A Mild One-Pot Reduction of Phosphine(V) Oxides Affording Phosphines(III) and Their Metal Catalysts.

The metal-free reduction of a range of phosphine(V) oxides employing oxalyl chloride as an activating agent and hexachlorodisilane as reducing reagent has been achieved under mild reaction conditions. The method was successfully applied to the reduction of industrial waste byproduct triphenylphosphine(V) oxide, closing the phosphorus cycle to cleanly regenerate triphenylphosphine(III). Mechanistic studies and quantum chemical calculations support the attack of the dissociated chloride anion of intermediated phosphonium salt at the silicon of the disilane as the rate-limiting step for deprotection. The exquisite purity of the resultant phosphine(III) ligands after the simple removal of volatiles under reduced pressure circumvents laborious purification prior to metalation and has permitted the facile formation of important transition metal catalysts.

Introduction

Applications of Phosphine(III) Ligands and Synthesis

class="Chemical">Phosphines and their derivatives are of class="Chemical">pan class="Chemical">significant importance to both academic and industrial chemistry. In particular, within organic chemistry phosphine(III) compounds have a distinguished history, mediating classical transformations such as the Appel,[1] Mitsunobu,[2] and Wittig[3,4] reactions. Additionally, the ready modulation of electronic and steric properties of phosphine(III) has made them excellent ligands for the formation of well-defined transition metal complexes,[5] although recalcitrant phosphine(V) oxides arise, when phosphine(III) compounds are employed as labile ligands[6] or the metal complexes are simply decomposed, in the presence of a suitable oxidant.[7] Arguably, the stoichiometric formation of phosphine(V) oxide waste from the above-named organic reactions presents an even greater issue, especially on the industrial scale,[3,4] as the conversion of P(V)=O to the P(III) oxidation state is nontrivial (vide infra).

Direct Reduction of Phosphine(V) Oxide

Given the class="Chemical">significance of class="Chemical">pan class="Chemical">phosphine(III) compounds, a variety of anaerobic syntheses have been reported.[8,9] However, the sensitivity of phosphine(III) to oxidation (requiring only minutes to hours) has led to the widespread use of “protected” phosphines,[10] such as phosphine–borane adducts[11,12] and phosphine(V) sulfides[13,14] but predominantly phosphine(V) oxides.[15−17] These precursors tolerate the reaction conditions necessary to construct more complex architectures[18] although the protection must be removed in the penultimate[12,19] or final[20,21] step of the ligand synthesis. Thus, much attention has been focused on the conversion of P(V)=O to P(III)[15,16] (Scheme a), including the use of silanes and siloxanes such as HSiCl3,[22−25] HSiCl3/Ph3P,[26] Si2Cl6,[24,27] Si2Me6 with CsF/TBAF,[28] HSi(OEt)3/Ti(O-i-Pr)4,[29] PhSiH3,[30−32] 1,1,3,3-tetramethyldisiloxane (TMDS) with CuX2,[33] polymethylhydrosiloxane (PMHS),[34,35] 1,3-diphenyldisiloxane (DPDS),[36] and (EtO)2MeSiH/(RO)2P(O)OH;[37] aluminum hydrides such as LiAlH4,[38,39] LiAlH4/CeCl3,[40] AlH3,[41] and HAl(i-Bu)2;[42] low-valent metals such as SmI2/HMPA (hexamethylphosphoramide)[43] or Cp2TiCl2/Mg;[44] hydrocarbon/activated carbon;[45] and electrochemical reduction.[46−48] A mild iodine-catalyzed reduction of phosphine(V) oxides employing a sacrificial electron-rich phosphine was developed by Laven and Kullberg,[49] while Li et al.[50] employed less expensive phosphite, although in both cases P(V)=O-containing contaminants must be removed from the final products. Thus, disadvantages of these procedures include harsh reaction conditions, toxic and/or highly reactive, potentially explosive reducing agents, narrow scope or undesirable side reactions, e.g., C–P,[51,52] C–O,[52] or P–N[53−56] bond cleavage, and laborious column chromatography to purify the desired phosphine(III).

Scheme 1

Phosphine Synthesis: Background and This Work

Left: (a) direct reduction of P(V)=O or P(V)=S affording P(III); (b) conversion of P(V)=O or P(V)=S to activated phosphonium salt; (c) reduction of activated phosphonium salt to P(III); (d) conversion of activated phosphonium salt to phosphine–borane; (e) deprotection of phosphine–borane affording P(III). Right-top: (f) BASFs conversion of Ph3PO to Ph3P using phosgene and silicon. Right-bottom: (g) Paradies et al. recent conversion of Ph3PO to Ph3P using oxalyl chloride and pressurized hydrogen. Center-bottom: (h) this work.

Left: (a) direct reduction of P(V)=O or P(V)=S affording P(III); (b) converclass="Chemical">sion of P(V)=O or P(V)=S to activated class="Chemical">pan class="Chemical">phosphonium salt; (c) reduction of activated phosphonium salt to P(III); (d) conversion of activated phosphonium salt to phosphine–borane; (e) deprotection of phosphine–borane affording P(III). Right-top: (f) BASFs conversion of Ph3PO to Ph3P using phosgene and silicon. Right-bottom: (g) Paradies et al. recent conversion of Ph3PO to Ph3P using oxalyl chloride and pressurized hydrogen. Center-bottom: (h) this work.

Reduction of Activated Chlorophosphonium Salts

The inherent stability of the P(V)=O has compelled others to explore sequential activation reduction methods, i.e., the converclass="Chemical">sion of the class="Chemical">pan class="Chemical">phosphine(V) oxide to more reactive chlorophosphonium salts (CPS) and subsequent reduction (Scheme b,c). Horner, Hoffmann, and Beck first published the reduction of chlorotriphenylphosphonium chloride (Ph3PCl2) in 1958,[57] with both LiAlH4 and sodium. The following year a sequential activation and deprotection was published, converting triphenylphosphine(V) oxide (Ph3PO) first to activated CPS, Ph3PCl2, before it was reduced to triphenylphosphine (Ph3P) with sodium metal.[58] Being readily afforded via inexpensive chlorinating reagents,[59] CPSs have also been reduced with aluminum/metal salts,[60] alkali metals,[57,58] LiAlH4,[57,61,62] thiols/Et3N,[63] activated carbon,[45] Hantzsch ester/Et3N,[64] electrochemically,[46−48,65,66] elemental aluminum[67,68] or silicon,[69] and hydrogenolysis,[70] which may be catalyzed by frustrated Lewis pairs (FLPs).[71,72] Harsh metal bases and Grignard reagents have even been used to deprotect certain CPSs.[73] Alternatively, CPS can be converted to phosphine–boranes by either NaBH4[74,75] or LiBH4,[76−79] although ultimately the borane “protecting group” itself requires removal (Scheme b,d,e).

Motivation to Develop a New Facile Reduction of Phosphine(V) Oxides

Our interest in class="Chemical">phosphine(V) oxides reduction originates from our declass="Chemical">pan class="Chemical">sire to explore bulky N-phosphinomethyl-functionalized N-heterocyclic carbene ligands (NHCPs)[80,81] as potential ligands for new olefin metathesis catalyst (Scheme ).[19] Progress has been severely hampered due to difficulties accessing azolium salt 5, with the problematic reduction of 4 being achievable only with a large excess of trichlorosilane (27.0 equiv) in anhydrous degassed chlorobenzene at elevated temperature over 2 days.[19] As well as the lengthy reaction time, we experienced some reproducibility issues, with the unsuccessful reduction being accompanied by the decomposition of the precious azolinium 4, previously obtained via a multistep synthesis.[19] In light of this, a simple procedure for the conversion of 4 to 5 would be a great advantage. Such a process might also permit access to other challenging phosphine(III) and metal catalysts as well as permitting the recovery of the valuable phosphine(III) ligands: “closing the phosphorus cycle” is of increasing importance due to environmental and availability concerns.[82−84] Herein, we report a new activation/deprotection of phosphine(V) oxides without the use of harsh reaction conditions, metals, or sacrificial phosphanes. Intermediate CPSs are directly converted to desired phosphines by reaction with hexachlorodisilane. Mechanistic details have been elucidated by experimentation and supported by computation. The “one-pot” procedure affords excellent yields of pure phosphine(III) ligands that can be telescoped into formation of transition metal catalysts without the prior need for silica gel chromatography.

Scheme 2

Problematic Reduction of NHCP Precursor

Synthesis of NHCP 6via the challenging reduction of phosphine(V) oxide in azolium salt 4 to phosphine(III) 5.

Syntheclass="Chemical">sis of class="Chemical">pan class="Chemical">NHCP 6via the challenging reduction of phosphine(V) oxide in azolium salt 4 to phosphine(III) 5.

Results and Discussion

Reduction of Activated CPSs with Disilane

In 1996, BASF reported the generation of class="Chemical">tetrachlorosilane (class="Chemical">pan class="Chemical">SiCl4) when the CPS, Ph3PCl2 (2), was heated with elemental silicon at 185 °C.[69] Not wanting to expose our ligand precursor to such harsh reaction conditions, we hypothesized that hexachlorodisilane might serve as a suitable surrogate for elemental silicon and similarly generate 2 equiv of SiCl4 on reactions with a CPS. The abundant industrial byproduct Ph3PO (1) appeared to be the ideal test substrate,[3,4] and was easily converted to activated Ph3PCl2 (2) with inexpensive oxalyl chloride.[59] Gratifyingly on reaction with 1.1 equiv of hexachlorodisilane (Si2Cl6) at room temperature, both 1H NMR and 31P NMR indicated the immediate, clean, and complete formation of Ph3P (3), with 29Si NMR showing only the formation of tetrachlorosilane, SiCl4 (δ = −18.8 ppm). Motivated by the ability of Si2Cl6 to reduce 2, we chose to explore other disilanes (Table , entries 2–10): 1,1,2,2-tetrachloro-1,2-dimethyldisilane (Si2Me2Cl4), hexamethyldisilane (Si2Me6), and hexaphenyldisilane (Si2Ph6), which might generate the corresponding attractive byproducts trichloromethylsilane (MeSiCl3), trimethylsilyl chloride (Me3SiCl), or triphenylsilyl chloride (Ph3SiCl). However, the more electron-rich and sterically hindered disilanes generated the desired phosphines in either lower yield, over extended reaction times or not at all. For instance, the addition of a single electron-donating methyl group to each of the silicon atoms in Si2Me2Cl4 drastically decreased the rate of reaction, with only a 28% conversion to 3 after 24 h, eventually reaching completion after 144 h. In contrast, the reaction with Si2Cl6 was complete in under 5 min.a No reaction was observed for even more electron-rich and sterically shielded Si2Ph6 or Si2Me6.

Table 1

Reaction of Phosphonium Salts with Disilanes

entry	CPS 2a–c, X =	disilane	equiv	time	conv to 3 [%]a
1	Cl	Si2Cl6	1.1	5 min	100
2	Cl	Si2Me2Cl4	1.1	5 min	0
3	Cl	Si2Me2Cl4	1.1	1 day	28
4	Cl	Si2Me2Cl4	1.1	2 days	55
5	Cl	Si2Me2Cl4	1.1	3 days	72
6	Cl	Si2Me2Cl4	1.1	4 days	78
7	Cl	Si2Me2Cl4	1.1	5 days	83
8	Cl	Si2Me2Cl4	1.1	6 days	100
9	Cl	Si2Me6	1.0	1 day	0
10	Cl	Si2Ph6	1.0	1 day	0
11	OTf	Si2Cl6	1.1	10 min	7
12	OTf	Si2Cl6	1.1	1 day	80
13	OTf	Si2Cl6	1.1	2 days	100
14	BArCl	Si2Cl6	4	2 days	0

Conversion judged by 31P NMR of 2a–c relative to 3.

Converpan class="Chemical">sion judged by class="Chemical">pan class="Chemical">31P NMR of 2a–c relative to 3.

Scope of the New Procedure

With class="Chemical">Si2Cl6 class="Chemical">proving to be the reductant of choice, we exclass="Chemical">panded the aclass="Chemical">pclass="Chemical">plication of the class="Chemical">procedure to other class="Chemical">pan class="Chemical">phosphine(III) compounds.b Aliphatic tricyclohexylphosphine (7) was afforded in 97% yield, in contrast to the recently reported hydrogenation at 130 °C, which notably afforded none of the desired phosphine(III) complexes.[71] Cyclic alkene 2-phospholene oxide was also converted to P(III) 2-phospholene (8)[77] (98%) without the reduction or isomerization of the C=C bond. Reduction of phosphinamides without the P–N bond scission is particularly challenging;[53−56] while Gilheany et al. synthesized “protected” aminophosphine–borane adducts from CPSs in excellent yields,[75] we were able to furnish the free aminophosphine 9 directly (89%). The dimethylamino group in DavePhos 11 (93%) was also tolerated well, with fellow Buchwald ligand CyJohnPhos 10 being cleanly afforded in 95% yield. Chiral phosphines[8] are still of great significance, and we chose to explore binaphthyl systems as the CPSs of P-chirogenic phosphines are known to racemize.[85] The oxides of chiral phosphepines permit structure elaboration,[86] and our new method rapidly afforded (S)-Ph-BINEPINE (12)[87] (96% yield). (R)-MeO-MOP (13)[88] was also readily synthesized (99%). It is of note that the direct reaction of MeO–MOP oxide with Si2Cl6 in acetonitrile led exclusively to scission of the C–O bond without reduction of P(V)=O,[52] highlighting the divergence in the reactivity of the activated P(V)Cl2 compared to recalcitrant P(V)=O. Moreover, we observed no racemization in the case of either 12 or 13.

Having established the optimal conditions for the generation of a range of class="Chemical">phosphine(III) comclass="Chemical">pounds, we turned our attention back to class="Chemical">pan class="Chemical">azolinium 5. The reaction of 4 with excess oxalyl chloride yielded a new chlorophosphonium bearing azolinium salt 15 (after removal of 4-toluenesulfonyl chloride produced by chlorination of the 4-toluenesulfonate; see the Supporting Information) which was readily transformed to the desired azolium 5 with hexachlorodisilane (1.5 equiv). The identity of both salts 5 and 15 was established by single-crystal X-ray diffraction analysis. Crystals suitable for this purpose were obtained by layering methylene chloride with hexane and storing at −30 °C. The salts crystallize in the monoclinic P21/c (CPS 15) and P21/n (azolium 5) space group, respectively. Graphical representation of molecular structure of both compounds is shown in Figure . The tetravalent phosphorus atom effectively means each molecule of CPS 15 has two dissociated chloride counteranions: one for each of the cationic phosphonium and the azolinium constituent parts. Interestingly, the asymmetric unit of the crystal lattice of 15 also contained a molecule of hydrochloride (Figure S2).c The additional chloride counterion has important implications for the deprotection of 15, which thus requires 1.5 equiv of hexachlorodisilane to fully convert the CPS to P(III)5: presumably, the extra Cl– counterion of the imidazolium moiety also reacts with Si2Cl6 (vide infra). Finally, mesityl-substituted 5 could be facilely synthesized in an excellent 94% yield, without implementing harsh reaction conditions. In addition, we further demonstrated the usefulness of the new procedure at generating phosphine-bearing azolium salts with the synthesis of the 2,6-diisopropylphenyl analogue 14, in a comparable 92% yield. More details concerning the crystal structure of CPS 15 and azolium 5 can be found in the Supporting Information (Figures S58–S66).

Figure 1

Graphical representation of molecular structure, where (a) CPS 15 and (B) azolium 5. Displacement ellipsoids are drawn at the 50% probability level. The H atoms, the HCl molecule (CPS 15), and the ionic pair “B” (azolium 5) were omitted for clarity.

The class="Chemical">azolium salts were reacted with 5.0 equiv class="Chemical">pan class="Chemical">(COCl)2. The resultant CPS was separated from TsCl and then reacted with 1.5–1.6 equiv of Si2Cl6.

Graphical representation of molecular structure, where (a) class="Chemical">CPS 15 and (B) class="Chemical">pan class="Chemical">azolium 5. Displacement ellipsoids are drawn at the 50% probability level. The H atoms, the HCl molecule (CPS 15), and the ionic pair “B” (azolium 5) were omitted for clarity.

Experimental and Computational Mechanism Studies

class="Chemical">CPSs in class="Chemical">pan class="Chemical">methylene chloride form a cationic phosphonium with a noncoordinated anionic chloride counteranion,[89−93] while it has been demonstrated that Cl– (e.g., from ammonium salts) leads to scission of the Si–Si bond in Si2Cl6 (Scheme ).[94−99] This lead us to surmise that the reaction is initiated by the attack of chloride anion at silicon of Si2Cl6 generating an equivalent tetrachlorosilane (SiCl4) and a reactive transient trichlorosilanide anion [:SiCl3]− which then abstracts the remaining phosphorus bound chloride from intermediated 17 to generate the second and final equivalent of SiCl4.

Scheme 3

Reaction Mechanism of Si2Cl6 with Dissociated Chloride Anions

Known formation of anion [:SiCl3]– from Si2Cl6.

Stepwise reaction mechanism (bottom left).

Concerted mechanism (bottom right).

Known formation of anion [:pan class="Chemical">SiCl3]– from class="Chemical">pan class="Chemical">Si2Cl6.

To explore this mechanistic proposal, chlorotriphenylclass="Chemical">phosphonium triflate (class="Chemical">pan class="Chemical">Ph3PClOTf) 2b was synthesized.[100] The triflate anion is a superb nucleofuge, being a far more stable leaving group than chloride;[101] therefore, the dissociated triflate ion (TfO–) of 2b would be expected to react much slower with hexachlorodisilane than Cl– of 2a. Indeed, after reaction for 10 min, 31P NMR indicated 5b had generated only 7% Ph3P 6, progressing to 80% and 100% after 24 and 48 h, respectively (Table , entries 11–13; Figure ), significantly slower than the dichloride analogue 2a which appears to react instantly. As with 2a, 29Si NMR analysis of the reaction mixture of monotriflate 2b with Si2Cl6 showed the generated of SiCl4 (singlet at δ = −18.8 ppm) but in addition a singlet at δ = −38.2 ppm. 13C NMR spectra showed a quartet at δ = 118 ppm (J = 320 Hz) and 19F NMR a singlet at δ = −75.6 ppm; these signals are tentatively attributed to trichlorosilyl triflate, SiCl3OTf (see the Supporting Information). Finally, CPS 2c bearing the non-nucleophilic tetrakis(3,5-dichlorophenyl)borate anion, [BArCl]−, was mixed with Si2Cl6 in methylene chloride. As anticipated, no triphenylphosphine 3 was formed, even with an excess of Si2Cl6, demonstrating that the reaction is initiated by the attack of a dissociated anion at silicon.

Figure 2

31P NMR (CD2Cl2) of Ph3PClOTf (2b) + Si2Cl6 → PPh3 (3). Reaction times = 10 min (bottom), 24 h (center), and 48 h (top).

class="Chemical">31P NMR (class="Chemical">pan class="Chemical">CD2Cl2) of Ph3PClOTf (2b) + Si2Cl6 → PPh3 (3). Reaction times = 10 min (bottom), 24 h (center), and 48 h (top).

To gain further inclass="Chemical">sight, quantum-chemical calculations emclass="Chemical">ploying the TURBOMOLE class="Chemical">program were class="Chemical">performed to study the thermodynamics and kinetics of the reaction. By use of the harmonic oscillator and rigid rotator aclass="Chemical">pclass="Chemical">proximation with a reference class="Chemical">pressure of 1 bar, Gibbs free energies are given at the PBE0-D3/def2-TZVPP//PBE-D3/dhf-SV(P) level of theory.[102−109] Our calculations show that the disclass="Chemical">proclass="Chemical">portionation of class="Chemical">pan class="Chemical">CPS into free phosphine with liberation of chlorine is uphill in free energy by 94 kJ/mol; similarly, formation of (unstabilized):SiCl2 by disproportionation of Si2Cl6 is also expected to be very unfavorable, ΔG = 107 kJ/mol. However, the formation of the free phosphine with Si2Cl6 releasing two SiCl4 molecules is thermodynamically favorable, ΔG = −246 kJ/mol (Scheme b,c).

Activated with 3.0 equiv of pan class="Chemical">(COCl)2, declass="Chemical">protected with 2.1 equiv of class="Chemical">pan class="Chemical">Si2Cl6.

Activated with 6.0 equiv of pan class="Chemical">(COCl)2, declass="Chemical">protected with 4.1 equiv of class="Chemical">pan class="Chemical">Si2Cl6.

A Telescoped Synthesis of Metal Complexes from Their Corresponding Phosphine(V) Oxides

With the new method of generating class="Chemical">phosphine(III) ligands with high yield and class="Chemical">purity in hand, we attemclass="Chemical">pted to telescoclass="Chemical">pe[110] the class="Chemical">procedure for the syntheclass="Chemical">pan class="Chemical">sis of organometallic catalysts. As such, after deprotection and removal of SiCl4 by evaporation, “intermediate” phosphine(III) compounds were filtered through Celite and then reacted with a suitable metal precursor to yield a selection of prominent phosphine-bearing catalysts. The resultant monodentate triphenylphosphine, tricyclohexylphosphine, and CyJohnPhos were reacted with the dichloro(p-cymene)ruthenium(II) dimer, Umicore M31, and (η3-allyl)palladium(II) dichloride to afford the versatile dichloro(p-cymene)(triphenylphosphine)ruthenium(II) catalyst, 18,[111] olefin metathesis catalyst Umicore M2 (Grubbs catalyst M202), 19,[112] and the palladium Buchwald complex, CyJohnPhos(η3-allyl)PdCl, 20,[113] respectively, in excellent yields (91–98%). Moreover, the oxides of multidentate ligands where similarly reduced and successfully metalated, thus affording bidentate nickel 21(114) and tetradentate palladium complexes 22(115) in good yields of 83% and 86%, respectively.

Conclusions

We have developed a class="Chemical">simclass="Chemical">ple mild one-class="Chemical">pot activation/declass="Chemical">protection class="Chemical">procedure in which class="Chemical">pan class="Chemical">phosphine(V) oxides are converted to their corresponding phosphine(III) ligands cleanly and efficiently at ambient temperature without the use of metals or the need for silica gel chromatography. The reduction of activated CPS 2 was investigated with a range of disilanes, and Si2Cl6 was demonstrated to be the best reductant. A reaction mechanism for the transformation has been elucidated through experimentation and supported by computation calculations, with the reduction being initiated by attack of the CPS’s dissociated chloride anion at the silicon of hexachlorodisilane. The new method was successfully applied to a range of aryl and alkyl phosphines, including state-of-the-art ligands, and found to be compatible with alkene, ether, and amine function groups. Challenging phosphine-bearing azolium salts were readily furnished. Furthermore, the high purity of resultant phosphine(III) compounds allowed the procedure to be telescoped for the formation of some prominent transition metal catalysts. We believe this research will facilitate the synthesis of both known and novel new phosphine(III) ligands as well as their corresponding complexes, while the catalytic use, reuse, or recycling of valuable phosphine(III)-based reagents is of importance for sustainability and is likely to be of only greater significance as increased demands or restrictions are placed upon finite phosphorus resources.[82−84]

Table 2

Conversion of Phosphine(V) Oxides to Phosphine(III) Ligands via CPS Intermediates

The azolium salts were reacted with 5.0 equiv (COCl)2. The resultant CPS was separated from TsCl and then reacted with 1.5–1.6 equiv of Si2Cl6.

Table 3

Conversion of Phosphine(V) Oxides to Their Corresponding Phosphine(III) Ligands and Metal Complexes

Activated with 3.0 equiv of (COCl)2, deprotected with 2.1 equiv of Si2Cl6.

Activated with 6.0 equiv of (COCl)2, deprotected with 4.1 equiv of Si2Cl6.