Phosphirenium Ions as Masked Phosphenium Catalysts: Mechanistic Evaluation and Application in Synthesis

Abstract

The utilization of phosphirenium ions is presented; optimized and broadened three-membered ring construction is described together with the use of these ions as efficient pre-catalysts for metal-free carbonyl reduction with silanes. Full characterization of the phosphirenium ions is presented, and initial experimental and computational mechanistic studies indicate that these act as a “masked phosphenium” source that is accessed via ring opening. Catalysis proceeds via associative transfer of {Ph₂P⁺} to a carbonyl nucleophile, H–SiR₃ bond addition over the C=O group, and associative displacement of the product by a further equivalent of the carbonyl substrate, which completes the catalytic cycle. A competing off-cycle process leading to vinyl phosphine formation is detailed for the hydrosilylation of benzophenone for which an inverse order in [silane] is observed. Experimentally, the formation of side products, including off-cycle vinyl phosphine, is favored by electron-donating substituents on the phosphirenium cation, while catalytic hydrosilylation is promoted by electron-withdrawing substituents. These observations are rationalized in parallel computational studies.

1. Introduction

The chemistry of the three-membered phosphacycle, the phosphirenium ion, first appeared in the literature at the beginning of the 1980s,¹ along with a number of theoretical studies that considered their stabilization by σ*-aromaticity.² The chemistry to date is dominated by the use of these species as reagents in synthesis; exploitation of their catalytic Lewis acidic properties has yet to be reported. The first example of a stabilized three-membered phosphirene–tungsten ring was reported by Marinetti and Mathey in 1982³ (Scheme 1a). A year later, Hogeveen and Breslow reported and characterized phosphirenium salts by formal addition of halophosphines to alkynes,⁴ a method that was further improved using amine-phosphenium ions [iPr₂NPCl]⁺[AlCl₄]⁻ (Scheme 1b).⁵ Other routes for their synthesis or methods for structural elaboration have been developed,⁶⁻⁸ but it was only in 2017, when Hirano and Miura reported the synthesis of highly functionalized organophosphines postulated to proceed via a phosphirenium intermediate, that the application in practical organic synthesis was reported (Scheme 1c).⁹ They have since expanded their methodology to the synthesis of dibenzophospholes,^10,11 while other groups have reported functionalization of phosphine oxides in the presence of Tf₂O.¹²

Scheme 1. (a,b) Different Approaches to the Synthesis of Phosphirenium Ions; (c) Functionalization via a Phosphirenium Ion; (d) This Work, where Mechanistic Investigations Indicate that a Phosphirenium Pre-catalyst Acts as a Masked Phosphenium Ion, Which is Itself the Active Catalyst.

Phosphirenium ions pose an interesting mechanistic challenge: they possess both electrophilic phosphorus and a low-lying π*-acceptor centered on the highly activated carbon–carbon double bond in the ring. Both could behave as Lewis acidic sites for the activation of organic substrates and so possess exciting opportunities for selective or divergent reactivity. However, we herein report on the unexpected ability of phosphirenium ions to undergo ring opening and subsequent alkyne displacement via nucleophilic trapping of a phosphenium ion {Ph₂P⁺} by carbonyl reagents (Scheme 1d). Onward reaction with a silane and a further equivalent of carbonyl releases the reduced carbonyl product and regenerates the highly active phosphenium adduct, which is the active catalyst. Divalent phosphenium ions are highly elusive species, relying on lone pair stabilization of the empty phosphorus p-orbital by adjacent heteroatoms (e.g., N- or P-donors) without which they are challenging to detect and characterize.¹³ It is worth noting that divalent phosphenium species, in the form of N-heterocyclic phosphenium ions, have recently started to gather interest,¹⁴ not least due to their catalytic capabilities, which are as yet unproven with a simple {Ar₂P⁺} catalyst. Here, our synergistic experimental and computational approach strongly supports our assertion of easy-to-prepare and -handle phosphirenium ions acting as a masked phosphenium catalyst of the form {Ph₂P⁺}, stabilized by a Lewis basic carbonyl oxygen atom. An advantage of the work reported here is the ease of synthesis, with a range of phosphirenium pre-catalysts being readily prepared or formed in situ for use in catalysis.

2. Results and Discussion

2.1. Synthesis and Characterization of Phosphirenium Ions

Due to the practicality of the reaction, Hirano and Miura’s procedure (where the phosphirenium ion was detected but not isolated)^9,11,15 was employed for the synthesis of phosphirenium ion 3·OTf (Table 1, entry 1). Formation of 3·OTf is observed by following the reaction via ³¹P{¹H} NMR spectroscopy, where a characteristic peak at −108.3 ppm is obtained.⁹ To optimize the protocol, different activating agents were reacted with diphenyl acetylene 1 and secondary phosphine oxide (SPO) 2 in CDCl₃. 3·OTf does not form when using triflate salts, such as silver or iron triflates (entries 2 and 3), or by Brønsted acid activation with triflic acid¹⁶ (entry 4) or activated triflic acid (entry 5). Acetic anhydride does activate 2, with the formation of Ph₂P(OAc)¹⁷ and Ph₂P(O)PPh₂.¹⁸ However, these species do not react further with 1 (entry 6). Di-tert-butyl dicarbonate (Boc₂O) is not effective for this transformation, and even at higher temperatures, 1 and 2 are recovered (entry 7). Following a brief solvent and temperature screening,¹⁹ the optimal conditions were determined to be CH₃CN at 60 °C for 30 min, which give full conversion to 3·OTf (entry 8).

Table 1. Optimization of the Synthesis of Phosphirenium Ions^a.

entry	activating agent	T (°C)	conversion to 3·OTf
1	Tf2O	80	60%
2	Fe(OTf)2	80	0
3	AgOTf	80	0
4	HOTf	80	0
5	HOTf/PhSiH3	80	Ph2PH (major)
6	Ac2O	80	several productsb
7	O(CO2tBu)2	80	0
8	Tf2O	60	>99%c (98%d)

^aReaction conditions 1 (0.2 mmol), 2 (0.2 mmol), activating agent (0.2 mmol), and CDCl₃ (0.4 mM).

^bSee the Supporting Information.

^cCD₃CN.

^dSpectroscopic yield analyzed by inverse-gated ³¹P{¹H} NMR spectroscopy with PPh₃ as an internal standard.

Adding 1 equiv of NaB_Ar^F to the optimized reaction mixture of 1, 2, and Tf₂O in CD₃CN forms 3·B_Ar^F, which is isolated in 66% yield (Figure 1). A small downfield shift is observed in the ³¹P{¹H} NMR spectrum from −108.3 ppm for 3·OTf to −109.1 ppm for 3·B_Ar^F; the disappearance of the triflate peak at −79.3 ppm is observed by ¹⁹F NMR spectroscopy and substituted by a ^–B_Ar^F counterion at −62.5 ppm. Counterion exchange is not observed when using sodium or silver salts, such as NaBF₄ or AgNTf_2, and 3·OTf is recovered; with AgSbF₆ and AgBF₄, decomposition of the starting material is observed via ³¹P{¹H}, ¹⁹F, and ¹H NMR spectroscopy.²⁰ The structure of 3·B_Ar^F was confirmed by single-crystal X-ray diffraction analysis. The bond metrics associated with this structure are in line with other examples of phosphirenium cations, for example, the ring P–C distances are 1.748(2) and 1.746(3) Å, and the C=C bond distance is 1.340(4) Å; both are similar to reports from Mathey, Wild, and Cowley.^3,6,7,21 The alternating single and double bonds show that the structure depicted is the dominant resonance form of the phosphirenium ion in the solid state. This species represents the first isolated phosphirenium ion with a borate counterion.²²

Figure 1.

Counterion exchange at the phosphirenium ion. Isolated yield presented. ORTEP view of 3·B_Ar^F showing a thermal ellipsoid at 30% probability levels. H, F, and minor components of disordered anion C atoms were omitted for clarity.²³

With the optimized conditions in hand, the generality of the methodology was explored (Figure 2). Some of the phosphirenium ions were very sensitive and could only be characterized in situ (5·OTf, 6·OTf, 7·OTf, 8·OTf, and 10·OTf proved particularly challenging; see the Supporting Information for spectral data). The use of different substituted SPOs, bearing bearing aryl or alkyl groups, is well tolerated. For example, using Cy₂P(O)H, 4·OTf and 4·B_Ar^F are isolated in high yields (97 and 98%, respectively). These species display downfield ³¹P{¹H} NMR signals compared to 3, with a singlet at −85.4 and −86.5 ppm, respectively, due to the increased electron donation of the alkyl substituents. There is only a very modest difference in the ³¹P chemical shift observed between the ions bearing different anions, 3·OTf versus 3·B_Ar^F Δδ = 0.8 and 4·OTf versus 4·B_Ar^F Δδ = 1.1 ppm, which is as expected given the non-coordinating nature of the counterions.

Figure 2.

Synthesis of phosphirenium ions; alkyne (0.2–0.5 mmol), phosphine oxide (0.2–0.5 mmol), Tf₂O (0.2–0.5 mmol), NaB_Ar^F (0.2–0.5 mmol), CD₃CN, 60 °C. Isolated yield presented; ^aspectroscopic yield analyzed by inverse-gated ³¹P{¹H} NMR spectroscopy with PPh₃ as an internal standard; ^bCD₃Cl. ORTEP view of 10·OTf showing a thermal ellipsoid at 30% probability levels. H atoms were omitted for clarity.²⁷

Mixed SPOs bearing alkyl and phenyl groups, such as tBuPhP(O)H and CyPhP(O)H, are also well tolerated with full conversion after 1 h to form 5·OTf(24) and 6·OTf, with ³¹P{¹H} NMR signals at −97.1 and −93.3 ppm, respectively. Varying the alkyne was also explored; phenyl acetylene and Cy₂P(O)H lead to full conversion to 7·OTf with a phosphorus signal at −80.2 ppm and a downfield signal for the terminal C–H at 8.4 ppm. For all phosphirenium species presented, there are deshielded signals in the proton and carbon NMR spectra, which can be explained by the resonance structure of the phosphirenium ion.^25,26 1-Phenyl-1-propyne reacts well with 2 to form 8·OTf with a ³¹P signal at −103.6 ppm. Further functionalization of the alkyne is tolerated with 9·OTf (38%) and 10·OTf (64%) isolated using the optimized conditions. The phosphorus signals for 9·OTf and 10·OTf appear at δ −104.3 and −110.9 ppm, respectively, versus −108.3 ppm for 3·OTf; thus, by this measure, the alkyne substituents only have a minor impact on P⁺ electronics, but as we show below, the same substituents have a dramatic effect on catalytic reactivity. Crystallization is not limited to B_Ar^F counterions, and X-ray single-crystal analysis of 10·OTf confirms the three-membered structure of this phosphirenium species (Figure 2, bottom), with similar metrics as observed for 3. Short contacts between the counterion and the phosphirenium substituents are observed, although no direct P···O interaction was found, probably due to the steric effect of the two phenyl rings compared to, for example, the mixed Ph and Me system reported by Wild.⁷

2.2. Hydrosilylation Optimization

With these phosphirenium species in hand, we explored the use of phosphirenium ions as pre-catalysts for the hydrosilylation of aldehydes and ketones. We opted for this transformation because it would allow comprehensive mechanistic investigations using a range of experimental techniques.

Initial studies were conducted with 10 mol % phosphirenium species 4·B_Ar^F, prepared and activated in situ prior to the addition of the substrates, benzophenone 11 and dimethylphenyl silane 12, in an equimolar ratio. The reaction was monitored by ¹H NMR spectroscopy and examined after 18 h of reaction at r.t. (Table 2). Initial results were promising with 44% total conversion of 11 (entry 1) as 28% 13a and 16% 14a. By increasing the temperature to 80 °C (entry 2), full conversion of the starting material is observed with 84% formation of 13a, while 16% 14a persists. Using 10 mol % isolated 4·B_Ar^F, lower conversions (73%) are observed after 18 h (entry 3), with 58% 13a. Isolated pre-catalyst 3·B_Ar^F shows higher reactivity, with increased conversion to 14a versus 13a at 10 mol % loading (entry 5). When phosphirenium triflate species are used, such as 3·OTf or 4·OTf, only deoxygenation occurs with full conversion into 14a; this is due to hydrogen release from silane reacting with co-catalytic TfOH (formed from Tf₂O).²⁸ Thus, in line with Miura’s recent studies on phosphirenium-derived heterocycles,^9,11 we also note that the use of a base is crucial to quench TfOH released during pre-catalyst formation and therefore increase the selectivity toward 13a. See Table S2 in the Supporting Information where comprehensive benchmarking and optimization were undertaken demonstrating that not only does TfOH lead exclusively to the deoxygenation product 14a but also that superior hydrosilylation catalysis takes place using a phosphirenium pre-catalyst compared to the equivalent reaction using TfOPPh₂. A total of 5 mol % 2,6-lutidine was added to 5 mol % in situ-prepared pre-catalyst 4·B_Ar^F prior to addition of substrates 11 and 12, and the selective formation of 13a is observed (5 h, 80 °C, CD₃CN, entry 5). Changing to 3·B_Ar^F, with 5 mol % 2,6-lutidine, gives good reactivity at lower temperature (entry 6), and further improvement is seen with the electron-poor phosphirenium species 9·B_Ar^F, for which reaction with 5 mol % 2,6-lutidine leads to full conversion into 13a at 60 °C after only 1 h (entry 7). In contrast, electron-rich 10·B_Ar^F reacted at a slower rate with full conversion to the hydrosilylation product requiring 58 h at 80 °C (entry 8). Clearly the electronics of the alkyne used to prepare the phosphirenium pre-catalyst play a substantial role in catalytic competency. Screening of the base with 10 mol % 3·B_Ar^F as the catalyst of choice shows lower conversion to 13a when using NEt₃ and pyridine (entry 9 and 10) with 59 and 52%, respectively, and recovery of the starting materials with K₂CO₃ (entry 11) after 18 h at 80 °C. These results show that a slightly bulkier, aromatic organic base like 2,6-lutidine is most effective.²⁹

Table 2. Optimization Reduction of Carbonyls with Silanes^a.

entry	catalyst (mol %)	temp.	13ab	14ab
1	4·BArF (10)	r.t.	28	16
2	4·BArF (10)	80 °C	84	16
3b	4·BArF (10)	80 °C	58	15
4	3·BArF (10)	80 °C	21	75
5b,c	3·BArF (10)	80 °C	>99
5c	4·BArF (5) + 2,6-lutidine (5)	80 °C	>99
6c	3·BArF (5) + 2,6-lutidine (5)	60 °C	>99
7d	9·BArF (5) + 2,6-lutidine (5)	60 °C	>99
8e	10·BArF (5) + 2,6-lutidine (5)	80 °C
9	3·BArF (10) + NEt3 (10)	80 °C	59
10	3·BArF (10) + pyridine (10)	80 °C	52
11	3·BArF (10) + K2CO3 (10)	80 °C

^aReaction condition: (i) activation step R₂P(O)H (5–10 mol %), Tf₂O (5–10 mol %), NaB_Ar^F (5–10 mol %), 1 (5–10 mol %), CD₃CN (0.41 M), 60 °C, and 30 min; (ii) benzophenone 11 (0.25 mmol), dimethylphenyl silane 12 (0.25 mmol), 18 h, and conversion.

^bIsolated pre-catalyst.

^c5 h.

^d1 h.

^e>99% after 58 h.

2.3. Mechanistic Studies

We initiated our mechanistic investigations by analyzing the catalytic reaction mixture, using in situ-prepared 3·B_Ar^F, by NMR spectroscopy (Scheme 2a). After 30 min, our standard reaction in CD₃CN only shows the starting material (H₂SiMePh 12, −16.9 ppm) and product (13a, 8.3 ppm) in the ²⁹Si{¹H} NMR spectrum. However, a plethora of species are observed in the ³¹P{¹H} NMR spectrum, with the major species being 3·B_Ar^F at −108.5 ppm, along with P₂Ph₄ at −17.8 ppm, (E)-(1,2-diphenylvinyl)diphenylphosphine at 7.1 ppm,⁹ smaller amounts of (Z)-(1,2-diphenylvinyl)diphenylphosphine at −7.4 ppm,³⁰ and benzhydryl diphenylphosphinate at 36.5 ppm.³¹ While phosphines are known to facilitate hydrofunctionalization of carbonyls,³² noteworthy is the lack of catalytic activity observed using 10 mol % HPPh₂ or P₂Ph₄.

Scheme 2. ²⁹Si{¹H} and ³¹P{¹H} NMR Spectroscopy Speciation Studies of (a) Standard Catalytic Reaction of 3·B_Ar^F, 2,6-Lutidine, 11, and 12; (b) Stoichiometric Reaction of 3·B_Ar^F, 2,6-Lutidine and 12; and (c) Stoichiometric Reaction of 3·B_Ar^F, 2,6-Lutidine, 11, and 12.

A stoichiometric reaction to form 3·B_Ar^F with subsequent addition of 2,6-lutidine and 12 results in two new species by ²⁹Si{¹H} NMR spectroscopy. The major species is identified as Me₂SiHOTf at 23 ppm, formed by the facile cleavage of the silane phenyl group.³³ Analysis of the mixture by ³¹P{¹H} NMR spectroscopy shows 3·B_Ar^F, P₂Ph₄, and vinyl phosphine as per the catalytic reaction, along with HPPh₂ with its characteristic peak at −40 ppm (Scheme 2b). Finally, when performing the same stoichiometric study in the presence of benzophenone 11 and analyzing the mixture after 10 min by ²⁹Si{¹H} NMR spectroscopy, we observe 13a at 8.3 ppm but no traces of silyl triflate (Me₂SiHOTf or Me₂PhSiOTf).³³ The latter mixture was also analyzed by ³¹P{¹H} NMR spectroscopy, which shows the presence of 3·B_Ar^F, vinyl phosphine, and P₂Ph₄, but no traces of HPPh₂ are visible at any point of the analysis (Scheme 2c).

When performing hydrosilylation to form 13a using 10 mol % 9·B_Ar^F pre-catalyst and 2,6-lutidine, we also observe the formation of substituted vinyl phosphine [i.e., (E)-(1,2-bis(4-(trifluoromethyl)phenyl)vinyl)diphenylphosphine] in situ. Hydrosilylation to form 13a using 10 mol % 10·B_Ar^F pre-catalyst and 2,6-lutidine gives a complex array of peaks in the ³¹P NMR spectrum, making identification difficult although likely including vinyl phosphines.

The use of 20 mol % (E)-(1,2-diphenylvinyl)diphenylphosphine and 20 mol % 2,6-lutidine in a catalytic reaction of 11 and 12 does not give an appreciable quantity of 13a after 18 h at 80 C, indicating that the vinyl phosphine is not an active catalyst in the hydrosilylation reaction, and this is further supported by computational studies.³⁴

Next, we undertook kinetic studies using NMR spectroscopic monitoring. By pre-forming the catalyst in situ, we have analyzed the order in substrates 11 and 12 by varying the amount of the starting material using 10 mol % 3·B_Ar^F. The data obtained suggest first-order dependence in 11, while an inverse order is observed for 12(35) (Figure 3). Product inhibition is not observed, with similar reaction rates obtained when adding 0.5 equiv of product 13a to the catalytic reaction of 11 and 12 in a 1:1 mixture (5.32 × 10^–4 vs 5.63 × 10^–4 mmol/dm³ min). Therefore, the negative order in silane can be attributed to silane involvement in side-product formation or catalyst degradation. Comparable data are obtained when using isolated pre-catalyst 3·B_Ar^F in the absence of 2,6-lutidine, further supporting the assertion that the role of 2,6-lutidine is simply to mop up TfOH released on formation of 3·B_Ar^F from reaction of Tf₂O with alkyne, SPO, and NaB_Ar^F.²⁶ It is important to note that the kinetic data found for this transformation are in contrast with Piers’ hydrosilylation-type activation via Lewis acids [e.g., B(C₆F₅)₃],³⁶ where the increase in ketone concentration slows down the reaction. Indeed, interrogating the ¹¹B and ¹⁹F NMR spectra from catalytic reactions also shows that only borate is present in solution; no borane signals are observed, and so we feel that borane-mediated hydrosilylation is unlikely in this case.

Figure 3.

a) Graphical analysis showing inverse order relationship for the formation of 13a at different phenyldimethylsilane 12 loadings (yellow ◆ 0.75, blue ● 1, red ◆ 1.25, and green ■ 1.5 equiv); (b) plot showing first-order dependence for the formation of 13a at different benzophenone 11 loadings (0.75, 1, 1.25, and 1.5 equiv); and reaction conditions: (i) activation step Ph₂P(O)H 2 (10 mol %), Tf₂O (10 mol %), NaB_Ar^F (10 mol %), diphenylacetylene 1 (10 mol %), CD₃CN (0.4 M), 60 °C, and 30 min and (ii) 2,6-lutidine (10 mol %), ketone or aldehyde (0.20 mmol), silane (0.20 mmol), 80 °C, and PhMe used as an internal standard.

At this stage, we have three distinct pieces of experimental data that necessitate further investigation:

• (i)Observation of vinyl phosphine both during catalysis and in stoichiometric studies indicates that phosphirenium ring opening could be involved, either as a catalyst deactivation process or on-cycle ring-opening/ring-closure sequence;
• (ii)The inverse order in silane indicates that there is a catalytically inactive pathway dominating at higher silane loadings.
• (iii)Phosphirenium ions with different electronic substituents have very different capabilities in hydrosilylation (9·B_Ar^F > 3·B_Ar^F ≫ 10·B_Ar^F), but these changes in electronics only have a modest impact on the phosphorus center (e.g., compare 3·OTf, 9·OTf, and 10·OTf³¹P NMR chemical shift). We therefore believe that the role of the phosphirenium pre-catalyst is not merely to act as a simple Lewis acid.³⁷

With these experimental results in hand, we undertook a computational study of the reactions of the phosphirenium cation, 3⁺ (denoted as I in the computational study) with acetone and Me₂PhSiH. Initial studies considered the direct nucleophilic attack of either the ketone or silane at the phosphirenium ring;²⁹ however, the characterization of these processes indicated that a ring-opening step occurs first to form an ethenylium intermediate II at +22.0 kcal/mol. From II, catalysis is initiated via the nucleophilic attack of the ketone at the P center (see Figure 4a). This proceeds through TS(II–III)_P at +28.6 kcal/mol and results in both P–O bond formation and displacement of the alkyne to form a phosphenium-ketone adduct III_P. Catalysis now proceeds (Figure 4b) via Si–H addition across the C=O moiety [TS(III–IV)_P, +27.9 kcal/mol] to form IV_P, in which partial hydride transfer has occurred (C···H = 1.29 Å; H···Si = 1.71 Å). Silyl group migration then readily occurs to form V_P at −7.8 kcal/mol, featuring a silylalkyloxonium adduct of the {Ph₂P} moiety. Ketone attack at V_P then displaces the hydrosilylation product and generates III_P that can then enter a second cycle with further reaction with silane.³⁸

Figure 4.

Free energy reaction profiles (kcal/mol) for the reactions of Me₂PhSiH with acetone mediated by the phosphirenium cation 3⁺ computed at the M052X-D3(CH₃CN)/def2-TZVP//TPSS/def2-SVP level of theory. (a) Initiation via ketone attack at P in the ring-opened ethenylium cation, II; (b) catalytic cycle based on intermediate III_P; (c) competing vinyl phosphine formation; and (d) competing phosphirenium reformation.

Overall, the calculations imply a relatively slow initiation process via TS(II–III)_P at 28.6 kcal/mol to form III_P, from which catalysis can then proceed with an energy span of only 10.8 kcal/mol. Catalysis from III_P should therefore be very efficient; however, the situation is complicated by at least two possible competing reactions. The first of these is vinyl phosphine formation (Figure 4c). This involves silane attack at the C1 position of intermediate II and proceeds via TS(II–III)_H at +30.7 kcal/mol with net insertion into the Si–H bond to give silylalkylphosphonium ion III_H,Z at −9.0 kcal/mol, featuring a short P–C bond (1.76 Å). Assuming facile rotation about the C–C bond leads to III_H,E (−7.3 kcal/mol), which can react with ketone with displacement of a silylether cation to form E-vinyl phosphine IV_H,E (−33.3 kcal/mol).³⁹ In principle, IV_H,E could re-enter catalysis via hydride transfer to the silylether cation, generating the hydrosilylation product and intermediate II. However, this process has a computed barrier of 48.3 kcal/mol, and so the formation of vinyl phosphine represents an unproductive off-cycle process. The second competing process (Figure 4d) involves the reaction of V_P with alkyne via TS(V–II)_P at +3.3 kcal/mol with the release of the hydrosilylation product and formation of II. II can then rapidly undergo ring closure to reform I.

The significance of these two competing processes is captured in the free energy differences (ΔΔG) between the rate-limiting transition states. For vinyl phosphine formation, these are TS(II–III)_H and TS(II–III)_P, where the latter is more stable and ΔΔG₁ = +2.1 kcal/mol. The productive onward reaction to form III_P should therefore dominate; however, the small energy difference suggests that vinyl phosphine formation may be significant at higher silane concentrations, and this is consistent with the experimental identification of an inverse first-order reaction with respect to silane. For the competing reactions of V_P with either the ketone or the alkyne, the key transition states are TS(III–IV)_P and TS(V–II)_P, respectively. In this case, TS(V–II)_P is only marginally disfavored with ΔΔG₂ = +0.4 kcal/mol, and this small energy difference suggests reaction with alkyne and therefore the reformation of I will be competitive. However, unlike the vinyl phosphine IV_E, I remains a competent catalyst precursor, albeit one that reacts with a relatively high overall barrier of 28.6 kcal/mol. There is therefore an ensemble of processes at play, in which efficient catalysis via III_P is balanced by catalyst loss through vinyl phosphine formation and a less efficient cycle based on phosphirenium precursor I. The net effect of these various processes is qualitatively consistent with a reaction that requires several hours at 60 °C to reach full conversion.

2.3.1. Substituent Effects

Reaction profiles computed for cations 9⁺ and 10⁺ were similar to those for the parent species 3⁺ (see the Supporting Information). However, subtle changes in the relative energies of the key competing transition states are computed that have a significant impact on the behavior of these systems (see Figure 5). For 9⁺, the electron-withdrawing p-CF₃ substituents cause an increase in both ΔΔG₁ and ΔΔG₂. As a result, catalysis becomes more efficient both in terms of a reduced likelihood of vinyl phosphine formation and a lower propensity to reform the phosphirenium cation. This is reflected in the short (ca. 1 h) times to full conversion seen experimentally with 9⁺. In contrast, the electron-donating p-OMe substituents switch the preference toward vinyl phosphine formation from II (ΔΔG₁ = −1.8 kcal/mol) and favoring the reaction of V_P with alkyne that leads to the formation of I over the more direct catalytic reaction with ketone. Both factors should render catalysis less efficient, consistent with the very sluggish catalytic reaction that is observed with 10⁺ (58 h to full conversion). The calculations indicate that while the p-OMe substituents in 10⁺ promote ring opening [ΔG_I→II = +17.6 kcal/mol (p-OMe), +22.0 kcal/mol (p-H), and +24.1 kcal/mol (p-CF₃)], they also disfavor the subsequent {Ph₂P⁺} transfer via TS(II–III)_P, and this results in hydride transfer via TS(II–III)_H dominating.

Figure 5.

Comparison of key transition states (free energies, kcal/mol) for the competing processes outlined in Figure 4. ΔΔG₁ = TS(II–III)_H – TS(II–III)_P and ΔΔG₂ = TS(V–II)_P – TS(III–IV)_P.

These combined experimental and computational studies allow us to propose a reaction mechanism for the hydrosilylation of carbonyls catalyzed by phosphirenium cations. These species act as a masked phosphenium source that is accessed via ring opening. Productive turnover then involves associative phosphenium transfer to the ketone, Si–H addition over the C=O bond, and associative displacement of the hydrosilylation product with further ketone. Spectroscopic and computational insights show that the formation of vinyl phosphine is an irreversible off-cycle process that will sequester the {Ph₂P⁺} moiety through reaction with silane, in accordance with inverse order kinetics. Para-substituents on the phosphirenium aryl groups can significantly impact the behavior; electron-donating p-OMe substituents undermine catalysis by favoring the formation of side products.

Experimentally, the observation of HPPh₂ and P₂Ph₄ by in situ reaction monitoring also supports the presence of the ring-opened form in the catalytic reaction. Given the capricious nature of unstabilized phosphenium ions, we do not observe free {Ph₂P⁺} in the catalytic reaction. Calculations indicate that loss of {Ph₂P⁺} from the parent phosphirenium is uphill by 34 kcal/mol; thus, observation of the free species is unlikely, lending further support to it being transferred directly between nucleophiles during the cycle. Catalytic reactions do show the presence of a phosphinate peak in the ³¹P NMR spectrum (−36.5 ppm, corresponding to benzhydryl diphenyl phosphinate³¹), which gives tentative support for the presence of phosphenium adducts.

2.4. Hydrosilylation Substrate Scope

With the optimal conditions for catalytic hydrosilylation and choosing to use commercially available diphenyl acetylene for in situ pre-catalyst formation (Table 2, entry 6), we explored the generality of the reaction. Compound 13a can be isolated in 73% yield, after 5 h at 60 °C using 2.5 mol % 3·B_Ar^F (Figure 6). Aryl-substituted ketones react well under the reaction conditions, giving 13b-13d in excellent 83–97% isolated yields. Electron-withdrawing and electron-donating groups in the para-position are well tolerated. These results are interesting because diaryl ketones have been reported to typically react to yield fully deoxygenated products when subjected to higher temperatures or longer reaction times.⁴⁰

Figure 6.

Reaction conditions: (i) activation step Ph₂P(O)H 2 (5 mol %), Tf₂O (5 mol %), NaB_Ar^F (5 mol %), diphenylacetylene 1 (5 mol %), CD₃CN (0.41 M), 60 °C, and 30 min and (ii) 2,6-lutidine (5 mol %), ketone or aldehyde (0.25 mmol), silane (0.25 mmol), 80 °C, and 5 h, reported as an average of two runs; ^aconversion; ^b10 mol %, 18 h; ^cspectroscopic yield.

In this procedure, the higher temperature does not affect the reaction outcome with hydrosilylation of diaryl ketones occurring selectively in 5 h. Unsymmetrical ketones with an aryl substituent and C_sp³ and C_sp² substituents yield 13e and 13f in full conversion (77 and 80% isolated yields, respectively). Alkyne-containing carbonyl substrates are not suitable for this transformation, possibly due to alkyne scrambling of the propargyl group.⁴¹ Cyclohexanone and acetone react well using 10 mol % 3·B_Ar^F, giving 13g and 13h in 87 and 62% yields, respectively, while acetophenone reacts sluggishly, and even with 10 mol % catalyst loading, only 23% of 13i is obtained. The use of other tertiary silanes was found to be suitable, and 13j is obtained in full conversion and high yield (85%). However, triethoxy silane and triphenyl silane do not convert, but the phosphirenium pre-catalyst is unaffected and is still observed in the reaction mixture. Secondary or primary silanes do not react with benzophenone 11 under the reaction conditions, and decomposition of the phosphirenium is observed by NMR spectroscopy.

Several aldehydes were subjected to the optimized reaction conditions, and the results are good to excellent when using electron-poor aryl aldehydes, with 60–98% isolated yields for 13k–13p. Cinnamaldehyde reacts smoothly to yield 13q in 75% yield. The reaction also works well with 4-methylbenzaldehyde, furnishing 13r in 89% isolated yield, but is quite limited with other electron-rich aldehydes. 4-Dimethylamino benzaldehyde shows low conversion into 13s (22%) after 18h with 10 mol % 3·B_Ar^F; 4-methoxybenzaldehyde does not show any conversion after 24 h, which suggests a potential deactivation of the carbonyl or catalyst poisoning. Electron-withdrawing groups in the ortho- and meta-position are detrimental, giving low conversion and yield of 13t, 13u, and 13v. Tiglic aldehyde reacts using 10 mol % catalyst at 80 °C with full conversion of silane to 13w in only 40% spectroscopic yield, with traces of the product derived from the reduction of the double bond observed in situ. Alkyl aldehydes such as pentanal, heptanal, and dodecanal give 13x, 13y, and 13z in 72–88% isolated yields, respectively, and cyclopropane aldehyde reacts to give 13aa in 77% isolated yield with no ring opening observed, highlighting that radicals are not involved in the reaction.

3. Conclusions

The ease of synthesis and the newly discovered reactivity of phosphirenium ions allow us to conclude that they are effective organocatalysts for the hydrosilylation of ketones and aldehydes; to the best of our knowledge, this is the first report of a phosphirenium ion being used as a pre-catalyst. Full characterization of the phosphirenium ions bearing triflate or borate counterions is reported, and further details on their synthesis were explored. The practical pre-catalyst synthesis from typically cheap and readily available starting materials such as phosphine oxides and alkynes allows a straightforward catalytic procedure that is easily scalable. Our mechanistic investigations give insights into the principles governing their reactivity; this is underpinned by DFT studies and experimentally determined kinetics and stoichiometric studies. The ability of these phosphirenium ions to undergo ring opening paves the way for several reaction pathways, with the dominant and most favorable pathway being the release of the active catalyst, which is a phosphenium ion adduct. Increasing the electron-donating ability of the alkyne used to prepare the pre-catalyst slows catalysis, allowing an off-cycle deactivation process to dominate. These elusive phosphenium species are highly reactive and challenging to isolate and characterize; therefore, by disguising a phosphenium catalyst as a more stable and manageable phosphirenium adduct, we open up a multitude of opportunities in organic synthesis. Work is still ongoing in our laboratory to shed light on diverse reactivity and the activation modes of phosphirenium rings and the broader applicability of masked phosphenium catalysts.

4. Experimental Section

4.1. General Procedure for Synthesis of Phosphirenium Ions

In a J-Young NMR tube or a Schlenk tube, alkyne (0.25 mmol), SPO (0.25 mmol), triflic anhydride (1 equiv.), and NaB_Ar^F (1 equiv.) were stirred in CD₃CN or CDCl₃ at 60 °C for 30 min. The reaction was monitored by ³¹P{¹H} NMR spectroscopy until complete disappearance of the phosphine oxide peak. After the reaction was completed, the mixture was concentrated under vacuum. The product was washed with pentane and recrystallized at −20 °C.

4.2. General Procedure for the Hydrosilylation of Ketones and Aldehydes

To a J-Young NMR tube, dicyclohexyl phosphine oxide (5 mol %, 12.5 × 10^–3 mmol, 2.7 mg) or diphenyl phosphine oxide (5 mol %, 12.5 × 10^–3 or 25 × 10^–3 mmol, 2.5 mg), diphenylacetylene (5 mol %, 12.5 × 10^–3 mmol, 2.3 mg), triflic anhydride (5 mol %, 12.5 × 10^–3 mmol, 2.1 μL), and NaB_Ar^F (5 mol %, 12.5 × 10^–3 mmol, 11.1 mg) were stirred in CH₃CN (3 mL) for 30 min at 60 °C. After formation of pre-catalyst 3·B_Ar^F, the reaction mixture was cooled down, and 2,6-lutidine (5 mol %, 12.5 × 10^–3 mmol, 1.5 μL), benzophenone (0.25 mmol, 45.5 mg), and dimethylphenylsilane (0.25 mmol, 38.3 μL) were added, and the reaction was heated at 80 or 60 °C for 18 and 5 h, respectively. After drying the volatiles, the pure product was obtained by purification by flash chromatography (9:1 petroleum ether/EtOAc) or vacuum distillation.

4.3. DFT Calculations

Gaussian 09 (Revision D.01) was used with geometry optimizations carried out at the TPSS/def2-SVP level and all stationary points characterized with analytical frequency calculations. Electronic energies were recomputed at the M052X-D3(CH₃CN)/def2-TZVP level, which has been shown to be a reliable approach for main group kinetics, thermodynamics, and non-covalent interactions.⁴² These corrected energies were then combined with the thermodynamic corrections from the gas-phase calculations to give the free energies quoted in the text. Full details and references are supplied in the Supporting Information.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acscatal.1c01133.

• Analysis data and NMR spectra for all products and computational supporting information which contains all data relevant to the computational studies reported here, including methodology, references, benchmarking studies, energetics, and coordinates of characterized stationary points (CIF)

• General considerations, synthesis and characterization of the starting material, synthesis of phosphirenium ions, substrate scope hydrosilylation, and optimization of the catalytic reaction (PDF)

• Computed geometries for all stationary points (XYZ)

Author Contributions

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

The authors declare no competing financial interest.