000000pt" viewBox="0 0 16.000000 16.000000" preserveAspectRatio="xMidYMid meet"><metadata>
Created by potrace 1.16, written by Peter Selinger 2001-2019
</metadata><g transform="translate(1.000000,15.000000) scale(0.005147,-0.005147)" fill="currentColor" stroke="none"><path d="M0 1440 l0 -80 1360 0 1360 0 0 80 0 80 -1360 0 -1360 0 0 -80z M0 960 l0 -80 1360 0 1360 0 0 80 0 80 -1360 0 -1360 0 0 -80z"/></g></svg>
 C 
<svg xmlns="http://www.w3.org/2000/svg" version="1.0" width="16.000000pt" height="16.000000pt" viewBox="0 0 16.000000 16.000000" preserveAspectRatio="xMidYMid meet"><metadata>
Created by potrace 1.16, written by Peter Selinger 2001-2019
</metadata><g transform="translate(1.000000,15.000000) scale(0.005147,-0.005147)" fill="currentColor" stroke="none"><path d="M0 1440 l0 -80 1360 0 1360 0 0 80 0 80 -1360 0 -1360 0 0 -80z M0 960 l0 -80 1360 0 1360 0 0 80 0 80 -1360 0 -1360 0 0 -80z"/></g></svg>
 C 
<svg xmlns="http://www.w3.org/2000/svg" version="1.0" width="16.000000pt" height="16.000000pt" viewBox="0 0 16.000000 16.000000" preserveAspectRatio="xMidYMid meet"><metadata>
Created by potrace 1.16, written by Peter Selinger 2001-2019
</metadata><g transform="translate(1.000000,15.000000) scale(0.005147,-0.005147)" fill="currentColor" stroke="none"><path d="M0 1440 l0 -80 1360 0 1360 0 0 80 0 80 -1360 0 -1360 0 0 -80z M0 960 l0 -80 1360 0 1360 0 0 80 0 80 -1360 0 -1360 0 0 -80z"/></g></svg>
 O) to form [{Ti2(μ:η5,η5-Pn†)2}{μ-C9O6}], which contains a 4-pyrone core, via the monoadduct [Ti2(μ:η5,η5-Pn†)2 (η2-C3O2)].


Introduction
Unlike the plethora of catalytic and stoichiometric transformations of carbon's most common oxides (i.e. CO and CO 2 ) promoted by well-dened molecular complexes, [1][2][3][4] there is a disproportionate lack of examples featuring the activation and subsequent transformation of the sub-oxides of carbon. C 3 O 2 is the rst in the series of the synthetically available carbon suboxides featuring an odd number of carbons 5,6 (predicted to augment their stability 7 ), and its spectroscopic 8 and physical properties 9,10 have been extensively studied. Its molecular structure in the solid state has been reported and shows a linear structure, 11 whereas in the gas phase computational 12 and spectroscopic studies 13 conrm a bent structure with a bond angle of 156 . Similarly, aspects of its reactivity with a variety of organic substrates 14 (e.g. ylides [15][16][17] ) and main-group 18 bonds have been reported since its rst synthesis. Although C 3 O 2 (hereaer referred to as carbon suboxide) is relatively unstable (it auto-polymerises above 0 C but can be stored indenitely below À35 C), it is moderately straightforward to prepare via the dehydration of malonic esters 5 or malonic acid 19 with phosphorus pentoxide. The polymer produced by its selfpolymerisation has a band-like structure with condensed apyrone rings and has been studied for its electronic properties. 20,21 Carbon suboxide is also formed in small quantities in vivo during biochemical processes that normally produce carbon monoxide, for example, during heme oxidation by heme oxygenase-1 (HO-1). It is then rapidly oligomerised into macrocyclic structures, predominantly cyclic hexamers and octamers ( Fig. 1), which contain fused 4-pyrone rings and are potent inhibitors of Na + /K + -ATP-ase and Ca-dependent ATP-ase; larger carbon suboxide based macrocycles are proposed to be natriuretic and endogenous digitalis like factors (EDLFs). [22][23][24] In terms of coordination chemistry, it was proposed that the thermal decomposition of Ag 3 C 3 O 2 to produce C 3 O 2 involved a coordination complex of Ag, 25 and subsequent studies of the reactivity of C 3 O 2 towards Pt(0), Pt(II) and Rh(I) complexes by Pandolfo et al. proposed the formation of C 3 O 2 complexes but lack of structural data plagued these early investigations. 26 Nevertheless, later studies from the same group 27,28 as well as that of Hillhouse 29 demonstrated some aspects of the reactivity of C 3 O 2 with organometallic fragments by isolating, for example, the products of its insertion into M-H bonds. A main problem of these early studies was the propensity of C 3 O 2 to act as a source of ketene (:C]C]O) and CO. Thus, in the presence of phosphorous containing ligands in the coordination sphere of the metal centre, this led to the formation of the corresponding phosphorous-ylides, as shown by Hillhouse et al. by the reaction of C 3 O 2 with WCl 2 (PMePh 2 ) 4 furnishing WCl 2 (CO)(PMePh 2 ) 2 {C,C 0 :h 2 -C(O)CPMePh 2 }. 30 List and Hillhouse further showed that C 3 O 2 can displace COD (COD ¼ 1,5-cyclooctadiene) in (PPh 3 ) 2 Ni(COD) to yield (PPh 3 ) 2 Ni{C,C 0 :h 2 -C 3 O 2 } where the C 3 O 2 ligand coordinates via the central and one of the terminal carbons, 31 although this could not be conrmed crystallographically. Gas phase spectroscopic 32 and theoretical 33 studies on the bonding of C 3 O 2 to late transition metal centres have also been reported, but, and to the best of our knowledge, there have been no other reports investigating the interaction of C 3 O 2 with organometallic or other coordination compounds. Undoubtedly one of the reasons is its capricious nature, which has favored in silico studies of its reactivity especially towards transition metals. 34,35 Indeed, C 3 O 2 is one of the least explored 'small molecules' from a synthetic chemist's point of view, a fact underlined by only two short reviews in the current literature. 14, 36 We have previously reported on the synthesis, 37 and diverse reactivity 38-41 of the syn-bimetallic complex [Ti 2 (m:h 5 ,h 5 -Pn † ) 2 ] (Pn † ¼ C 8 H 4 (Si i Pr 3 ) 2 ) (1) towards CO, CO 2 , and heteroallenes and therefore envisioned that (1) might be a good candidate for the binding and activation of C 3 O 2 . Herein we present the unprecedented trimerization of C 3 O 2 promoted by (1), Fig. 2, as well as experimental and computational investigations into the mechanism of this reaction.

Results and discussion
Exposure of a crimson-red toluene solution of (1) to C 3 O 2 at À78 C, instantly produced a homogeneous brown solution which, upon warming to À35 C and then slowly to room temperature, deposited some C 3 O 2 polymer, together with a brown supernatant. Filtration of the reaction mixture and work up of the ltrate afforded a brown-green solid, which was isolated in moderate to good yields (yields are dependent on the nal temperature of the solution and vary between 40 and 66%), and proved to be a diamagnetic, spectroscopically pure new compound (2). The 1 H-NMR spectrum of ( Fig. 3) (2) consisted of 16 doublets in the aromatic region signifying the formation of a dimer exhibiting four inequivalent pentalene environments; this was further substantiated by the observation of eight peaks in the 29 Si{ 1 H}-NMR spectrum of (2).
The 13 C{ 1 H}-NMR spectrum of (2) displayed 41 resonance in the region between 389-96 ppm, 32 of which were assigned to the four inequivalent pentalene environments (our empirical observation is that resonances associated with this type of pentalene ligand scaffold in the m:h 5 ,h 5 coordination geometry appear in the region between 90 and 145 ppm (ref. [38][39][40][41]), with the nine remaining signals found in the downeld part of the spectrum (150-400 ppm) and which corresponded to quaternary carbons. At this point, it is interesting to note that such high eld resonances (300-400 ppm region) in 13 C-NMR spectra have been observed for complexes of Zr(IV) and Th(IV) featuring dihaptoacyl ligands with substantial oxy-carbene character. 42,43 The IR spectrum (thin lm) of (2) showed a strong absorption at 2061 cm À1 characteristic of a C]C]O moiety along with bands at 1658, 1591 and 1532 cm À1 characteristic of carbonyl functionalities, but also in agreement with haptoacyl ligands with a strong oxo-carbene character. 42,43 An X-ray diffraction study revealed the molecular structure of this new complex (2) (Fig. 4), which is consistent with the solution NMR data discussed above, and unequivocally demonstrates the rst example of the  The Ti-Ti bonds in (2) (2.4648(14) and 2.4834(16)Å) are retained but have been slightly elongated in comparison to that in (1) (2.399(2)Å). 37 As can be seen from Fig. 4, two titanium centres (Ti3 and Ti2) bind to two CO moieties in an h 2 fashion. This bonding mode is best described as an haptoacyl with a considerable carbenoid contribution to the resonance structure. We base this on the observed metrics of the corresponding bond lengths and angles (Ti2-C1: 2.17 (3) (2) is that the pyrone 6membered ring is not planar (deviates from planarity by 0.055 A, Fig. 6), and therefore lacks aromaticity. As a result, the C-C bond lengths of this 6 member ring are elongated in comparison to the ones found in 4-pyrone, 45 although the C]O bond distances (i.e. C6-O4: 1.207(18)Å vs. 1.253(12)Å in 4-pyrone) are similar within esd's. Unfortunately, due to the mixed occupancy of the CCO moiety and O6 over the two sides of the [C 9 O 6 ] core in (2) and the resulting crystallographic restraints used to model this disorder, we cannot talk with certainty about the bond lengths and angles of these two ligating moieties to this 6member ring. Nevertheless, upon inspection of the corresponding bond lengths of these two atoms to the Ti centres, we can deduce that the bonding situation is far from straightforward. For instance, the Ti1-C 0 bond resembles the ones found in Ti-NHC complexes, although closer to the high end of the spectrum (2.2-2.35Å), 46 and is in the same range as the ones discussed for the oxy-carbene moieties discussed above. A comparison with the corresponding lengths and angles found for free C 3 O 2 11 shows that the C 0 -C9 bond is elongated (1.2475(15)Å in free C 3 O 2 ) while the C-O bond remains unchanged (1.442(13)Å in free C 3 O 2 ). The same trend (i.e. C-C elongation) applies when compared with the corresponding bond lengths found in ketene (C]C: 1.314Å, C]O: 1.162Å). 47 Similarly, the O6-Ti4 bond distance is closer to the ones found in the C]O-Ti dative interaction, e.g. in [cis-Ti(OEt 2 ) 2 (h 2maltolato) 2 ]. 48 (2) is diamagnetic and the elongation of the Ti-Ti bonds might suggest an increase of the formal oxidation state of each Ti centre by one (i.e. Ti(II) in (1) to Ti(III) in (2)). However, the oxy-carbene character of the C7-O5 and C1-O1 units as well as the non-aromatic 6 membered heterocycle of the [C 9 O 6 ] core  suggest that complicated resonance structures are in play, and assignment of formal oxidation state and Ti-Ti bond order in (2) is therefore problematic.
In order to gain a better understanding of the formation of (2), the reaction was probed computationally. Density functional calculations using the ADF program suite (BP86/TZP) were carried out on model systems in which the Si i Pr 3 groups on the pentalene ligands were replaced by H atoms to increase computational efficiency. The computational analogues of experimental structures are denoted by italics; calculations on the analogue of the starting material 1, Ti 2 (C 8 H 6 ) 2 1, have been described previously. 37,40 Geometry optimisation of the possible addition product of the rst molecule of C 3 O 2 to 1, Ti 2 (C 8 -H 6 ) 2 (C 3 O 2 ), led to two local minima, 3 and 3 0 (Fig. 7).
Isomer 3 was the more stable being lower in energy by 0.59 eV, and resembled more closely the structure inferred from the disordered X-ray data (vide infra). The alternative structure, 3 0 , could possibly be formed as a kinetic product. The structure of 3 clearly suggests that formation of 2, with two C atoms and one O atom bound to the two Ti atoms, proceeds by the le hand side of the molecule depicted in Fig. 5 being the initial product rather than the right hand side where only one C atom and two O atoms are bound to the two Ti atoms.
The coordination mode of C 3 O 2 proposed here differs from some others calculated which indicate bonding primarily to the central carbon. In the cases of metal carbonyls 32 and AuCl 33 for example the metals function primarily as electron pair acceptors whereas Ti 2 Pn 2 has very high energy electrons and acts as a electron pair donor through its Ti-Ti bond. [37][38][39][40][41] The HOMO of C 3 O 2 has electron density on the central C (Fig. 8a), hence this carbon is preferred for bonding by a Lewis acid, whereas the LUMO is localised on the outer carbons (Fig. 8b) leading to asymmetrical bonding by a Lewis base. Bending C 3 O 2 to the geometry calculated for 3 concentrates the LUMO on an outer carbon (Fig. 8c).
The HOMO of 3 is a largely unperturbed Ti-Ti bonding orbital (Fig. 8d). The HOMO-1 and HOMO-2 correspond to TiPn bonding orbitals. The HOMO-3 (Fig. 8e) is responsible for C 3 O 2 binding and is formed by donation from the other Ti-Ti bonding orbital into the LUMO of bent C 3 O 2 . A suitable analogy for the bonding situation in (3) can be found in that described for the adduct of (1) with CO 2 [Ti 2 (m:h 5 ,h 5 -Pn † ) 2 (m-CO 2 )] (6) which has been studied computationally due to the instability of (6) in solution (one can conceptually replace the C 2 C 3 O 2 moiety with O and vice versa). 40 Indeed, a computational analysis of the orbitals involved in the coordination of C 3 O 2 in (3) reveals a similar picture to the one found in (6). The Ti-O distance in 3 (2.19Å) is shorter than that of 6 (2.27) indicating increased donation to O. This analogy between (3) and (6) is further reected by the short Ti-Ti bond distances that are characteristic of both these computational models. It should be noted that the HOMO-3 retains Ti-Ti bonding character hence there is only a slight lengthening of Ti-Ti distance from 1 to 3 (2.37Å to 2.41Å). A CBC 49 representation of 3 has an arrow going from the Ti]Ti double bond to C 3 O 2 acting as a Z ligand, in the same way as CO 2 behaves in [Ti 2 (m:h 5 ,h 5 -Pn † ) 2 (m-CO 2 )]. 40 The computed bond distances (ESI Table S2 ‡) of the coordinated C 3 O 2 in this model are in good agreement with the ones determined crystallographically vide infra.
In order to examine the energetics for the formation of 2, and to investigate possible intermediates in the reaction, the geometries of Ti 2 (C 8 H 6 ) 2 (C 3 O 2 ) 2 , 4, Ti 2 (C 8 H 6 ) 2 (C 3 O 2 ) 3 , 5, and 2 were optimised (Fig. 9). Key bond lengths for all calculated species are given in the ESI (Table S2 ‡).
The free energies for possible reaction pathways are shown in Fig. 10, and activation energies, where identied, are given in italics. The barriers to C 3 O 2 and to 5 reacting with 1 appear to be purely entropic.   The calculated mechanism highlights two key steps in the formation of (2): (a) the formation of the intermediate [Ti 2 (m:h 5 ,h 5 -Pn † ) 2 (m:h 2 ,h 1 -OCC 0 CO)] (3) (i.e. the adduct between (1) and C 3 O 2 that would arise from the addition of 1 eq. of the latter to the former) and (b) the termination of the consecutive two additions of C 3 O 2 to (3) by the capping of (5) by (1) or alternatively (c) the reaction of (4) with (3). The barriers for the two pathways, (b) and (c), are too similar to distinguish between them energetically. In all possible pathways leading to (2), the formation of adduct (3) is the common denominator. The high activation energy calculated for the reaction of (3) with a further molecule of C 3 O 2 to form (4) indicates that (3) should be isolable at low temperature. Indeed, repeating the reaction in the same manner as for the synthesis of (2), but removing volatiles at ca. 0 C, resulted in the formation of no carbon suboxide polymer and 1 H-NMR analysis showed the formation of an extra species along with (2), exhibiting two inequivalent pentalene ligand scaffolds (i.e. 8 doublets in the aromatic region). Encouraged by this observation, the reaction between (1) and C 3 O 2 was repeated under higher dilution conditions to prevent the last step of the formation of (2) and the reaction mixture was kept below À10 C throughout. Upon removing volatiles at low temperature (ca. À25 C), and lyophilising the residue with benzene (below À10 C), this new species was isolated in almost quantitative yields and with spectroscopic purity of >98%. More conclusive evidence that (3) is indeed that predicted by calculations was provided by 13 (3) is found at much lower eld (7.03 ppm) compared with the ones assigned to the terminal carbons (see above) and follows the trend observed in previous studies (À12.3 and À16.2 ppm for [M(h 2 (C,C 0 )-C 3 O 2 )(PPh 3 ) 2 ] with M ¼ Ni, Pt respectively); it has to be noted though that is shied downeld compared to these reported values. These discrepancies are expected as the documented examples concern electron rich monometallic metal fragments of d 10 transition metals, unlike the present case where a syn-bimetallic Ti-Ti core is involved. The coordination of C 3 O 2 was further corroborated by IR spectroscopy (thin lm) that showed characteristic bands for CCO (2060 cm À1 ) and CO functionalities (1588, 1575 and 1510 cm À1 ) (for free C 3 O 2 2280 cm À1 ) which are in good agreement with values reported for the complexes [M(h 2 (C,C 0 )-C 3 O 2 )(PPh 3 ) 2 ] (M ¼ Ni, Pt). 28,51 Unfortunately, mass spectrometry was not informative and microanalysis was hampered by the thermal instability of (3) even in the solid state. Nevertheless, based on the spectroscopic data discussed above, (3) was assigned as the adduct of C 3 O 2 with (1), i.e. the rst intermediate towards the  formation of (2). This was unequivocally established by a single crystal XRD study (Fig. 11).
As can be seen from Fig. 9, C 3 O 2 coordinates via one of the terminal CO's in an h 2 fashion to one of the Ti centers (Ti2) and h 1 via that same carbon to the other one (Ti1). The latter also coordinates to the central carbon (C2) of the C 3 O 2 ligand. The molecular structure of (3) represents the rst example of a crystallographically authenticated example of C 3 O 2 coordination and conrms the coordination modes of C 3 O 2 predicted by Pandolfo and Hillhouse based on spectroscopic evidence. 28,51 However the coordination mode calculated for M(PH 3 ) 2 C 3 O 2 (ref. 44) is closer to that found for 3 0 where an O is not coordinated. Presumably the bimetallic nature of Ti 2 Pn 2 allows more extensive donation to the unsaturated substrate. The Ti-Ti bond in (3) (2.4293(14)Å) is similar to the one found in parent (1) (2.399(2)Å) within esd's; a similar invariance in the Ti-Ti bond length has been observed in the adducts of (1) with CO ([Ti 2 (m:h 5 ,h 5 -Pn † ) 2 (m:h 2 ,h 1 -CO)] d Ti-Ti ¼ 2.4047(5)Å; [Ti 2 (m:h 5 ,h 5 -Pn † ) 2 (CO) 2 ] d Ti-Ti ¼ 2.4250(10)Å). 40 The ligation of C 3 O 2 has a profound effect on its bond angles, with the most prominent changes being the signicant deviation of the O1-C1-C2 and C1-C2-C3 angles from linearity (179.93(11) and 178.32 (12) respectively in free C 3 O 2 11 ) to 137.0(7) and 132.5(10) respectively. The C2-C3-O2 also deviates from linearity (172.0(10) vs. 179.57 (12) in free C 3 O 2 11 ) but to a much lesser extent. On the other hand, the bond distances in the ligated C 3 O 2 are similar within esd's to the ones found in free  11 ). In comparison to the CCO moiety found in (2) (Fig. 4), the corresponding C-C bond (i.e. C2-C3) is shorter (1.300(15)Å vs. 1.40(3)Å in (2)) while the C-O bond lengths are the same within esd's. In the case of the corresponding angles, the CCO angle in both (2) and (3) are identical (172.4(16)Å and 172.0(10) A respectively) (Fig. 11).
In conclusion, we report the rst example of the trimerisation of C 3 O 2 promoted by a well-dened molecular complex leading to the formation of (2). The core structure between the two [Ti 2 Pn † 2 ] moieties is reminiscent of biologically relevant compounds responsible for the regulation of ion concentrations in cells. This transformation was studied computationally revealing that the rst step is the formation of (3), which was conrmed experimentally by its isolation and structural characterization.

Conflicts of interest
There are no conicts of interest to declare.