Synthesis and structural characterization of isolable phosphine coinage metal π-complexes

Abstract

The chemical community has recently witnessed a dramatic increase in the application of cationic gold(I)-phosphine complexes as homogeneous catalysts for organic synthesis. The majority of gold(I)-catalyzed reactions rely on nucleophilic additions to carbon–carbon multiple bonds, which have been activated by coordination to a cationic gold(I) catalyst. However, structural evidence for coordination of cationic gold(I) complexes to alkynes has been limited. Here, we report the crystal structure of a gold(I)-phosphine η²-coordinated alkyne. Related Ag(I) and Cu(I) complexes have been synthesized for comparison. The crystallization of these complexes was enabled by tethering a labile alkyne ligand to a strongly coordinating triarylphosphine. This approach also proved applicable to crystallization of the first gold(I)-phosphine η²-coordinated alkene.

The development of new synthetically useful methodology often rests on an understanding of the mechanistic underpinnings of the desired transformation. For example, the isolation and characterization of Zeise's salt, K[PtCl₃(C₂H₄)], provided direct evidence for the ability of Pt(II) to remove electron density from ethylene, thereby rendering it susceptible to nucleophilic attack (1–4). This insight provided the impetus for the development of numerous platinum and palladium catalyzed reactions. In contrast, homogeneous Au(I) and Au(III) complexes have only recently emerged as highly competent and selective catalysts for the activation of π-bonds (5–8). In light of the recent synthetic advances concerning Au(I) catalysis, a thorough understanding of the mechanistic and structural basis for these reactions has lagged behind.^‡

The coordination of an alkyne to a cationic Au(I)-phosphine complex represents the prototypical mechanistic starting place for Au(I)-catalyzed reactions, despite the fact that little structural evidence exists for this assertion. In fact, to date there have been no reported crystal structures of linear-phosphine-Au-(η²-alkyne) complexes.^§ Here, we report the characterization of the 14-electron Au(I)-phosphine-(η²-alkyne) complex 1, as well as its silver(I) and copper(I) analogues (2 and 3), and the cationic phosphine Au(I)-alkene complex (4) for comparison (Fig. 1).^¶ With these structures in hand, we can begin to understand the unique ability of Au(I) complexes to serve as effective π-activation catalysts, especially in understanding why gold is often more effective than copper or silver.^‖

Fig. 1.

Schematic representations of the coinage metal alkyne complexes 1 (M = Au), 2 (M = Ag), and 3, together with a portion of Au-alkene coordination polymer 4.

Results and Discussion

Although Au(I)-π-complexes have traditionally been difficult to isolate and characterize, we hypothesized that these difficulties might be overcome by employing a tethered phosphine-alkyne ligand. To this end, reaction of alkynyl phosphine 5 with (dimethylsulfide)gold(I) chloride afforded the phosphinegold(I) chloride complex in 93% yield (Fig. 2). This complex was converted into cationic phosphinegold(I) complex 1 in 98% yield by abstraction of the chloride with silver hexafluoroantimonate. Crystals of complex 1 were obtained when a layered CH₂Cl₂/hexanes solution of 1 was allowed to stand at 0°C. The corresponding Ag(I) and Cu(I) complexes were obtained in quantitative yields by the reaction of ligand 5 directly with cationic metal precursors. The structures of the complexes 1–3 were firmly established by x-ray crystallographic analysis (Figs. 3 and 4). In all cases, solvent molecules and counterions are completely separated from the cationic metal centers. The Au(I) and Ag(I) complexes 1 and 2 are structurally analogous dimers, both displaying pseudo-linear geometry about the metal.^†† Although, the dimeric structure of 1 and 2 was unexpected, it is not surprising considering the structure of the ligand and the preferred linear Au(I)-coordination geometry. In contrast, the Cu(I) complex 3 is monomeric, with pseudo-trigonal planar geometry about copper.

Fig. 2.

Synthesis of coinage metal alkyne complexes 1–3, from alkynyl phosphine 5.

Fig. 3.

ORTEP drawings of the Au- and Ag-alkyne complexes 1 and 2, respectively, shown as 50% ellipsoids. Hydrogens, solvent, and counterions (SbF₆) omitted for clarity. For selected bond lengths and angles see Table 1.

Fig. 4.

ORTEP-drawing of the Cu-alkyne complex 3, shown as 50% ellipsoids. Hydrogens, solvent, and counterion (PF₆) are omitted for clarity. For selected bond lengths and angles, see Table 1.

Finally, a crystal structure of the first (η²-alkene)-Au(I)-phosphine complex 4 was obtained using an analogous alkene-tethered phosphine ligand (15–19) (Fig. 5. In agreement with previous calculations, the Au(I)-alkene bond is longer than the Au(I)-alkyne bond, despite the fact that the former is calculated to be stronger (25, 26).

Fig. 5.

ORTEP drawing of the Au-alkene complex 4, shown as 50% ellipsoids. Hydrogens, solvent, and counterions (SbF₆) are omitted for clarity. Selected bond lengths (Å): P(1)-Au(1), 2.272(3); Au(1)-C(1), 2.250(10); and Au(1)-C(2), 2.34(1).

A comparison of the some relevant structural features of 1–3 is given in Table 1). The Cu(I)-bound alkyne has the smallest C(1)-C(2)-C(3) and C(2)-C(1)-Si(1) angles [159.8(3)° and 159.5(2)°] and the lowest stretching frequency (2,020 cm⁻¹). In contrast, the Au(I)- and Ag(I)-bound alkynes exhibit significantly less distortion from linearity [C(1)-C(2)-C(3) angles of 167.2(6)° and 172.9(3)°, respectively]. The observed deviations of the alkyne from linearity correspond well to a decrease in the bond's stretching frequency. It is also important to note that in both the Au(I) and Ag(I) complexes (1 and 2), the metal center is significantly “slipped” to one side of the π-bond. Previous research suggests that such η² → η¹ migration should accompany an increase in ligand electrophilicity (27, 28). In contrast, the Cu-alkyne complex is highly symmetrical [2.029(2) vs. 2.024(2) Å], although this may be a result of the altered geometry and electron count of this complex.

Table 1.

Selected bond lengths (Å), bond angles (°), and IR stretching frequencies (cm⁻¹)—experimental and calculated

Entry†	Parameter	1 (M = Au)‡	2 (M = Ag)	3 (M = Cu)
1a	d C(1)-C(2)	1.221(8)	1.211(4)	1.229(4)
1b		1.258	1.252	1.265
2a	∠C(1)-C(2)-C(3)	167.2(6)	172.9(3)	159.8(3)
2b		171.0	173.6	158.9
3a	∠Si(1)-C(1)-C(2)	164.4(5)	165.3(3)	159.5(2)
3b		160.0	160.8	161.3
4a	νC(1)≡C(2)§	2053	2095	2020
4b		2068	2100	2014
5a	d M(1)-C(1)	2.197(5)	2.294(3)	2.029(2)
5b		2.249	2.292	2.077
6a	d M(1)-C(2)	2.270(5)	2.445(3)	2.024(2)
6b		2.386	2.493	2.070
7a	d P(1)-M(1)	2.271(1)	2.3934(8)	2.2571(7)
7b		2.337	2.431	2.338

^†na, experimental value; nb, calculated value.

^‡Data from one half of the unsymmetrical Au-dimer is shown. For the other half, see SI Appendix.

^§The ν_C(1)≡C(2) of the free ligand is 2,171 cm⁻¹.

Experimentally, we have observed that Au(I) complexes are generally superior π-activation catalysts, especially when compared with analogous Cu(I) and Ag(I) complexes. Intuitively, the superior catalytic activity of Au(I) in these reactions suggests that Au(I)-coordinated π-bonds are more activated toward nucleophilic attack than π-bonds that are coordinated to Ag(I) or Cu(I). Under the Dewar–Chatt–Duncanson bonding model, this further suggests increased π-to-metal σ-donation to gold and/or decreased metal-to-π* back-donation from gold (29, 30). Therefore, we sought to compare the relative importance of these two types of bonding for complexes 1–3. The situation is complicated by the fact that both π-to-metal σ-donation and metal-to-π* back-bonding elongate and distort a coordinated π-bond. As a result, one cannot directly correlate the degree of alkyne distortion with the degree of metal-to-π* back-donation. To circumvent these problems, we turned to DFT calculations (17, 31–35).

Beginning with the crystal structure of Au(I) complex 1, we initially simplified the structure to monomeric triphenylphosphine-metal-alkyne complex 1b (Fig. 6). The experimental geometric features were well produced with the B3PW91/LANL2DZ(Au), LANL2DZdp(Si,P), ccpVDZ(C,H) level of theory (Table 1) [for further details on the synthesis and characterization of complexes 1–4, see supporting information (SI) Appendix]. Importantly, this indicates that the geometry of the metal center in 1 is not significantly influenced by the constraints of crystallization. The structures of 2 and 3 were trimmed and optimized similarly; the experimental and calculated values are compared in Table 1. The same structural trends were observed in both the calculated and experimental geometries. With these optimized geometries in hand, we turned to natural bond order calculations to investigate the nature of the metal-alkyne bond (Table 2) (for similar studies, see refs. 25 and 31–34). Second-order perturbative analysis revealed that π-to-metal σ-donation is of the largest magnitude for Au (56.6 kcal/mol), as is metal-to-π* back-donation (13.3 kcal/mol). For all three metals, σ-donation to the metal dominates, augmenting the electrophilicity of the alkyne, although the difference of the two bonding interactions is the largest for Au. These calculations also provide further insight into the observed distortion of the Cu(I)-bound alkyne. Noting that the calculated Cu-alkyne orbital interaction energies underestimate the degree of distortion suggests that some of the distortion of the Cu-bound alkyne may be due to geometrical constraints of that system.

Fig. 6.

Structures of 1b and 2b along with the calculated lowest unoccupied molecular orbitals (LUMOs) and corresponding energies.

Table 2.

Natural bond order orbital interaction energies

Parameter	1b (M = Au)	2b (M = Ag)	3b (M = Cu)
π → M	56.6	38.5	46.9
M → π *	13.3	6.4	12.0
Difference	43.3	32.1	34.9

In addition to NBO analysis, we also examined the frontier molecular orbitals of 1b and 2b (Fig. 6).^‡‡ As expected, the lowest unoccupied molecular orbitals (LUMOs) contain the π*-metal interaction. Furthermore, the LUMO of 1b was calculated to be 4 kcal/mol lower in energy than the LUMO of 2b. Although the energies of unoccupied orbitals must be considered with caution, this result is again qualitatively in agreement with Au(I) being a superior π-activation catalyst.

In agreement with the above claims, we have also found that complexes 1 and 4 are active in typical gold(I)-catalyzed reactions, whereas 2 and 3 are not (Fig. 7) (36–38). It is worth mentioning that phosphine-Au(I)-chloride complexes (16-electron complexes upon alkyne coordination) are typically much less catalytically active than their cationic congeners. Thus, Ph₃PAuCl is typically activated by addition of a silver salt (such as AgSbF₆), which serves to remove the chloride ligand and replace it with a less coordinating counteranion. Complexes 1 and 2 represent 14-electron Au(I) and Ag(I) complexes, whereas similar crystallization conditions result in 16-electron Cu(I) complex 3. Furthermore, the observation that the alkene complex 4 is also catalytically active supports the assumption that coordination of cationic gold(I) complexes to π-bonds is reversible and that the observed selectivity for alkynes in many gold-catalyzed reactions does not arise as a result of preferred bonding to the alkyne.

Fig. 7.

Reactivity of coinage metal complexes 1–4 in typical gold(I)-catalyzed reactions.

In summary, we have isolated and characterized, by x-ray crystallography, linear (η²-alkyne)- and (η²-alkene)-Au(I)-phosphine complexes. The isolation of these complexes was enabled by covalently linking kinetically labile alkenes and alkynes to triphenylphosphine.

Experimental Procedures

1.

To a solution of the corresponding phosphine-AuCl complex. (37 mg, 0.053 mmol) in d₃-acetonitrile (0.4 ml) was added AgSbF₆ (18 mg, 0.053 mmol). Silver chloride precipitated immediately, providing a solution of the corresponding cationic phosphine-Au-acetonitrile complex. ¹H NMR (400 MHz, d₃-MeCN) δ 7.64–7.51 (m, 12H), 7.27–7.22 (m, 1H), 6.80 (dd, 1H, J = 13.4, 7.8 Hz), 3.12 (t, 2H, J = 7.3 Hz), 2.51 (t, 2H, J = 7.3 Hz), 1.08–0.92 (m, 21H). ¹³C NMR (100 MHz, d₃-MeCN) δ 144.5 (d, J = 12 Hz), 135.3 (d, J = 14 Hz), 134.2 (d, J = 9 Hz), 133.5 (d, J = 2 Hz), 133.2, 131.8 (d, J = 9 Hz), 130.6 (d, J = 12 Hz), 128.4 (d, J = 67 Hz), 128.1 (d, J = 10 Hz), 126.8 (d, J = 62 Hz), 108.9, 82.5, 34.3 (d, J = 12 Hz), 21.1, 19.0, 11.9. ³¹P NMR (162 MHz, d₃-MeCN) δ 24.44.

This solution was then filtered through celite and concentrated in vacuo. The resulting solid (49 mg, 98% yield) was dissolved in dichloromethane and again filtered through celite. X-ray quality crystals were grown by layering the resulting solution with hexanes. IR: 2,053, 1,472, 1,439, 1,102, 883 cm⁻¹. Attempted NMR analysis of the crystals was hampered by broad signals and instability of the sample to disproportionation. ¹H NMR (400 MHz, CDCl₃) δ 7.64–7.50 (m, 12H), 7.32 (m, 1H), 6.89 (m, 1H), 3.4–3.1 (m, 4H), 1.20–0.90 (m, 21H). ³¹P NMR (162 MHz, CDCl₃) δ 29.3. HRMS (FAB) calculated for [C₃₁H₃₉PSiAu]⁺ 667.2224, found 667.2218.

X-Ray Data.

C_31.5H₄₀F₆SiPClSbAu, M_r = 945.88, triclinic, space group P1 (#2); a = 13.963(2), b = 14.920(2), c = 18.703(2) Å, α = 69.422(1), β = 77.264(2), γ = 75.161(2)°, V = 3489.2(7) ų, Z = 4; ρ_calcd = 1.80 g/cm³, μ (Mo-Kα) = 51.97 cm⁻¹, Mo-Kα radiation (λ = 0.71069 Å), 153 K, 2θ_max = 52.0°, 9,559 unique reflections of 31,450 measured, R_int = 0.028, R = 0.032, Rw = 0.035. Crystal dimensions: 0.08 × 0.08 × 0.17 mm.