¹⁸³W NMR Spectroscopy Guides the Search for Tungsten Alkylidyne Catalysts for Alkyne Metathesis

Tungsten Alkylidynes with a Tripodal Silanolate Ligand Architecture

When seen against this backdrop, it is perhaps surprising that tungsten alkylidynes with silanolates as ancillary ligands are conspicuously underrepresented in the literature (Scheme 1). The neopentylidyne complex 8 a is accessible from 7 a by salt metathesis but was found to degrade rapidly in solution as well as in the solid state; the derived phenanthroline adduct [8 a⋅(phen)] is stable but kinetically inert and hence no adequate precatalyst either.^{25, 26} Unligated 8 a succumbs to a C−H activation process that converts the alkylidyne into an alkylidene unit: while this transformation is clearly manifested in the NMR data,²⁷ the resulting complex could not be obtained in crystalline form and hence remains structurally undefined.²⁵ In this context, reference is made to a similar C−H activation observed for 5 a on treatment with 11.²² This decomposition pathway may be the reason why yet another complex carries siloxide ligands devoid of aromatic substituents in the periphery; once again, the fairly small number of applications that 10 a has found so far, mostly to conjugated 1,3-diyne substrates, make an accurate assessment difficult.^{28, 29}

On top of this stability issue comes a practical problem: other than the trichloro alkylidyne 7 a, the much more readily available tribromo complexes of type 7 b³⁰ with an aryl substituent on the alkylidyne invariably led to formation of the corresponding ate-complexes such as 9 on treatment with Ph₃SiOK in toluene.^{1a, 31} For the intrinsically high Lewis-acidity of tungsten, they do not lose the fourth silanolate to vacate the necessary coordination site for the incoming substrate under conditions that work well in the molybdenum series.¹ Although 9 exhibits modest catalytic activity at high temperature in the homo-metathesis of 1-phenyl-1-propyne, it cannot compete at all with its molybdenum counterpart 2.^{1, 25} Interestingly, however, 9 and relatives cleave the triple bond of aryldiazonium salts [Ar−N≡N]⁺ with surprising ease at a rate that outcompetes loss of dinitrogen.³²

These issues notwithstanding, we conjectured that the tripodal silanolate ligand framework of the new “canopy catalysts” such as 3 might provide an opportunity for the development of truly competent tungsten catalysts. The phenyl ring forming the basal plane blocks the coordination site trans to the alkylidyne and should hence prevent detrimental C−H activation as observed for 5 a and 8 a from occurring. Moreover, the problem with competing ate-complex formation can likely be solved by a new synthesis route that avoids salt metathesis (Scheme 2):¹¹ Specifically, it was found that 3 is best prepared on treatment of [(tBuO)₃Mo≡CAr] with trisilanole 15; the ligand exchange is entropically driven and benefits from the different acidity of alcohols and silanols. Since deprotonation of the ligand is unnecessary, irreversible ate-complex formation is effectively prevented.

In consideration thereof, the analogous complex 4 b³³ was chosen as the starting point for the preparation of a series of new tungsten alkylidynes; it differs from the classical Schrock catalyst 4 a¹⁴ only in the substituent of the alkylidyne but is easier to make on scale. In line with our expectation, reaction of 4 b with 15 a (R=Ph) in toluene afforded 16 a in 79 % yield (Scheme 2),³⁴ whereas 4 c lacking the 2,6-dimethyl substitution on the benzylidyne furnished a complex mixture. The structure of 16 a in the solid state (Figure 2) shows great resemblance to the analogous molybdenum canopy complexes,¹¹ even though the short W−O bonds bear testimony to the higher Lewis-acidity of tungsten. The catalytic activity, however, was disappointing and the likely reason readily identified. Addition of 3-hexyne (5 equiv) to a solution of 16 a in [D₈]toluene caused an instant color change from yellow to deep red: Upon cooling to −20 °C, a sharp signal set was recorded that allowed the new species to be identified as the mixed metallacyclobutadiene complex 17 formed in the initial [2+2] cycloaddition step (Scheme 2).³⁵ Although 3-hexyne was present in excess, it took one week at ambient temperature for 17 to convert into metallacyclobutadiene 18 carrying three ethyl substituents by reaction with a second equivalent. The striking inertia of 17 as the “snapshot” of the “initiation” step of the catalytic cycle implies that cycloreversion is not only rate-determining, as expected,³⁶ but is much too slow for a competent catalyst. Such hyper-stabilization of the metallacycle is well known to be the result of an overly Lewis-acidic metal fragment.^{20, 37-39} In this context, it needs to be emphasized that treatment of an analogous molybdenum canopy complex with 3-hexyne unexpectedly furnished a metallatetrahedrane complex;^{11, 12} such disparate behavior of tungsten and molybdenum alkylidynes with the same ancillary ligand sphere is striking and unprecedented.

The coordination geometry about the tungsten center of 17 (Figure 3) is much closer to square pyramidal (τ₅≈0.14) than to the trigonal bipyramidal arrangement, which basically all other known tungstenacyclobutadiene complexes adopt;^{14b, 20, 23, 24b, 37-40} this difference is likely a consequence of the chelate ligand framework and may contribute to the low reactivity of 17.⁴¹ The equatorial siloxide serves as a notably better π-electron donor to tungsten, since the W1-O2 bond (1.903(1) Å), which is essentially coplanar with the metallacycle, is significantly shorter than the orthogonal W1-O1 (1.988(1) Å) and W1-O3 (1.952(1) Å) bonds. The metallacyclobutadiene ring itself is highly distorted, in that all bond lengths and angles are uneven. This finding is in accord with the spectroscopic data which show that the C_α/C_α′ atoms of the metallacycle in 17 are inequivalent (Figure 4). Most notable are the different ¹J_C,W coupling constants: although a detailed interpretation is difficult,⁴² we note that the shorter W1-C1 “double” bond features the much larger ¹J_C,W coupling constant. In the derived complex 18, each ethyl substituent gives rise to distinct ¹H and ¹³C NMR signals. Importantly, ROESY-NMR experiments show a dynamic process that mutually interconverts the C_α/C_α′ positions and their ethyl substituents on the NMR timescale,⁴³ which is interpreted as interconversion of the two tautomers 18 a and 18 b.⁴⁴ The observation of two different valence isomers implies that the metallacyclobutadiene formed by [2+2] cycloaddition and the metallacyclobutadiene releasing the product in the [2+2] cycloreversion step are discrete entities that must not be mistaken for a mesomeric resonance form.⁴⁵

Ligand Design Revisited: Guidance by ¹⁸³W NMR Spectroscopy

The exceptional kinetic stability of 17 and 18 shows that a simple extrapolation of the successful ligand design from molybdenum to tungsten is to no avail. We had previously recognized that a direct look at the central metal via ⁹⁵Mo NMR spectroscopy was tremendously useful for catalyst design.¹¹ Therefore it was hoped that similar advantage could be taken from ¹⁸³W NMR in our quest for truly competent tungsten alkylidynes; the very large spectral range of this spin- nucleus and the pleasingly narrow signal line widths suggest so. However, the dramatically poor receptivity is a serious handicap in that the low-γ ¹⁸³W isotope is ≈49 times less sensitive than the already insensitive ⁹⁵Mo nucleus.⁴⁶ This issue had previously been addressed in various ways, of which 2D shift correlation spectroscopy is arguably the most successful approach.⁴⁷ Unfortunately, many tungsten complexes show very small scalar couplings to protons (hydrides), even when directly bound.⁴⁸ It is hence truly remarkable that the ¹⁸³W NMR signal of 16 a (δ_W=517 ppm) can be detected with ease by ¹H,¹⁸³W HMBC measurements using long-range couplings (⁵J) between the tungsten center and the protons on the ortho-methyl groups of the benzylidyne unit (Figure 5).⁴⁹ The fact that such ⁵J-coupling does exist is testament to an efficient orbital overlap between the alkylidyne and its aryl cap.

We are aware of a single prior study into tungsten alkylidynes based on 1D ¹⁸³W NMR spectroscopy.⁵⁰ Surprisingly though, the reported ¹⁸³W NMR shift of [(tBuO)₃W≡CPh] (δ_W=2526 ppm) is drastically different from what we record by the ¹H,¹⁸³W HMBC pulse sequence for the closely related complex 4 b (δ_W=166 ppm) carrying two remote methyl substituents on the phenyl ring. Struck by the discrepancy, we sought to rigorously confirm our data point by a number of control experiments. To this end, spectra were recorded with different offset and sweep widths to exclude possible folding of the signals.⁵¹ In addition, a 1D ¹⁸³W NMR spectrum of 4 b was recorded using an INEPT magnetization transfer, which proved that the inversely detected and the directly recorded δ_W are identical within ±0.2 ppm.

After this validation, ¹H,¹⁸³W HMBC experiments were used to record spectra with an excellent signal-to-noise ratio in less than 20 min on a Bruker AVIIIHD 400 MHz NMR spectrometer. Figure 6 shows that the catalytically poorly active podand complex 16 a is much more deshielded than the traditional Schrock-type complex 4 b from which it is derived. Since the over-stabilization of the metallacyclobutadienes 17 and 18 had shown that the metal center in 16 a is too Lewis-acidic, it is tempting to take the deshielding of the ¹⁸³W NMR signal as a proxy.⁵² To test this hypothesis more closely, additional tungsten alkylidyne complexes were prepared by the standard methods (Schemes 2 and 3). It is of note that solvolysis of 4 b by Ph₃SiOH furnished complex 8 b bearing three monodentate Ph₃SiO- ligands, which had been beyond reach of salt metathesis because of instant ate-complex formation. Since the ¹⁸³W NMR signal of 8 b (δ_W=548 ppm) is even more deshielded than that of 16 a, this complex is no promising candidate for catalytic alkyne metathesis.⁵³ Crystals suitable for X-ray diffraction could not be grown but repeated attempts furnished a different crop that turned out to be the tungsten d⁰ oxo alkyl complex 19 (Figure 7), inadvertently formed by addition of moisture to the alkylidyne unit of 8 b. Indeed, this complex can be deliberately made from 8 b and one equivalent of water in toluene; the strong deshielding of the tungsten center (δ_W=758.1 ppm, see the SI) is well befitting a W^VI=O unit. To the best of our knowledge, 19 is only the second example for the formation of a alkyl ligand by controlled double-protonation of an alkylidyne and the first reported case in which a monomeric oxo-complex is obtained in this way.^{54, 55, 56, 57}

Next, the phenyl groups on the silicon bridges of the tripodal framework were replaced by more electron-donating alkyl substituents (note that complex 3 as the currently best canopy catalyst of the molybdenum series carries methyl rather than phenyl groups).¹¹ As expected, the spectral response is a shift of the ¹⁸³W signal to higher frequencies, even though complexes 16 b (R=Et) and 16 c (R=CH₂CHMe₂) are still much more deshielded than the parent complex 4 b (Figure 6). The same is true for 10 b³³ carrying permethylated siloxide rather than silanolate ligands, which differs from complex 10 a²⁸ as the only previously disclosed tungsten alkylidyne with a silicon-based ancillary ligand sphere solely in the substitution pattern of the aryl cap on the alkylidyne. The expectation that the almost identical ¹⁸³W NMR shifts of 10 b, 16 b and 16 c might translate into similar performance proved correct: all of them effect the homo-metathesis of 1-methoxy-4-(prop-1-yn-1-yl)benzene at ambient temperature in toluene,⁵⁸ but the mass balance was invariably poor because of substantial competing polymerization, independent of whether MS 5 Å was added or not.⁵⁹

To complete the picture, Figure 6 also includes the ¹⁸³W NMR spectrum of [ArC≡W{OC(CH₃)(CF₃)₂}₃]⋅THF (5 b, Ar=2,6-dimethylphenyl) for comparison. Despite the presence of THF as an additional neutral donor ligand, the observed signal at δ_W=396 ppm speaks for a highly Lewis-acidic alkylidyne. This conclusion fits well to the previously reported finding that tungsten alkylidynes with hexafluoroalkoxide ligands lead to (over)stabilized metallacycles and are hardly adequate for alkyne metathesis because they lack activity^{38, 60} or cause competing polymerization.^{20, 61} We hence conclude that silanolates (and highly fluorinated alkoxides) as poor π- and σ-donors synergize exceedingly well with molybdenum alkylidynes, but the rather disappointing results obtained with 8 b, 10 b and 16 a–c in concord with their strongly deshielded ¹⁸³W NMR signals indicate a fundamental mismatch for the tungsten series that cannot be rectified by peripheral modifications.

Based on these data, it became increasingly clear that the Si-linker strategy had to be abandoned altogether and the search re-directed toward a ligand with stronger (rather than weaker) donor properties than tBuO- that the inherently more electropositive W^VI seems to mandate; if one overshoots, however, catalytic activity will also get lost as a certain Lewis-acidity is necessary for the [2+2] cycloaddition to proceed.^{3, 4, 24b, 24c, 36} Once again, we reasoned that a chelate structure might provide an adequate balance: the W-O-C angles are expected to be more obtuse and the O-atoms therefore more sp-hybridized than those of monodentate tBuO-, which should gently upregulate their π-donor character; an improved stability of a chelate architecture comes on top. However, triol 20 as the carbinol variant of trisilanole 15 is inadequate for this purpose because it is incapable of engaging all three -OH groups with a single Mo^VI or W^VI atom (Scheme 4).^{11, 62} As a consequence, the ligand design was revisited and the expanded C₃-symmetric triol 22 prepared from 14 in two high-yielding steps. Despite the additional rotatable bond and the resulting higher conformational freedom, this ligand adopts a favorable “all-up” conformation in the solid state (see the SI); treatment with 4 b furnished the targeted chelate complex 23 in almost quantitative yield (Scheme 4).⁶³

Using the ¹H,¹⁸³W HMBC technique, a ¹⁸³W resonance at δ_W=114 ppm was recorded for 23, which is upfield of the signal of the parent complex 4 b (δ_W=166 ppm). In line with our expectation that a chelate carbinol ligand is a somewhat better donor for the geometrically enforced substantial widening of the W-O-C angles,⁶⁴ this data point suggests that the tungsten center of 23 is indeed less Lewis-acidic than that of the classical Schrock-type complex 4 b, even though the first coordination sphere is formally made up of (substituted) “tert-butoxy” groups in either case. This view is corroborated by computational data. Specifically, the calculated energies of the canonical MO's are notably different: the tripodal ligand as the better donor raises the energy of the π-symmetric orbitals. This effect is slightly manifested in the π(W-C) orbitals (HOMO, HOMO-1) but particularly notable in the energy of the π*(W-C) orbitals (for details, see the SI). An important consequence of this uplifting of the π*(W-C) levels is the fact that the canonical LUMO of 23 is almost entirely ligand-based, with the largest lobes on the phenyl group that forms the basal plane (Figure 8).

A detailed analysis of the ¹³C NMR data by 2-component ZORA DFT calculations⁶⁵ confirms the conclusions (for full computational details, see the SI): the alkylidyne C-atom of 23 is clearly more shielded than that of 4 b (Table 1). Because the computed isotropic shifts (δ_iso) and ¹J_W,C and ²J_W,C coupling constants nicely reproduce the experimental values,⁴² one can confidently break the data down to the individual components δ_ii of the chemical shift tensor (δ_iso=(δ₁₁ + δ₂₂ + δ₃₃)/3), which themselves are linked to the computed shielding tensor σ (δ_ii≈σ_iso,ref−σ_ii). Details of chemical shift tensor (CST) analysis apart,⁶⁶ it has been shown that magnetic shielding reflects an increasing energy gap between high-lying occupied and the low-lying vacant orbitals. For metal alkylidynes, the key contributions come from the pairs π(M-C)→σ*(M-C), σ(M-C)→π*(M-C), and σ(C-Ar)→π*(M-C) (Figure 8 c).⁶⁷

Table 1. Tabulation of recorded and computed NMR data ([D₈]toluene) of the alkylidyne unit of different tungsten 2,6-dimethylbenzylidyne complexes.

Complex	δW [ppm]	δC [ppm]		1JW,C [Hz]		2JW,C [Hz]
		exp.	computed	exp.	computed	exp.	computed
16 a	517	287	285	270	282	37.1	37.3
8 b	548	283	274	273	289	30.1	38.9
10 b	458	281	280	274	n.d.	38.7	n.d.
4 b	166	273	281	297	310	44.8	43.4
23	114	264	262	293	304	44.4	43.4

• n.d.=not determined

The computed data shown in Figure 8 and the results of a Natural Chemical Shielding (NCS) analysis of all complexes compiled in Table 1 (see also the Supporting Information Table S-13 and Figure S-35) can be summarized as follows: The contributions arising from π(M-C) orbitals (and to a lesser extent those of the σ(M-C) orbitals) decrease upon going from 4 b to 23. This reflects the higher-lying π*(M-C) orbitals that render 23 less electrophilic, as well as the equally higher-lying σ*(M-C) orbital induced by the tripodal ligand. It is also interesting to note that the lower-lying of the π*(M-C) orbitals in 4 b and 23 is associated with the less-deshielding σ₂₂ component. This presumably results from the delocalization of the π(M-C) system over the dimethylphenyl moiety, which decreases the efficiency of the orbital coupling. The component oriented along the M-C axis (σ₃₃) is particularly shielded in 23 and points to a highly cylindrical electron density distribution around this M−C bond, as is observed for acetylene and molecules with a C_∞ axis,⁶⁸ making it difficult to draw a direct relation to the HOMO–LUMO gap for this component.

One can therefore safely conclude that complex 23 must not be mistaken for just a tethered version of the classical Schrock catalyst; rather, the tripodal ligand topology itself constitutes an intrinsic electronic determinant. With the canonical LUMO delocalized on the ligand framework and a relatively high lying metal-centered π*(W-C) orbital (LUMO+2), over-stabilization of a derived metallacycle is improbable; a good level of activity and selectivity may therefore be expected.

Benchmarking

Indeed, the direct comparison shows that 23 outperforms the Schrock catalyst 4 b in terms of rate; moreover, polymerization has essentially ceased, as evident from the mass balance data (Figure 9).⁶⁹ When performed in the presence of silanized MS 5 Å to sequester the released 2-butyne,⁷⁰ the reaction furnished the desired homo-metathesis product in 85 % yield (NMR). Good results were also achieved in a number of demanding ring closing alkyne metathesis (RCAM) reactions (Scheme 5). Since acid-sensitive epoxides and aldol subunits as well as many heteroatom donor sites are known to be incompatible with the parent Schrock catalyst 4 a,^{16, 17, 71, 72} it is notable that complex 23 opens new opportunities. The successful formation of cycloalkyne 30 bears witness of this notion, since this compound comprises a very acid- and base-sensitive vinylepoxide subunit, a labile allylic ether as well as a skipped and hence non-thermodynamic array of double and triple bonds.⁷³ The formation of cycloalkyne 31 is equally indicative: for the presence of a highly elimination-prone aldol subunit and the additional donor site of the thiazole ring in the side chain, this particular compound had previously been reported to lay beyond the reach of the classical Schrock catalyst 4 a.⁷⁴