Hydrophosphination of Activated Alkenes by a Cobalt(I) Pincer Complex

Herein we report the synthesis of three heteroleptic first-row transition metal(II) complexes containing carbazolido NNN pincer ligands and conversion to the corresponding metal(I)-carbonyl complexes via a reductive carbonylation route. These complexes are precatalysts for the hydrophosphination of activated alkenes, affording a cobalt-catalysed hydrophosphination process that solely and selectively yields the β addition (anti-Markovnikov) product. The scope of this transformation has been investigated using a variety of activated alkenes. Isolation and characterisation of substrate-coordinated intermediates reveal available coordination sites, which provide insight into the proposed catalytic cycle.


Introduction
Phosphorus-containing compounds are a precious commodity, finding use in numerous areas such as organocatalysis, [1] bulk and fine chemical production, [2] and the pharmaceutical industry. [3] Aiming to access new synthetic routes for their preparation, these industries have stimulated the development of more effective, atom-economical routes and viable strategies for their preparation. However, a continuing challenge in this area is the ability to selectively and cleanly access compounds of interest.
From 3 and 4, the metal(I)-carbonyl complexes could be formed by treatment with one equivalent of LiBHEt 3 and exposure of the in situ generated hydride complex (3-H/4-H) to an atmosphere of CO(g) (3-CO and 4-CO; Scheme 2 and Figure 1). [47,48] Extraction of the reaction mixtures into toluene afforded 3-CO and 4-CO, allowing for characterisation by NMR spectro-  scopy, mass spectrometry, IR spectroscopy, and (for 4-CO) single crystal X-ray diffraction ( Figure 1). Samples of analytical purity were not obtained, and these carbonyl complexes were generated and utilised in situ for all subsequent catalytic reactions.

Catalytic Olefin Hydrophosphination and Reaction Optimisation
An initial assessment of the catalytic activity for the hydrophosphination of alkenes (Scheme 3, Table 1) was performed via the reaction of acrylonitrile with diphenylphosphine in C 6 D 6 using 5 mol% of 3, (Entry 3, Table 1). No evidence for the formation of hydrophosphination products was observed at rt. However, upon heating at 60°C for 22 h, selective formation of the linear isomer (A, compound 10) was observed (20% conversion). Heating for four days afforded ca. 50% conversion of the starting HPPh 2 , with the linear isomer A, as the main product (49% vs. the branched isomer B 2%). Looking to aid deprotonation of HPPh 2 , and to ensure the solubility of complex 3 (5 mol%) during catalysis, an excess of NEt 3 was added to the reaction. However, under these conditions, there was a decrease in conversion relative to the analogous experiment in absence of base (11% yield of the linear isomer, Entry 4, Table 1). However, the cobalt containing 4 exhibited a higher activity than iron containing 3 when NEt 3 was used as an additive (Entry 6, Table 1) giving 61% total conversion after 18 h, with an overall 58% yield of the linear isomer.
Under identical conditions, manganese complex 5 only achieved a 13% conversion after 18 hours, affording the β addition (anti-Markovnikov) isomer in 12% yield (Entry 7, Table 1). Returning to the more promising precatalyst 4, increasing the reaction temperature to 80°C afforded 88% conversion with the formation of the linear product in 74% yield (Entry 8, Table 1). The replacement of the bromide in 4 with a hydride was expected to increase the catalytic activity of the metal complex (Entry 9, Table 1). [30] However, the in situ generated hydride 4-H exhibited decreased reactivity when compared with 4 under similar conditions (70% conversion after 22 h). Surprisingly, when the in situ generated cobalt(I)-carbonyl complex 4-CO was tested in toluene, high conversions were achieved in < 2 h at 80°C, with excellent (ca. 100%) selectivity for β-addition (Entry 10, Table 1, see SI31-SI32). The same reaction using the  (4) (11). analogous iron complex (3-CO) only afforded a 59% conversion after 17 hours, with significant formation of the branched isomer B and the dehydrocoupling product C (Entry 11, Table 1). It is noteworthy that, whilst catalytic hydrophosphination of alkenes by first-row transition metals is known, [23,25,32] to our knowledge this is the first cobalt-based example. However, the analogous insertion in the more reactive alkynes is well documented for both hydrophosphination [26,27] and hydrosilylation. [29][30][31]49] Based on the catalyst screening and optimisation experiments, it was decided to proceed using the conditions described in Entry 10 (Table 1). In order to determine the nature of the catalytic process, poisoning experiments with Hg [24,50] and CS 2 , [4,51] were performed (Table SI1). No changes were observed in the reaction rate or product selectivity in the presence of Hg or CS 2 suggesting that the reaction most likely occurs through a homogenous mechanism. Moreover, the reaction does not appear to be radical mediated, since the presence of cumene [4,23,52,53] or 1,4-cyclohexadiene, [54,55] does not diminish the activity of the catalysts (Table SI1).

Substrate Scope
Having established complex 4-CO as the most selective/ active metal precatalyst, the reaction scope was investigated with a variety of activated unsaturated substrates. Results of the hydrophosphination reactions catalysed by cobalt(I) are shown in Table 2. Using an α-carbonyl unsaturated substrate such as methyl acrylate, the reaction reaches full conversion to the linear product 11 in 5 h (Entry 1, Table 2). Substitution of the α-carbon with a methyl group results in a significant drop in conversion (26%) despite prolonged reaction times (Entry 2, Table 2). This is most likely a consequence of steric hindrance. For activated non-terminal alkenes (fumaronitrile and dimethylfumarate) only the less sterically hindered fumaronitrile was susceptible to substitution, with full conversion to 13 in 1 h (Entry 3, Table 2). When the double bond was located in the β-position, even when activating functional groups are present (Table SI2, entries 5 and 6), no hydrophosphination was observed. It was noted that in all cases where coordination of the substrate to the catalyst is not possible, the dehydrocoupling product Ph 2 PÀ PPh 2 (C) is obtained as the sole product. This suggests that, in substrates where hydrophosphination is unfavourable, an alternative pathway is enabled (vide infra) in which the stoichiometric transformation of HPPh 2 to Ph 2 PÀ PPh 2 (C) occurs (See Table SI2). When vinylpyridines were used (Entries 4 and 5, Table 2), full conversion to the linear isomers (14 and 15) was achieved in 40 and 20 h, respectively. Similarly, the α-unsaturated ketone 2-cyclohexen-1-one gave selective substitution in the β-position with up to 81% conversion after 24 h (Entry 6, Table 2). In line with the reactivity described for Entries 1-3, a preference for activated terminal alkenes was observed when the isomeric lactones 5,6-dihydro-2H-pyran-2-one (Table SI2, entry 14) and α-methylene-γ-butyrolactone were employed, the latter yielding exclusively the β addition (anti-Markovnikov) product 17 in 4 h (Entry 7, Table 2). The low conversion of 5,6-dihydro-2H-pyran-2-one could be due to a combination of disfavoured attack at a secondary carbon, and the reduced electrophilicity of conjugated esters vs. alkenes (comparing to the sterically   Table 2. Summary of substrates susceptible to hydrophosphination catalysed by 4-CO. [a] Reaction conditions: 5.0 mg, 6.92 × 10 À 3 mmol of 4-CO (5 mol%), 0.6 mL of toluene, 20 equiv. of alkene/HPPh 2 . Samples were heated in an oil bath at 80°C, progress was monitored by NMR spectroscopy.   Table 2) are also substrates for substitution, with moderate conversions in under 6 hours. However, an aromatic imine with ortho-iPr substituents was found to give no conversion under the same conditions (see Table SI2, entry 18). This may result from steric hindrance preventing coordination of the imine N to the catalyst. A methyl-substituted imine, by contrast, achieved relatively modest conversions when compared to the less bulky aryl imines (38% conversion in 5 h, Table 2, entry 10). This could be due to the more electron rich C=N bond disfavouring nucleophilic attack on the substrate. For the initial steps in transition metal catalysed hydrophosphination reactions, two possible reaction pathways have been proposed. The first involves initial coordination of the secondary phosphine to the metal centre, yielding a metal-phosphide complex, and is favoured in systems catalysed by platinum, [56][57][58][59][60][61] lanthanides [62][63][64][65] and in some cases iron. [5,16] An alternative mechanism, involving the formation of a metal-olefin complex, is less common [66] and is most commonly reported for late first-row transition metals [9,67] and in limited cases by noble metals. [67][68][69] For analogous reactions involving the hydroamination of alkenes, the η 2 -olefin is not directly activated upon coordination by the metal centre. Activation occurs when the olefin changes from η 2 to η 1 coordination and results in localisation of the LUMO for the ligand-M-{alkene} complex on the most distant carbon atom. This enhances the interaction with the incoming nucleophilic phosphine, as expected for phospha-Michael type reactions, a key feature in the reaction mechanism that explains the selective formation of β addition (anti-Markovnikov) product. [69][70][71] Attempts to form a metal-phosphide complex between HPPh 2 and 4/4-CO were unsuccessful, even when stoichiometric amounts of NEt 3 or K[N(SiMe 3 ) 2 ] were employed. Nevertheless, we have been able to prepare and structurally characterise examples of metal-substrate complexes using 4 and the unsaturated substrates acrylonitrile (6), dimethyl fumarate (8) and methyl acrylate (9) (Figure 2 and SI39-SI42 for further details) to form five-coordinate compounds of general formula [Naph 2 carbCo{substrate}Br] (6-9). In these, the substrate is bound to the metal via an electron-rich nitrogen or oxygen. These examples highlight the space available for coordination and reactivity around the low-coordinate metal centre.

Proposed Mechanism
The mechanism depicted in Scheme 4 attempts to reconcile our observations. Initially, 4-CO coordinates the unsaturated substrate, forming the coordinated complex I, similar to complexes 6-9 (Scheme 4 and SI39-SI42). As it has been formally reduced, one

FULL PAPER
asc.wiley-vch.de might expect Co(I) species 4-CO to be a weaker Lewis acid than Co(II) complex 4. However, this is not necessarily the case, as 4-CO has undergone a geometry change relative to 4 (square planar vs. seesaw) and a π-donating ligand (Br) replaced with a strong π-acceptor (CO). Indeed, there are several reports of Co(I) species behaving as Lewis acids under an appropriate ligand environment. [72][73][74] Initial loss of carbon monoxide from the parent metal complex, via an initial carbonylation mechanism, is unlikely since NMR ( 1 H and 13 C{ 1 H}) studies show neither the formation of carbonylic products nor free CO and mass spectrometric determinations do not give molecular ions consistent with CO homologation into the substrate. This suggests that 4-CO is the catalytically active species. After coordination, intermediate I undergoes nucleophilic substitution by the secondary phosphine (Michael addition). Kinetic determinations previously reported using an isostructural (four coordinate/square planar) and isolectronic (d 8 ), Ni(II) complex to 4-CO, [Ni(k 3 -Pigiphos)(N � CMeCCH 2 )] 2 + , have shown this initial step is reversible due to low energy barriers for nucleophilic attack and elimination by phosphines. [66] This step affords intermediate II, in a regioselective manner, with subsequent proton transfer to yield III. It should be noted that attempts to isolate and characterise intermediate III, or any of its analogues, were unsuccessful (stoichiometric reactions). This suggests that rapid elimination of product 10 occurs and regenerates the cobalt(I) catalyst (4-CO), which then coordinates an additional molecule of acrylonitrile, completing the catalytic cycle. An alternative reaction mechanism involving a Co(I)/Co(III) cycle, as proposed in some recent examples of cobalt hydrofunctionalisation reactions, [75,76] cannot be ruled out with our experimental data; with current investigations in our group targeting a detailed analysis of the reaction mechanism for this transformation. Finally, our catalytic experiments show that for reactions where a metal-olefin complex cannot be formed, or if the unsaturated substrate is too sterically hindered or inactive (e. g. styrene), then the reaction produces stoichiometric quantities of the dehydrocoupling product, Ph 2 PÀ PPh 2 . A similar observation has been previously reported by Webster et al., in processes catalysed by β-diketiminate iron complexes. [5,21]

Conclusions
We have reported the synthesis and characterisation of three heteroleptic metal(II) NNN pincer complexes. From these, two metal(I)-carbonyl complexes have been prepared using a reductive carbonylation process. All of these complexes are catalytically active in the hydrophosphination of activated olefins, with the cobalt(I) carbonyl complex showing significant promise for this reaction. The method has been extended to a range of substrates, yielding selective β addition (anti-Markovnikov) products. Experiments to characterise the mechanism by which the reaction takes place has allowed us to draw similarities with equivalent heteroatom insertion reactions, founded in crystallographically authenticated analogues.

Experimental Section
Apart from the synthesis of the ligand (Naph 2 carbH, 1) and the phosphine oxides, all products described were treated with rigorous exclusion of air and water using standard air-sensitive handling techniques which included bench-top operations (Schlenk line) and glove-box techniques. NMR samples of air and moisture sensitive compounds were prepared using glove box techniques and contained in Young's tap modified borosilicate glass NMR tubes. NMR data were collected on either a Bruker DPX300, DPX400, AV400, AV(III)400, AV(III) 400HD or AV(III)600 spectrometer. Chemical shifts are quoted in ppm relative to TMS ( 1 H, 13 C{ 1 H}) and H 3 PO 4 ( 31 P, 31 P{ 1 H}). Reaction progress was monitored by quantitative 31 P, 31 P{ 1 H} NMR spectroscopy (inverse gated decoupled) of samples prepared in dry, non-deuterated toluene, with a C 6 D 6 insert for locking. Apart from the substrates employed in Entries 8-9, Table 2, which were prepared following reported procedures; [77] all reagents were used as received. Magnetic moments were calculated through the Evans method at 298 K, employing C 6 D 6 as solvent. Diamagnetic corrections were calculated according to Pascal's constants. [78] CCDC 1874664-1874672 contain the supplementary data for 3-9, 4-CO and 20 a. These data can be obtained free of charge from The Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/data_request/cif. Safety warnings: 1-Naphthylamine is highly toxic and is suspected to be a carcinogen; great care must be taken during synthesis and adequate handling of waste should be procured. Carbon monoxide is an extremely toxic and flammable gas, good ventilation within a fumehood should be procured when running a Schlenk line with this gas as source, away from open flames and with a carbon monoxide detector operating at all times.
Solutions containing Li[BHEt 3 ] (Super-Hydride ® ), are extremely pyrophoric and flammable upon exposure to air, ensuring handling of this compound under inert conditions is paramount.

Typical Procedure for the Formation of Metal(I) Complexes (Reductive Carbonylation; 3-CO and 4-CO) For Catalytic Scale Reactions
In a glovebox, 3 (5 mg, 6.68 × 10 À 3 mmol) or 4 (5 mg, 6.92 × 10 À 3 mmol) was dissolved in dry toluene (0.6 mL). The resulting dark purple solution was transferred to a Young's tap modified borosilicate glass NMR tube. Shortly after, LiBHEt 3 (1 M solution in THF, 7 μL, 7 × 10 À 3 mmol) was added to the NMR tube rendering the solution black. The NMR tube was sealed, removed from the glovebox and connected to a Schlenk line running on CO(g). The tubing connecting the NMR tube to the line was thoroughly cycled before the tube was opened and the atmosphere within the tube was replaced with CO(g) via three freeze-pump-thaw cycles. The resulting mixture was left under this atmosphere for 22 hours. Crystals suitable for XRD, for 4-CO, were grown in a glovebox from a saturated C 6 D 6 solution via hexane vapour diffusion, at rt. [Repeated attempts at growing crystals of 3-CO suitable for study by single X-ray diffraction were unsuccessful. Despite this, multinuclear structures relying on bridging carbonyls have been ruled out, due to absence of characteristic υ CO signals at lower wavenumber (1800-1500 cm À 1 )]. Brown powder (14.8 mg, 55%, see ESI for large-scale synthesis).

Typical Procedure for the Hydrophosphination of Activated Alkenes
To a solution of the isolated or in situ generated 4-CO (5 mg, 6.92 × 10 À 3 mmol) in toluene (with C 6 D 6 insert), acrylonitrile (9.06 μL, 0.138 mmol, 20 eq.) and then HPPh 2 (24 μL, 0.138 mmol, 20 eq.) were added. The resulting NMR sample was transferred to an oil bath set at the desired temperature ( Table 1 and Table 2) and the reaction was monitored until the resonances attributed to the starting material disappeared ( 31 P NMR spectroscopy) or when no significant reaction progress was observed (10). In order to isolate the hydrophosphination products, the reaction mixtures were purposely oxidised (10 a). [16] In all cases the crude mixture was opened to air and added to a silica gel plug (petroleum ether 40-60) to remove the unreacted HPPh 2 . The product was eluted with Et 2 O, this fraction was exposed to H 2 O 2 (30% w/w, 5 mL) and stirred at rt for 10 min. The reaction mixture was quenched with deionised water and the organic phase was separated, dried over MgSO 4 and evaporated. In the particular case of products 15 a and 16 a, the crude product after oxidation with H 2 O 2 , was extracted with CH 2 Cl 2 due to the increased solubility of the products in H 2 O. For full characterisation of the hydrophosphination products, see the supporting information.