Cleavage of carbon suboxide to give ketenylidene and carbyne ligands at a reactive tungsten site: a theoretical mechanistic study

Abstract

Reaction of (MePPh₂)₄WCl₂ with C₃O₂ has been shown experimentally to result in stepwise cleavage of the two C=C double bonds in C₃O₂ to give successively tungsten complexes containing phosphinoketenylidene and phosphinocarbyne ligands. The mechanism of such processes has been elucidated by density functional theory methods for the L₄WCl₂ (L = PMe₃, PMePh₂) systems. The triplet L₄WCl₂ reagents are found to proceed to singlet intermediates and products in a reaction sequence involving dissociation of a phosphine ligand, a triplet → singlet intersystem crossing, an initial C=C bond cleavage and a free phosphine attachment transition state. The first step is the rate-determining step with a Gibbs free energy barrier of 19.8 kcal mol⁻¹, and the formation of the stable phosphinoketenylidene intermediate is thermodynamically favorable. Further reaction of the phosphinoketenylidene intermediate to give the final phosphinocarbyne product is unusual because it is thermodynamically disfavored but kinetically feasible. The key steps involve loss of another phosphine ligand to give the transition state involving the cleavage of the second C=C bond of C₃O₂.

---

Introduction

Carbon suboxide, C₃O₂, was initially reported in 1906 as the product from double dehydration of malonic acid.¹ Despite the tendency of C₃O₂ to polymerize under relatively mild conditions,² its reactivity as the heterocumulene O=C=C=C=O, have led to its use as a reagent in organic chemistry as well as a ligand in transition metal chemistry. However, many interesting questions remain about its chemistry. Of particular interest is the use of carbon suboxide as a synthetic source of “C₂O” and “C” fragments. In this connection, C₃O₂ has been used as a precursor for the production of graphene and carbon nanotubes by thermal decomposition and polymerization at relatively low temperatures.³ Thus a temperature of only 180 °C is sufficient to produce carbon allotropes from C₃O₂ whereas nanotube synthesis otherwise requires much higher temperatures with an iron catalyst. The catalyst was found to play a key role in nanotube growth since without the catalyst the products instead were lamellar carbon nanoparticles.

In transition metal chemistry C₃O₂ functions as an activated olefin ligand, binding initially to the metal through a C=C double bond. The heavier congener C₃S₂ is also an olefin ligand, but it can bind to transition metals through either a C=C or a C=S double bond.^4,5 Subsequent reaction of the initially formed metal C₃O₂ complexes leads to ketenylidene complexes via oxidative addition. Thus in the reactions of C₃O₂ with platinum(0),⁶ rhodium(I),⁷ and nickel(0)⁸ complexes, cleavage of a C=C double bond occurs to give products with a CO ligand and a C=C=O ketenylidene group. For the rhodium(I) complexes,⁷ the ketenylidene group is liberated to undergo polymerization into (C₂O)_n polymers. In addition, the M–H bonds in tungsten and rhenium hydrides add to a C=C double bond in C₃O₂ to form complexes containing a ketenyl group linked to a carbonyl group through a C–C single bond.⁹ Theoretical calculations have explored the chemical bonding in C₃O₂ and its role as a π-acceptor in transition metal chemistry.¹⁰

Recent theoretical work by Esterhuysen and Frenking¹¹ have suggested an alternative view of C₃O₂ as C(CO)₂, i.e., a “carbon dicarbonyl” or “carbone”. Since each of the CO groups in C(CO)₂ furnishes a lone pair to the central carbon atom, all four original valence electrons of the central carbon atom are available as two lone pairs. Such lone pairs can function as Lewis bases [e.g., bonding to an AuCl electrophile in ClAu(C₃O₂)] and/or participate in π → π* back bonding to the π* antibonding orbitals of the two outer CO groups. Such bonding of C₃O₂ to a metal atom uses only the central carbon atom of the C₃O₂ ligand. The view of C₃O₂ as C(CO)₂ is related to the recently reported¹² stable boron dicarbonyl ArylB(CO)₂ [Aryl = 2,6-di(2,4,6-triisopropylphenyl)phenyl].

Another important reaction of C₃O₂ is the cleavage of its C=C double bonds leading ultimately to the formation of carbene, carbyne, or even carbide complexes. For example, C₃O₂ as well as some other heterocumulenes react with WCl₂(PMePh₂)₄ with subsequent cleavage of the C=C double bond initially coordinated to tungsten.^13–15 For C₃O₂, two CO molecules are sequentially lost consistent with its “carbon dicarbonyl” formulation by Esterhuysen and Frenking.¹¹ In this way C₃O₂ can provide a synthon for free carbon atoms^13,14 not unlike the use of simple binary transition metal carbonyls as reagents for metal deposition. The first loss of CO from C₃O₂ in this tungsten system gives the ketenylidene ligand C=C=O, which is trapped by coordination both with the central tungsten atom and coordination of a phosphine to the ketenylidene ligand. Further loss of the remaining CO group from the ketenylidene ligand occurs above 35 °C to give a carbide ligand which is stabilized by coordination with liberated phosphine to give a carbyne ligand of the type [W]≡C ← PR₃.

The metal–carbon double and triple bonds in carbene and carbyne complexes are so reactive that they are used both catalytically and stoichiometrically for many organic reactions.^16,17 For example, metal–vinylidenes (M=C=CR₂) or allenylidenes (M=C=C=CR₂), as well as carbynes are known to be common catalytic intermediates and can initialize polymerization.¹⁸ Thus metal vinylidenes are important species in many organic syntheses including alkyne polymerization and cycloaromatization.^19–21 Metal carbynes can function as polymerization catalysts for acetylenes and cycloalkenes.²² These considerations suggest that the formation of carbene or carbyne species from C₃O₂ might relate to the growth of graphene and carbon nanotube materials. For this reason, a detailed understanding of the mechanism of stepwise C=C double bond cleavage in C₃O₂ is critically important for generating the carbene and carbyne metal complexes involved in catalytic processes. Unfortunately, the mechanism of this process has not been well understood. For this reason we undertook a density functional theory study of the C=C binding of C₃O₂ to WCl₂(PMePh₂)₄ and subsequent transformations to phosphine-stabilized ketenylidene and carbyne complexes.

Theoretical methods

Density functional theory (DFT) methods have evolved as a practical and effective computational tool, especially for organometallic compounds.^23–29 In particular, the functional M06-L is found to be excellent for predicting geometries and vibrational frequencies. Furthermore, the M06-L method generally provides reliable results on the thermochemistry and thermochemical kinetics of main-group elements, transition metals, inorganic compounds, and organometallics.^30,31 Therefore we chose the M06-L functional for this study.

The basis set used for this work was a mixed double-ζ plus polarization (DZP) + Lanl2DZ basis set. Thus DZP is a double-zeta basis set, adding one set of d-functions on the heavy atoms and one set of p-functions on the hydrogen atoms. This DZP basis set for hydrogen, carbon, oxygen, sulfur, selenium, phosphorus and chloride was obtained by adding one set of pure spherical harmonic p(H)/d functions with orbital exponents α_p(H) = 0.75, α_d(C) = 0.75, α_d(O) = 0.85, α_d(S) = 0.70, α_d(Se) = 0.338, α_d(P) = 0.60 and α_d(Cl) = 0.60, to the standard contracted DZ sets.^32–36 For tungsten, the effective core potential (ECP) basis sets Lanl2DZ was used.^37,38

Geometry optimizations of all structures were performed at the DFT(M06-L)/DZP + Lanl2DZ level. Vibrational frequency calculations followed the optimizations to determine the thermodynamic properties. The transition state (TS) structures were located using TS and modredundant scan methods. Then the potential energy surface was obtained using the Gibbs free energy unless otherwise specified. Since the reported experimental work was carried out in toluene and chloroform, calculations in these solvents were also performed using the Polarizable Continuum Model (PCM).³⁹ All calculations were carried out with the Gaussian 09 program.⁴⁰

The experimental work^13,14 used methyldiphenylphosphine, PMePh₂, as the ligand. Since using PMePh₂ for all calculations was too computationally expensive, we chose a simpler phosphine, namely trimethylphosphine PMe₃, to locate the geometries of all of the structures and transition states in the reaction pathways. Subsequently the key structures were reconfirmed using PMePh₂ corresponding to the reported experimental work.

Upon checking the potential energy surfaces, we found one intersystem crossing between triplet and singlet spin states. Harvey's program⁴¹ was used to locate the structures of the minimum energy intersystem crossing points (MECP).

Results and discussion

Generation of the reactive tungsten complex WCl₂(PMe₃)₃

These tungsten complexes appear to favor 16-electron structures owing to the steric effect of the phosphine groups and thus can exist in either triplet or singlet spin states. For the initial reactant trans-W^IICl₂(PMe₃)₄ (1),^42,43 the triplet spin state is the lower energy spin state, lying 15.2 kcal mol⁻¹ below the singlet spin state structure (Fig. 1). The octahedral coordination of the tungsten atom in WCl₂(PMe₃)₄ is too crowded to accept other incoming ligands leading to a seven-coordinate tungsten atom. Therefore, one or more ligands (chloride or PMe₃) must be removed from WCl₂(PMe₃)₄ before C₃O₂ can enter the tungsten coordination sphere. Chloride elimination from the triplet WCl₂(PMe₃)₄ requires a very high Gibbs free energy of 132.4 kcal mol⁻¹ (gas phase) or 72.7 kcal mol⁻¹ (toluene), whereas PMe₃ elimination requires only 19.8 kcal mol⁻¹ (gas phase) or 18.3 kcal mol⁻¹ (toluene), thus indicating the relative lability of the PMe₃ ligands. The 14-electron complex WCl₂(PMe₃)₃, obtained by losing a PMe₃ ligand from 1, is also a triplet structure, lying 16.9 kcal mol⁻¹ in Gibbs free energy below the corresponding singlet structure (Fig. 2). The resulting vacant site in WCl₂(PMe₃)₃ can now bind a C₃O₂ ligand. For our reported neutral species, most of the relative Gibbs free energy differences between the gas phase and solvent are found to be less than 2.0 kcal mol⁻¹, as expected. Thus, solvent effects are minor for neutral species. Therefore, only gas-phase results are discussed in this paper.

Fig. 1 Loss of PMe₃ or Cl⁻ from the triplet and singlet spin states of WCl₂(PMe₃)₄. The red and black paths refer to the singlet and triplet, respectively. The colored and black values represent the relative Gibbs free energies (kcal mol⁻¹), compared to singlet 1, in the gas phase and in toluene solution.

Mechanism of phosphinoketenylidene complex formation

Fully understanding the mechanism of stepwise cleavage of the C=C bonds in C₃O₂ is critically important for preparing metal complexes containing the W≡C–PPh₃ ligand as well as related triply bonded W≡C–L analogues. Hillhouse and coworkers suggested a possible mechanism but not in detail.^13,14 We examined some possible pathways, both in the singlet and triplet spin states. We then selected the most favorable one shown in Fig. 2 as well as others given in Fig. S1 and S2 in the ESI.†

Fig. 2 A possible path from the reagents to the phosphinoketenylidene intermediate int3 including two electron spin states, the singlet (red) and triplet (black), crossing at a MECP. The relative Gibbs free energies, all compared to R plus C₃O₂, and the Gibbs free energy changes from the previous step (in parentheses) are given in kcal mol⁻¹. The colored and black values represent the Gibbs free energies (kcal mol⁻¹) in the gas phase and in toluene solution, respectively.

As stated above, the first step is an extrusion of one PMe₃ ligand from triplet WCl₂(PMe₃)₄ (1) to give the intermediate WCl₂(PMe₃)₃ (R). However, loss of a second PMe₃ ligand from WCl₂(PMe₃)₃ seems unlikely since the dissociation Gibbs free energy is found to be relatively high at 33.0 kcal mol⁻¹ (see Fig. S1 in the ESI†). The intermediate R (Fig. 2) is therefore more likely to bind C₃O₂, suggested by the binding free energy of −18.1 kcal mol⁻¹ for triplet int1 or −30.9 kcal mol⁻¹ for singlet int1. The triplet–singlet splitting is found to be 16.9 kcal mol⁻¹ for R, but decreases to only 4.1 kcal mol⁻¹ when C₃O₂ is added to R to give int1. This indicates that C₃O₂ in R is a strong field ligand like CO, thereby making the singlet reaction pathway more accessible.

The experimental work^13,14 led to the isolation and characterization of the two important complexes int3 and P in Fig. 2 and 4, respectively, which have been optimized in both singlet and triplet spin states. In both cases the singlet int3 and P structures were found to have lower energies than the corresponding triplet structures, as indicated by an average singlet–triplet splitting of 15.9 ± 1.6 kcal mol⁻¹ (see Fig. S1 in the ESI†). Therefore, from the initial triplet R to the final singlet P via the singlet int3, a MECP must be present in both channels. In this connection, all of the stationary points as well as transition states were examined in two different spin states (Fig. 2). The WCl₂(PMe₃)₂(CO)(C₂O) intermediate (int2), derived from int1 by simultaneous cleavage of the C1=C0 bond (Fig. 3) and loss of a PMe₃ ligand, is predicted to be a singlet because of its lower energy of 18.1 kcal mol⁻¹ relative to the corresponding triplet. It is thus reasonable to suspect that the structure of MECP is more similar to that of int1 (small splitting) rather than to that of int2 (large splitting).

Fig. 3 Structures along the possible reaction pathway in Fig. 2 with some bonding data in black and natural charges in red (arrows from the atoms). The methyl hydrogen atoms are omitted for clarity.

Harvey's program⁴¹ was then used to locate the MECP structure (Fig. 3). The obvious change in structure of the MECP compared with int1 is the decrease in the C1–C0–W–Cl dihedral angle from 52.0° for int1 to 3.5° for the MECP. This means that in int1 the C₃O₂ ligand is not in the same plane as the two chlorine atoms whereas in the MECP the C₃O₂ ligand and the two chlorine atoms become nearly planar. The W–C0 and W–C1 distances are found to decrease from 2.128 Å and 2.089 Å (triplet) to 2.061 Å and 2.037 Å (singlet), respectively, suggesting strengthening of the W–C bonds in going from int1 to MECP. Meanwhile, the increased C0–C1 distance of 1.484 Å in MECP from 1.427 Å in int1 suggests stronger activation (oxidation) of C₃O₂ in MECP relative to int1. This can be verified from the lower total natural charge of −0.281 on the C₃O₂ fragment in the MECP and larger charge of −0.309 in int1 (Table S3 in the ESI†). Moreover, the natural charges on C0 and C1 change from −0.516 and 0.561 in int1 to −0.491 and 0.557 in MECP, respectively. The trans W–PMe₃ bond in MECP is weakened owing to the trans-effect of C₃O₂ since the trans-W–P distance is elongated from 2.614 Å in int1 to 2.653 Å in MECP. Accordingly, the MECP structure is disfavored with respect to loss of the PMe₃ group trans to C₃O₂. At this point the energy of the triplet pathway increases and that of the singlet pathway decreases so that the two pathways cross. For the triplet pathway, WCl₂(PMe₃)₃(η_CC²-C₃O₂) first becomes another intermediate WCl₂(PMe₃)₂(η_CC²-C₃O₂) by losing a PMe₃ ligand and then goes through a transition state to become WCl₂(CO)(PMe₃)₂(C₂O) (see Fig. S1 in the ESI†). For the singlet pathway loss of the second PMe₃ ligand leads to simultaneous cleavage of the C=C bond of C₃O₂ thereby giving the ketene complex int2.

The intermediate int2 can be regarded as a singlet C=C=O carbene stabilized by coordination to tungsten. The very stable intermediate (int3) is formed by using free PMe₃ rather than the coordinated PMe₃ to attack the carbene carbon atom C0, simultaneously converting the monohapto η¹-C₂O ligand into a dihapto η²-C₂O ligand stabilized by coordination of PMe₃. This process needs to cross a transition state (TS1) with an activation free energy of 17.8 kcal mol⁻¹, then TS1 releases 25.7 kcal mol⁻¹. Thus the W=C=C=O unit is unstable in int2, which was not isolated in the experimental work. The natural charge on the carbene atom (C0) atom in the C₂O fragment becomes more negative from −0.373 in int2, to −0.497 in TS1 and −0.860 in int3. The charge on the C₂O fragment also becomes more negative in this reaction sequence, i.e., −0.099 in int2, −0.264 in TS1 and in −0.994 in int3. These observations indicate the enhancement of the M → C₂O back donation to the C₂O fragment in the sequence int2 < TS1 < int3 thereby strengthening the bonding between the metal center and the C₂O fragment. However, the bonding mode of the C₂O fragment changes in the sequence int2 → TS1 → int3. Thus in int2 the C0 atom is sp hybridized with a linear W=C=C=O bond, while in int3 the C0 atom has bent sp² hybridization after coordination with PMe₃. This bonding change is reflected by the W–C0 bond distance that changes from 1.881 Å in int2 to 1.930 Å in TS1 and 1.994 Å in int3 (Fig. 3). It is very difficult to break a chemical bond with sp hybridized atoms, whereas chemical bonds with sp² hybridized atoms are more readily broken. Accordingly, the C0–C2 bond distance lengthens from 1.310 Å to 1.321 Å and 1.392 Å in the sequence int2 → TS1 → int3, indicating that the second C0=C2 double bond of C₃O₂ is also activated in int3. However the C0–C2 bond (1.392 Å) in int3 is still viable so that int3 is a stable intermediate and can be isolated experimentally.

Conversion of the triplet reagent WCl₂(PMe₃)₄ to the singlet intermediate int3 occurs through an intersystem crossing. This process is thermodynamically favorable since the Gibbs free energy of int3 is 19.3 kcal mol⁻¹ below that of WCl₂(PMe₃)₄. For the entire pathway in Fig. 2, the rate-limiting step is then the first step, namely the loss of the first PMe₃ ligand from WCl₂(PMe₃)₄, requiring a Gibbs free energy of 19.8 kcal mol⁻¹. The transition state TS1 is the second important step, whereas the intersystem crossing is a minor energetic factor. Conversely, the reverse reaction has a higher energy barrier of 25.7 kcal mol⁻¹ from int3 to TS1. Thus the reverse reaction is unlikely to proceed under ambient conditions owing to thermodynamically and kinetically unfavorable factors.

Mechanism of phosphinocarbyne formation

The phosphinocarbyne complex, P, is obtained from int3 by CO elimination.^13,14 However, it appears to be thermodynamically disfavored since P has a higher free energy of 17.6 kcal mol⁻¹ than int3 (Fig. 4). In fact, the most common method of preparing a ketenylidene complex such as int3 is the carbonylation of a carbyne ligand such as that in P.^13,14 However, experimental work¹⁴ shows that P can be obtained from int3 by mild heating to 35 °C for 48 h in chloroform solution. Conversely, treatment of P with carbon monoxide at 20 °C does not give int3. These observations suggest that the reaction direction is kinetically rather than thermodynamically controlled. We therefore searched for a reaction pathway with a suitably high activation free energy for the conversion of P to int3.

Fig. 4 Possible reaction pathway from the phosphinoketenylidene intermediate to the phosphinocarbyne product, all in singlet states. The red and black values represent the relative Gibbs free energies (kcal mol⁻¹), compared to R plus C₃O₂ (Fig. 2), in the gas phase and in chloroform, respectively.

The tungsten atom in int3 is six-coordinate by being bonded to the phosphinoketenylidene ligand, one carbonyl group, two chlorides, and two trimethylphosphine ligands and has a 16-electron configuration similar to WCl₂(PMe₃)₄. Cleavage of the C=C bond of the ketenylidene ligand converts int3 into a complex with an 18-electron configuration but with an unfavorable tungsten coordination number of seven. We also scan the energy surface, with every step optimized, by slowly elongating the ketenylidene C=C bond. However we cannot locate a stable intermediate with seven ligands. Thus the first possible step in the pathway for conversion of int3 into P should be loss of one ligand. Therefore, the dissociation Gibbs free energies of CO, Cl⁻, and PMe₃ ligands from int3 were compared and the corresponding energy data and possible reaction paths, in the singlet and triplet spin states and in the gas phase and in chloroform solution, are supplied in Fig. 4 and S2 and Table S2 in the ESI.† The results show that dissociation of either CO or Cl⁻ from int3 in the gas phase is very difficult requiring a free energy greater than 40 kcal mol⁻¹ (41.2 and 123.8 kcal mol⁻¹ for CO and Cl⁻, respectively, see Fig. S2 in the ESI†). Even in chloroform, extrusion of Cl⁻ still requires 61.2 kcal mol⁻¹ (see Fig. S2 in the ESI†). However, the dissociation free energy of PMe₃ is only 18.0 kcal mol⁻¹ (gas) or 16.8 kcal mol⁻¹ (chloroform). This implies that the ligand dissociated from int3 in the first step of its conversion to P should be PMe₃. This information is used to derive the reaction pathway shown in Fig. 4. The corresponding intermediates are shown in Fig. 5.

Fig. 5 Structures along the possible reaction pathway in Fig. 2 with some bonding data in black and natural charges in red (arrows from the atoms). The methyl hydrogen atoms are omitted for clarity.

The first step in the conversion of int3 to P thus involves a loss of one PMe₃ ligand to give int4, requiring a free energy of 18.0 kcal mol⁻¹. The loss of PMe₃ is accompanied by migration of the chlorine atom trans to CO in int3 to trans to the remaining PMe₃ ligand owing to the large trans effect of CO and the steric effect of the phosphinoketenylidene ligand. Therefore the tungsten coordination in int4 is approximately square pyramidal with a vacancy trans to the CO group. However, the electron density on the tungsten atom is reduced in going from int3 to int4 owing to loss of a strong σ-donor PMe₃ ligand. This reduces the M→L back donation thereby slightly lengthening the W–C0 and W–C2 bonds from 1.994 Å and 2.148 Å (Fig. 3) to 2.014 Å and 2.218 Å (Fig. 5), respectively. In addition the C0–C2 bond is shortened from 1.392 Å (Fig. 3) to 1.369 Å (Fig. 5).

The phosphinoketenylidene C=C double bond in int4 then undergoes cleavage to give the phosphinocarbyne intermediate, int5, through a transition state, TS2. The activation free energy of this transition state is 19.5 kcal mol⁻¹, which is slightly larger than the PMe₃ dissociation free energy of int3. The structure of TS2 is so close to that of int5 that the TS2 → int5 conversion reduces the free energy by only 3.5 kcal mol⁻¹. This is mainly because the major structural change from TS2 to int5 is the substantial lengthening of the C0–C2 distance from 1.937 Å to 2.391 Å (Fig. 5). However, the C0–C2 distance in TS2 is already so much longer than a normal carbon–carbon bond that its further lengthening has little effect on the energy. In contrast, the very large stretching of the C0–C2 bond from 1.369 Å to 1.937 Å in going from int4 to TS2 changes a C=C double bond to a C⋯C non-bonding interaction. This is expected to lead to a substantial change in energy. Both TS2 and int5 contain the phosphinocarbyne ligand with short W≡C triple bond distances of 1.880 and 1.850 Å, respectively, as compared with a W–C0 distance of 2.014 Å in int4. However, the phosphinocarbyne ligand has a large trans-influence, lengthening the trans W–CO distance from 1.910 Å in int4 to 2.200 Å in int5. This trans-influence reduces the Gibbs free energy for dissociation of the CO ligand trans to the carbyne to only 4.4 kcal mol⁻¹. The resulting facile dissociation of this CO group gives another intermediate int6. Note that int6 and its successors as depicted in Fig. 5 are rotated approximately 90° counterclockwise on the paper plane compared with its predecessors.

This intermediate int6 is a relatively high-energy coordinatively unsaturated intermediate containing a five-coordinate tungsten atom (Fig. 5). The loss of the strongly back-bonding CO group in going from int5 to int6 strengthens the back donation of the tungsten atom to the remaining ligands leading to slight shortening of the bonds from the tungsten atom to both the phosphinocarbyne ligand and the remaining CO group. Addition of one of the liberated PMe₃ ligands from an earlier step to int6 is all that is needed to give the final product P. However, such addition is expected to be unfavorable since the vacant coordination site is trans to the phosphinocarbyne, which has a large trans effect. On the other hand, one vibrational mode in int6 around 70 cm⁻¹ includes a swing of the W–Cl bond, located trans to the only remaining PMe₃ ligand in the tungsten coordination sphere, between the cis and trans position of the phosphinocarbyne ligand. Thus the Gibbs free energy required for this chlorine to migrate trans to W≡C–PMe₃ is only 2.4 kcal mol⁻¹. This migration produces an intermediate int7, having the vacant coordination site trans to the PMe₃ ligand with a nearly linear Cl–W–C0 angle of 166.2°. Attaching a free PMe₃ ligand to the vacant coordination site in int7 to give the product, P, is now energetically favorable, releasing 23.2 kcal mol⁻¹.

The final product WCl₂(CO)(PMe₃)₂(≡CPMe₃) (P) is a stable octahedral hexacoordinate structure. The PMe₃ unit in the phosphinocarbyne ligand is essentially a phosphonium ion with a formal positive charge. Therefore, a neutral phosphinocarbyne ligand is formally a four-electron donor to the tungsten atoms with three electrons coming from the W≡C triple bond and the fourth electron indirectly from the phosphorus atom. The tungsten atom in P thus has the favored 18-electron configuration. In P the W≡C triple bond shortens to 1.824 Å and the W–C0–P angle to the phosphinocarbyne ligand is slightly bent at 162.7°.

In going from int3 to P, there seems to exist an obviously much larger free energy barrier, namely 37.5 kcal mol⁻¹, from int3 to TS2. However, between int3 and TS2, there lies an intermediate int4 that splits the large free energy barrier into two smaller ones and leads to a more feasible reaction (Fig. 4). Thus the rate-determining step in going from int3 to P is that from int4 to the transition state TS2 (Fig. 4), requiring a Gibbs free energy of approximately 20 kcal mol⁻¹, rather than the step from int3 to TS2. This Gibbs free energy barrier suggests that the reaction could proceed under ambient conditions. However, the experiments show that int3 is so stable at ambient temperature that it can be isolated in the crystalline state. The reaction with PMePh₂ as the phosphine actually requires slight heating at 35 °C for 48 h.^13,14 This is because the forward process is totally endothermic (17.6 kcal mol⁻¹). Therefore an energy compensation is needed from heating. In addition, vaporization of CO from chloroform is a key driving force for conversion of int3 to P. The overall Gibbs free energy of 17.6 kcal mol⁻¹ leads to a very small equilibrium constant (K^Θ) in chloroform solvent. The concentration of CO in the solvent is readily decreased by warming, thus driving the reaction equilibrium in the forward direction. Therefore, at ambient temperature int3 is stable enough to be isolated in the crystalline state for determination of its structure by X-ray crystallography. Additionally, we suspect that the forward reaction rate at ambient temperature might be very slow so that conversion of int3 to P needs 48 h although it is energetically preferred.

While looking back to the reverse reaction channel, i.e. from P to int3, this process can occur spontaneously since it is thermodynamically favored as releasing a Gibbs free energy of 17.6 kcal mol⁻¹, and is kinetically feasible owing to the dominant free energy barrier of 23.2 kcal mol⁻¹ (Fig. 4). However, the experimental work^13,14 showed that carbonylation of the phosphinocarbyne complex P back to the ketenylidene complex int3 does not occur at 20 °C and 1 atm CO. This discrepancy might arise from the following three factors. First, the larger activation Gibbs free energy barrier (23.2 kcal mol⁻¹) from P to int3 leads to an even lower reaction rate than for the forward process. As a result, a longer reaction time might be required to observe the carbonylation of the metal phosphinocarbyne intermediate int3. Second, carbonylation of phosphinocarbyne might be not a simple reaction, since the preparation of ketene ligands⁴⁴ from carbonylation of metal carbene requires high CO pressure for increasing the reaction rate as well as for making the equilibrium move to the backward. Finally, the W≡C triple bond in P is a high energy structural feature and therefore is suspected to be highly reactive towards carbonylation. However, Hillhouse and co-workers did not supply more details about the experimental conditions, even though Pandolfo's group⁴⁵ provided suggestions that this reaction was reversible. Recently, a bridged μ₂-CPh carbyne was reported to couple a CO to form spontaneously a PhC=C=O bridge.⁴⁶ However, losing a CO from the PhC=C=O bridge required heating (energy compensation). Therefore, a carbyne ligand is suspected to be more stable than a PhC=C=O or Ph₃PC=C=O ligand. We suggest that further experimental investigation is therefore necessary for preparing the novel Ph₃P=C=C=O ligand by carbonylation of a phosphinocarbyne ligand at a lower temperature, longer reaction time, and especially higher CO pressure.

Structures with PMePh₂ ligands

The discussion above is based on the simplest phosphine that is a reasonable model for the experimental chemistry, namely trimethylphosphine, PMe₃. However, the experimental work used methyldiphenylphosphine, PMePh₂. In order to determine the significance of the choice of phosphine for these reactions, the structures of the species actually isolated and characterized structurally, namely int3 and P with PMePh₂ ligands, labelled as int3-ph and P-ph, respectively, were optimized (Fig. 6 and Table S4 in the ESI†) and the optimized geometries were compared with experimental data. However, note that the experimental data for P-ph are taken from triphenylphosphinocarbyne complex WCl₂(CO)(PMePh₂)₂(≡CPPh₃) since experimental data are not available for the methyldiphenylphosphinocarbyne complex WCl₂(CO)(PMePh₂)₂(≡CPMePh₂). The key bonding data change very little when the phosphine is changed from PMe₃ in int3 to PMePh₂ in int3-ph. Thus the critical W–C0, W–C1, C0–C2, and C2–O2 distances in int3 (Fig. 3) and int3-ph (Fig. 6) are essentially unchanged within ±0.01 Å. The largest difference is the shortening of the W–C2 bond from 2.148 Å in int3 to 2.137 Å in int3-ph. For P-ph, the critical bond distances are also very similar to those of P, with almost identical W–CO and W–P bond distances. The W–C0 distance and the W–C0–P angle of the carbyne ligand are slightly shorter and smaller, respectively, in P-ph relative to P. These results show that the simplified PMe₃ ligand provides a reasonable model for our theoretical study.

Fig. 6 Geometries of int3-ph and P-ph with some bonding and natural charge data. The top (black) and the bottom (blue) numbers represent the present work and experiment,¹⁴ respectively, and the red numbers show the natural charge. Note that the experimental data for P-ph are from the complex WCl₂(CO)(PMePh₂)₂(≡CPPh₃).

The Gibbs free energy of the P-ph complex relative to int3-ph is 18.5 kcal mol⁻¹ (Fig. 6), which is slightly larger than the 17.6 kcal mol⁻¹ energy of P relative to int3 (Fig. 4). For int3-ph, the differences between the calculated key bonding interatomic distances and the experimental distances from X-ray crystallography¹⁴ agree within ±0.03 Å, suggesting that our DFT calculation is very reliable. In particular, the interatomic distances in the key fragment (1.999, 2.137, and 1.381 Å for W–C0, W–C2, and C0–C2, respectively), namely the phosphinoketenylidene ligand, agree within ±0.013 Å of the experimental values (1.996, 2.146, and 1.368 Å for W–C0, W–C2, and C0–C2, respectively) except for the C2–O2 distance. For P-ph, the calculated W–C0 distance of 1.818 Å is slightly shorter than the experimental value of 1.823 Å. However, our predicted W–C0–P angle for the phosphinocarbyne ligand in P-ph of 158.3° differs significantly from the experimental value of 171.2°. This discrepancy might arise from the fact that the phosphinocarbyne ligand in P-ph is W≡CPMePh₂ whereas that in the experimental complex is W≡CPPh₃. The greater steric demand of PPh₃ relative to PMePh₂ might account for this difference. The predicted interatomic distances for P-ph with a W≡CPMePh₂ ligand and the experimental interatomic distances for the corresponding complex with a W≡CPPh₃ ligand,¹⁴ including the critical W–Cl, W–P and W–C2 bonds, agree within ±0.02 Å.

Conclusions

Carbon suboxide (C₃O₂) is known experimentally to react with WCl₂(PMe₃)₄ to form a stable phosphinoketenylidene intermediate and a phosphinocarbyne product by sequentially breaking its two C=C bonds. The mechanism of this process has been investigated using density functional theory. Carbon suboxide strongly prefers bonding to tungsten as well as other transition metals through its C=C bond rather than its C=O bond. The mechanism of this reaction between C₃O₂ and WCl₂(PMe₃)₄ involves intersystem crossing from a triplet reaction channel to a singlet reaction channel, with a low free energy barrier. The MECP structure is found to be close to the first reaction intermediate where the smallest singlet–triplet splitting energy is located. The first C=C cleavage then occurs, followed by migration of a free phosphine ligand to the resulting C₂O ligand to give a phosphine-stabilized stable ketenylidene intermediate. Cleavage of the second C=C bond from the original C₃O₂ follows loss of a phosphine ligand from this phosphinoketenylidene intermediate with dissociation of one CO ligand. Addition of a phosphine ligand gives the final phosphinocarbyne product. The reaction pathway from the phosphinoketenylidene intermediate to the final phosphinocarbyne complex is unusual since it is thermodynamically disfavored. However, it can proceed because it is kinetically favorable during heating as shown by the calculated energy barriers.

Acknowledgements

We are indebted to the National Natural Science Foundation of China (Grant 21303138), the China Postdoctoral Science Foundation (Grants 2013M540660 and 2014T70817), the Chinese Universities Scientific Fund (Grant 2014YB028), the Fundamental Research Funds for the Central Universities (Grant 2452015425), the PhD Research Startup Foundation of Northwest A&F University (Grant 20100203185), and the U. S. National Science Foundation (Grant CHE-1057466) for support of this research.

Notes and references

1. O. Diels and B. Wolf, Ber. Dtsch. Chem. Ges., 1906, 39, 689.
2. J. S. Auf der Günne, J. Beck, W. Hoffbauer and P. Krieger-Beck, Chem.–Eur. J., 2005, 11, 4429–4440.
3. M. W. Shao, D. B. Wang, G. H. Yu, B. Hu, W. C. Yu and Y. T. Qian, Carbon, 2004, 42, 183–185.
4. A. P. Ginsberg, W. E. Lindsell and W. E. Silverthorn, Trans. N. Y. Acad. Sci., 1971, 33(3), 303–312.
5. A. P. Ginsberg and W. E. Silverthorn, J. Chem. Soc. D, 1969, 823–824.
6. G. Paiaro and L. Pandolfo, Angew. Chem., Int. Ed., 1981, 20, 288–289.
7. G. Paiaro and L. Pandolfo, Angew. Chem., Int. Ed., 1981, 20, 289–290.
8. A. K. List, M. R. Smith III and G. L. Hillhouse, Organometallics, 1991, 10, 361–362.
9. G. L. Hillhouse, J. Am. Chem. Soc., 1985, 107, 7772–7773.
10. M. Casarin, L. Pandolfo and A. Sassi, Organometallics, 2002, 21, 2235–2239.
11. C. Esterhuysen and G. Frenking, Chem.–Eur. J., 2011, 17, 9944–9956.
12. H. Braunschweig, R. D. Dewhurst, F. Hupp, M. Nutz, K. Radacki, C. W. Tate, A. Vargas and Q. Ye, Nature, 2015, 522, 327–330.
13. A. K. List and G. L. Hillhouse, J. Am. Chem. Soc., 1988, 110, 4855–4856.
14. A. K. List and G. L. Hillhouse, Organometallics, 1989, 8, 2010–2016.
15. J. C. Bryan, S. J. Geib, A. L. Rheingold and J. M. Mayer, J. Am. Chem. Soc., 1987, 109, 2826–2828.
16. M. I. Bruce, Chem. Rev., 1998, 98, 2797–2858.
17. C. Bruneau and P. H. Dixneuf, Angew. Chem., Int. Ed., 2006, 45, 2176–2203.
18. C. Bruneau and P. H. Dixneuf, Metal Vinylidenes and Allenylidenes in Catalysis: From Reactivity to Applications in Synthesis, Wiley-VCH, Weinheim, Germany, 2008.
19. C. Bruneau and P. H. Dixneuf, Acc. Chem. Res., 1999, 32, 311–323.
20. B. M. Trost, Acc. Chem. Res., 2002, 35, 695–705.
21. B. M. Trost, M. U. Frederiksen and M. T. Rudd, Angew. Chem., Int. Ed., 2005, 44, 6630–6666.
22. T. J. Katz, T. H. Ho, N. Y. Shih, Y. C. Ying and I. W. van Stuart, J. Am. Chem. Soc., 1984, 106, 2659–2668.
23. M. Brynda, L. Gagliardi, P. O. Widmark, P. P. Power and B. O. Roos, Angew. Chem., Int. Ed., 2006, 45, 3804–3807.
24. N. Sieffert and M. Bühl, J. Am. Chem. Soc., 2010, 132, 8056–8070.
25. P. Schyman, W. Lai, H. Chen, Y. Wang and S. Shaik, J. Am. Chem. Soc., 2011, 133, 7977–7984.
26. R. D. Adams, W. C. Pearl, Y. O. Wong, Q. Zhang, M. B. Hall and J. R. Walensky, J. Am. Chem. Soc., 2011, 133, 12994–12997.
27. R. Lonsdale, J. Olah, A. J. Mulholland and J. N. Harvey, J. Am. Chem. Soc., 2011, 133, 15464–15474.
28. L. Crawford, D. J. Cole-Hamilton, E. Drent and M. Bühl, Chem.–Eur. J., 2014, 20, 13923–13926.
29. H. Zhekova, M. Krykunov, J. Autschbach and T. Ziegler, J. Chem. Theory Comput., 2014, 10, 3299–3307.
30. Y. Zhao and D. G. Truhlar, J. Chem. Phys., 2006, 125, 194101.
31. B. S. Marendrapurapu, N. A. Richardson, A. V. Copan, M. L. Estep, Z. Yang and H. F. Schaefer, J. Chem. Theory Comput., 2013, 9, 2930–2938.
32. T. H. Dunning, J. Chem. Phys., 1970, 53, 2823–2834.
33. T. H. Dunning and P. J. Hay, in Methods of Electronic Structure Theory, ed. H. F. Schaefer, Plenum, New York, 1977, pp. 1–27.
34. S. Huzinaga, J. Chem. Phys., 1965, 42, 1293–1302.
35. A. J. H. Wachters, J. Chem. Phys., 1970, 52, 1033–1036.
36. D. M. Hood, R. M. Pitzer and H. F. Schaefer III, J. Chem. Phys., 1979, 71, 705–712.
37. W. R. Wadt and P. J. Hay, J. Chem. Phys., 1985, 82, 284–298.
38. P. J. Hay and W. R. Wadt, J. Chem. Phys., 1985, 82, 299–310.
39. J. Tomasi, B. Mennucci and R. Cammi, Chem. Rev., 2005, 105, 2999–3093.
40. M. J. Frisch, et al., Gaussian 09, Revision B.01, Gaussian, Inc., Wallingford CT, 2010.
41. J. N. Harvey, M. Aschi, H. Schwarz and W. Koch, Theor. Chem. Acc., 1998, 99, 95–99.
42. P. R. Sharp, Organometallics, 1984, 3, 1217–1223.
43. S. F. Nelsen, R. W. Alder, R. B. Sessions, K. D. Asmus, K. O. Hiller and M. Göbl, J. Am. Chem. Soc., 1980, 102, 1430–1431.
44. G. L. Geoffroy and S. L. Bassner, Adv. Organomet. Chem., 1988, 21, 1.
45. R. Berani, M. Casarin and L. Pandolfo, Coord. Chem. Rev., 2003, 236, 15–33.
46. M. A. Alvarez, M. E. García, D. García-Aivó, M. E. Martínez and M. A. Ruiz, Organometallics, 2011, 30, 2189–2199.

---

Footnote

† Electronic supplementary information (ESI) available: Fig. S1: reaction paths of C₃O₂ binding to WCl₂(PMe₃)₄; Fig. S2: reaction paths from int3 to P with an extrusion of one CO or Cl⁻ ligand; Table S1 Gibbs free energies of intermediates and TS from 1 to int3; Table S2 Gibbs free energies of intermediates and TS from int3 to P; Table S3: natural charges of the atoms in C₃O₂ molecule and its fragments; Fig. S3: geometries of P-ph with full PMePh₂ ligands with some labeled atoms; Table S4: bond lengths and angles for WCl₂(PMePh₂)₂(η_CC²-C₂OPMePh₂); Tables S5–S28: theoretical Cartesian coordinates for all complexes in the gas phase. See DOI: 10.1039/c5ra19003g

---