Isolation of Au-, Co-η1PCO and Cu-η2PCO complexes, conversion of an Ir–η1PCO complex into a dimetalladiphosphene, and an interaction-free PCO anion

Sodium phosphaethynolate reacts with [MCl(PDI)] (M = Co, Ir; PDI = pyridinediimine) to give metallaphosphaketenes, which in the case of iridium rearranges into a dimetalladiphosphene.


Introduction
Despite their early synthesis in 1992 by Becker, Westerhausen et al., 1 the reactivity of phosphaethynolate salts M + (PCO) À remained unexplored until the recent development of simple syntheses which allow for the preparation of large quantities of pure material. 2 These salts have been shown to undergo a variety of chemical transformations with organic substrates, 3 including cycloaddition reactions leading to phosphorus containing heterocycles. 4 In contrast, the strong reducing ability of PCO À salts 5 has hindered their exploration in transition metal chemistry. In most cases the reactions with metal complexes lead to decomposition products. 6 The only exceptions are the formation of a P 2 (C]O) 2 ring A by reacting Li(OCP) with (h 5 -C 5 R 5 )(CO) 2 FeBr, 7 the isolation of [Re(PCO)(CO) 2 (triphos)] B, 6 and uranium and thorium M(OCP)(amid) 3 complexes C, 8 in which the phosphaethynolate binds the metal via the oxygen center (Fig. 1). Noteworthy is an extensive computational study, showing that metallaphosphaketene complexes such as D may rearrange via CO migration into molecular metal phosphides E. 9 The limited number of studies concerning the coordination chemistry of PCO À is in marked contrast with the large number of reports dealing with the lighter analogue, namely the cyanate anion (NCO À ). 10 Herein we describe salt metathesis reactions leading to both unstable and stable terminal PCO transition metal complexes, featuring different coordination modes, and reactivity. We also report the experimental demonstration of the predicted conversion of D to E, and the displacement of the PCO unit of the copper complex by a CAAC, which leads to a PCO anion with no coordinating solvents or binding agents.

Results and discussion
Because of the strongly reducing character of Na(OCP), 6,7 we targeted complexes bearing pyridinediimine (PDI) ligands 11 and cyclic (alkyl)(amino)carbenes (CAACs) 12,13 which are known to efficiently stabilize metals in low oxidation states. 14 When the [CoCl(PDI iPr )] complex 1 was reacted with Na(OCP) in THF at À30 C the color changed from pink to deep purple, and a single broad resonance in the 31 P NMR spectrum at d ¼ À226 ppm [vs. d ¼ À392 ppm for Na(OCP)] indicated quantitative conversion. The metallaphosphaketene 3 was isolated in 61% yield and fully characterized. The IR spectrum showed the asymmetric stretching frequency of the phosphaketene unit at n asym ¼ 1851 cm À1 , intermediate between Na(OCP) (n asym ¼ 1755 cm À1 ) and Ph 3 Sn-P]C]O (n asym ¼ 1946 cm À1 ) 2d indicating a cobalt phosphaketene structure, Co-P]C]O. This is conrmed by a single crystal X-ray diffraction analysis (Fig. 2). The P-C [1.633(4)Å] and C-O [1.179(6)Å] bond distances, the rather large Co-P-C angle [116.2(1) ] and long Co-C distance [3.325(4)Å] indicate a h 1 -coordination via the phosphorus atom of the OCP À anion. 6 The structural parameters conrm that neither the cobalt center nor the PDI ligand are reduced by OCP À .
The corresponding iridium complex [IrCl(PDI Me )] 2 with the less sterically encumbered PDI Me ligand reacts with Na(OCP) at low temperatures to cleanly give product 4. A 31 P NMR resonance at d ¼ À316.7 ppm indicates a metallaphosphaketene (Ir-P]C]O) featuring a highly covalent metal phosphorus bond. Complex 4 could not be isolated. Keeping a THF solution at 20 C for about 6 h leads cleanly to complex 5 (d 31 P ¼ +682 ppm; l max ¼ 524.1 nm, 732.5 nm), which was isolated as red crystals. A single crystal X-ray structure analysis shows compound 5 to be a dimetalladiphosphene (Fig. 2). 15 The P-P distance [2.021(1)Å] is short and in the typical range of diphosphenes, R-P]P-R. The iridium centers are bound to a redox-inactive PDI ligand with short C]N bonds [N1-C2 1.340(2)Å; N3-C8 1.335(2)Å] and a carbonyl ligand. The rearrangement of 4 into 5 is the experimental conrmation of the computationally predicted transformation of D to E. 9 This transformation is also comparable to the conversion of an iridium azido complex, Ir-N 3 , to a transient terminal nitrido complex, Ir^N, which could be spectroscopically characterized but also dimerizes to an Ir-N]N-Ir complex. 16 Monitoring by 31 P NMR spectroscopy the reaction of gold and copper complexes 6a,b and 7 with Na(OCP) in benzene showed in each case the formation of a single product giving a 31 P NMR signal (8a: d ¼ À360; 8b: À364; 9: À387 ppm) slightly up-eld shied compared to that of Na(OCP) (d ¼ À392 ppm). Single crystals of 8a and 9 were grown and subjected to X-ray diffraction studies (Fig. 3). Only very subtle structural differences between both complexes were observed. The P-C [8a: 1.640(3); 9: 1.636(2)Å] and C-O bond lengths [8a: 1.176(4); 9: 1.184(2)Å] are similar, and the M-P-C angle is slightly more acute for the copper complex 9 [8a: 86.2(1), 9: 79.15 (5) ].
Despite the similarities of the solid state structures, natural bond orbital (NBO) analysis at the M06/6-311++G(2d,p)+SDD// M06/6-31+G(d)+LANL2DZ(+f) level of theory shows signicant differences between the electronic structures of 8a and 9. The NBO charges of Au and PCO in 8a are +0.39 and À0.56 a.u., respectively, whereas those of Cu and PCO in 9 are +0.58 and À0.68 a.u., suggesting that the PCO anion in 9 is more ionic and "free" than that in 8a. This is in agreement with the different 31 P NMR chemical shis of 8a (d ¼ À360 ppm) and 9 (d ¼ À387 ppm). The NBOs corresponding to the M(PCO) (M ¼ Au or Cu) fragments are quite different as shown in Fig. 4. The phosphorus center of 8a forms three bonds (Au-P s, P-C s and P-C p) (Fig. 4a-c). In contrast, for 9 no Cu-P s bond could be located. Instead, the phosphorus center of 9 forms one P-C s  bond and two P-C p bonds (Fig. 4f-h). Moreover, there are two bonds between the C and O atoms (C-O s and C-O p) in 8a ( Fig. 4d and e), while only one C-O s bond in 9 (Fig. 4i). These computational results suggest that the coordination modes of PCO with gold and copper are h 1 and h 2 , respectively.
The different bonding modes in 8 and 9 lead to a difference in reactivity. In solution 9 decomposes aer a few hours giving a complex mixture, whereas 8b rearranges into complex 10 over the course of a week when le standing in THF at room temperature (Fig. 5). The trinuclear nature of 10 [(L b Au) 3 P], as determined by an X-ray diffraction study, is reminiscent of the rearrangement product of Ph 3 Sn(PCO), namely (Ph 3 Sn) 3 P. 2d, 17 Three gold atoms surround a single P atom, leading to a phosphine supported solely by metals. The 31 P NMR spectrum displays a signal at d ¼ À200 ppm, which is considerably high-eld shied compared to alkyl and aryl phosphines. The electron rich nature as well as the steric crowding around the P center could make 10 an interesting redox active ligand for transition metals. 18 The gold complexes 8a,b are rather inert and do not react with heavier group 14 element halides to give R 3 E-PCO derivatives. Equally, no reaction with N,N 0 -dicyclohexylcarbodiimide or with carbene L a are observed. On the contrary, the copper salt 9 does react with these reagents similarly to Na(OCP) (see the ESI † for details). 2b,d Remarkably, 9 reacts with carbene L a to afford the cationic bis(CAAC)Cu complex 11, in which the PCO fragment is the anionic counterpart. The 31 P NMR signal appears at d ¼ À400 ppm, which is more downeld shied than Na(OCP) (d ¼ À392 ppm), implying that PCO À is less coordinated. The IR spectrum showed the asymmetric stretching frequency of the PCO unit at n asym ¼ 1791 cm À1 , suggesting a more cumulenic nature than in the two crystalline forms of Na(OCP) (n asym ¼ 1780 or 1755 cm À1 ). 6,19a,b Lastly, although a disorder precludes accurate determination of the geometric parameters, the X-ray diffraction study revealed that the PCO anion has no close contacts with the cationic part of the complex (Fig. 6). This is the rst time that the PCO anion has been structurally observed without any binding agents or coordinating solvents.
On the basis of DFT calculations, the HOMOs of 8a and 9 are mainly localized on the PCO fragments (see ESI †) and the oxygen atoms carry the largest negative charge (À0.56 a.u. in 8a    and À0.59 a.u. in 9). Thus, we were curious to see if the terminal oxygen atom could react with a Lewis acidic borane. Indeed, adding one equivalent of B(C 6 F 5 ) 3 to either 8b or 9 led to the same type of heterocycle 12 and 13, respectively (Fig. 7). In the case of gold, the corresponding product immediately crystallized out and became insoluble in the tested solvents. However, for copper the product was soluble and the 31 P NMR spectrum showed two broad peaks at d ¼ 261 and d ¼ 137 ppm. Single crystal X-ray diffraction studies of 12 and 13 revealed a fourmembered P 2 C 2 heterocycle that arose from borane coordination to oxygen followed by a dimerization and lastly a migration of a LM fragment from one phosphorus center to the other. The four-membered P 2 C 2 heterocycles have a planar geometry and, as expected, the bond lengths of the PCO fragments become elongated compared to 8b and 9 as a result of the delocalization over the ring. This small ring represents a novel bonding mode for the rapidly growing eld of molecular polyphosphorus clusters and cages. 20

Conclusions
PCO complexes with electron rich metal centers such as copper, gold and also cobalt can be prepared and are relatively stable. On the other hand, the iridium phosphaketene 4 rapidly rearranges via CO migration to give a genuine dimetalladiphosphene. The copper complex features an h 2 coordination mode, which leads to an active PCO fragment that can undergo further reactions. A free PCO anion, resulting from simple displacement of the PCO unit with a carbene, was also isolated. For both the copper and gold complexes, borane coordination to the oxygen of the OCP unit induces a [2 + 2] cycloaddition into a P 2 C 2 heterocycle. These results demonstrate that the uncharted chemistry of transition metal PCO complexes is rich and the formation of new metal phosphides and the mechanisms leading to them merit further exploration.

General considerations
All air-and moisture-sensitive manipulations were carried out using standard vacuum line Schlenk techniques or an MBraun dry-box under argon. THF was distilled over sodium benzophenone-ketyl before use. THF-d 8 , CD 2 Cl 2 , and C 6 D 6 were purchased from Cambridge Isotope Laboratories and dried over 4Å molecular sieves. ( iPr PDI)CoCl 1, 21 ( Me PDI)IrCl 2, 11d L a,b AuCl 6a,b 22 and L a CuOtBu 7 23 were synthesized according to literature procedures. 1 H, 13 C, 11 B, 19 F, and 31 P NMR spectra were recorded on a Varian VX 500, Bruker 300, Bruker 500 and Jeol 500 spectrometer at 25 C. All 1 H and 13 C NMR chemical shis are reported relative to SiMe 4 using the 1 H (residual) and 13 C chemical shis of the solvent as a secondary standard. NMR multiplicities are abbreviated as follows: Chemical shis are given in ppm and coupling constants J are given in Hz. Peak widths at half heights (in Hz) are given for broad signals. Infrared spectra were collected on a Perkin-Elmer-Spectrum 2000 FT-IR-Raman and Bruker ALPHA FT-IR spectrometer. Elemental analyses were performed at the Mikrolabor of ETH Zürich. Single crystals suitable for X-ray diffraction were coated with polyisobutylene oil in a dry-box, transferred to a nylon loop and then transferred to the goniometer of a Bruker X8 APEX2 diffractometer equipped with a molybdenum X-ray tube (l ¼ 0.71073Å) or on a Bruker Apex II-CCD detector using Mo-Ka radiation (l ¼ 0.71073Å) or Cu-Ka radiation (l ¼ 1.54178Å). The data were processed using the Bruker SAINT+ program and corrected for absorption using SADABS. The structures were solved using direct methods (SHELXS) completed by Fourier synthesis and rened by fullmatrix least-squares procedures. Mass spectra were performed at the UC San Diego Mass Spectrometry Laboratory. Melting points were measured with an electrothermal MEL-TEMP apparatus.
Preparation of ( iPr PDI)Co(PCO) 3. In the glove box, a 20 mL scintillation vial was charged with 0.200 g (0.347 mmol) of ( iPr PDI)CoCl 1 and 10 mL of THF. The solution was cooled to À35 C and Na(OCP) (0.130 g, 0.355 mmol) was added portionwise over the course of 5 minutes, eliciting a color change from pink to dark purple. The reaction was placed in the freezer at À35 C for one hour then ltered through Celite. The solution was concentrated, layered with hexane and placed at À35 C. This gave 0.152 g (48%) of a purple crystalline solid identied as [( iPr PDI)Co(PCO)] 3. The mother liquor was placed back in the freezer to obtain another 42 mg (13%) of product. X-Ray quality crystals were grown from the second fraction. Analysis for C 34 -H 43 CoN 3  Preparation of [( Me PDI)IrCO] 2 (m-P 2 ) 5. A 20 mL Schlenk ask was charged with 0.100 g (0.167 mmol) of (MePDI)IrCl 2 and 5 mL of THF and cooled in a dry-ice/acetone bath. A solution of Na(OCP) (0.060 g, 0.170 mmol) in THF (3 mL) was syringed into the stirring iridium solution, immediately causing a color change to dark purple. The reaction was warmed to room temperature, whereupon the color changed to deep pink, and stirred for an additional hour. The reaction was then ltered through Celite and then concentrated to roughly 3 mL. Storing at À35 C overnight produced a solid that was collected on a glass frit and dried under reduced pressure, yielding 0.076 g (74% yield) of red crystalline solid 5. X-Ray quality crystals were grown from the slow evaporation of the mother liquor at room temperature overnight. NMR analysis was performed in CD 2 Cl 2 due to the poor solubility of 5 in ethereal or aromatic solvents, but the compound slowly decomposed (if le overnight) in methylene chloride. Analysis for C 52 H 55 Ir 2 N 6  Preparation of (CAAC)Au(PCO) complex 8a. A mixture of (CAAC)AuCl 6a (1.0 g, 1.83 mmol) and [Na(PCO)(dioxane) 2.5 ] (0.55 g, 1.83 mmol) was cooled to À78 C before THF (10 mL) was added. The mixture was stirred for 15 minutes and then warmed to room temperature. Aer 30 min, the solvent was removed under vacuum and the resulting brown solid was extracted with 15 mL of benzene. Aer removing the solvent, 8a was obtained as a light yellow solid (0.73 g, yield: 70%). Colorless single crystals of 8a were obtained by vapor diffusion of pentane into a saturated benzene solution of 8a in the dark. IR (C 6 H 6 ): n PCO ¼ 1887 cm À1 . M.P. ¼ 193 C (dec.  5. HRMS was attempted but a peak corresponding to M + could not be located, probably due to the weak P metal bond.