Gold(I) and Silver(I) π-Complexes with Unsaturated Hydrocarbons.

Gold π-complexes have been studied largely in the past 2 decades because of their role in gold-catalyzed reactions. We report an experimental and theoretical investigation of the interaction between a wide range of unsaturated hydrocarbons (alkanes, alkynes, alkadienes, and allenes) and triphenylphosphine-gold(I), triphenylphosphine-silver(I), and acetonitrile-silver(I) cations. The bond dissociation energies of these complexes were determined by mass spectrometry collision-induced dissociations and their structures were studied by density functional theory calculations and infrared photodissociation spectroscopy. The results show that with the same phosphine ligand, gold binds stronger to the π-ligands than silver and thereby activates the unsaturated bond more effectively. Ligand exchange of phosphine by acetonitrile at the silver complexes increases the binding energy as well as the activation of the π-ligands. We also show that the substitution of an unsaturated bond is more important than the bond type.

Introduction

Gold catalysis is one of the important sub-topics of current homogenclass="Chemical">pan>eous catalysis.[1−15] Its main advantage is a mild and chemoselective activation of C–C multiple bonds for a nucleophile attack. Cationic two-coordinate π-complexes of gold with unsaturated hydrocarbons are considered as initial reaction intermediates in almost every reaction catalyzed by cationic gold(I) (Scheme ) and therefore they attract substantial attention.[16,17] Brooner and Widenhoefer highlighted the importance of those complexes in gold(I) catalysis with a focus on their structure, bonding, and ligand exchange behavior.[18] Hashmi and Jones analyzed the same topic in their subsequent reviews.[19,20]

Scheme 1

Proposed Role of π-Complexes in Gold Catalysis

The cationic n class="Chemical">gold(I) catalysts are usually prepared in situ from their n class="Chemical">gold chloride pren class="Chemical">cursors by an anion exchange with silver salts. Silver salts are typically added in excess; thus, possible (co)catalytic effects cannot be excluded. We have shown that excess silver cations did not affect the kinetics of a simple gold-catalyzed nucleophilic addition to alkynes.[21] However, an example of a “silver effect” was presented by Shi and co-workers in 2012.[22] They showed that the simultaneous presence of both gold and silver species was crucial for several reactions to proceed. Since this contribution, the discussion on the silver effect has increased significantly.[23−29]

Our research group has already used the combination of mass spectrometry, ion spectrometry, and theoretical calculations for determining of possible n class="Chemical">silver effects in gold catalysis.[30,31] The cationic nature of the reactive gold(I) and silver(I) complexes with π-ligands allows their detection by electrospray ionization (ESI) mass spectrometry. Once isolated in the gas phase, their thermodynamic properties can be assessed, their bimolecular reactions can be studied, and also their optical absorption spectra can be measured.[32−35] This work continues in the investigation of cationic π-complexes of silver and gold complexes with ancillary ligands with a wide range of unsaturated hydrocarbons. We aim at a systematic comparison of binding energies and modes of activation of hydrocarbons in gold and silver π-complexes. We use triphenylphosphine as the main ancillary ligand because it is among the most applied ligands in gold catalysis.[36] For a long time, the triphenylphosphine–gold π-complexes were considered unstable in solution.[37,38] However, we have recently shown that these complexes are much more stable than what was previously published and we were also able to characterize them by X-ray for the first time.[39] Other silver(I) and gold(I) π-complexes utilizing phosphines and other ligands were characterized previously as well.[40−48]

This work aims at the comparison of n class="Chemical">silver(I) and n class="Chemical">gold(I) π-complexes with the same supporting ligand to clearly distinguish the direct efn class="Chemical">fect of the noble metals. In addition, a change of the supporting ligand on silver simulates the more realistic situation in solution. We focus our attention on the strength of the coordination bond, the lengthening of the π-bond, and the mode of activation.

Experimental and Computational Details

The experiments were performed with a Paul-type ion-trap mass spectrometer n class="Chemical">LCQ Deca (ThermoQuest). Ions were generated by ESI from dichloromethane solutions of unsaturated hydrocarbon and either AgSbF6 with an additional ligand (PPh3, CH3CN) or Au(PPh3)Cl activated by AgSbF6 (see the Supporting Information for details). Conditions for ESI: the nebulizer gas was nitrogen, the spraying voltage was 5–6 kV, the heated capillary was kept at 150–200 °C, and the flow rate of the solution was set to 0.3 mL·h–1. Mass-selection was done with a uniform isolation width of 1.6 amu. Collision activation in the collision-induced dissociation (CID) experiments was achieved through radiofrequency excitation (30 ms) followed by collisions with the helium buffer gas (the pressure of helium within the ion trap was ∼10–3 mbar). The trapping parameter qz was 0.25. The “normalized” collision energy scale used in LCQ ion traps can be calibrated and transformed to the center-of-mass collision energies.[49] Therefore, the appearance energies of the fragments in collision-energy resolved experiments correspond to the bond dissociation energies (BDEs).[50−52] All experiments were measured six times on at least two different days in order to eliminate possible systematic errors.

All caln class="Chemical">culations were performed with a Gaussian 09 package.[53] Structures were optimized by density functional theory method mPW1PW91.[54] The basis set cc-pVTZ was used for H, C, n class="Chemical">N, O, P, and Cl atoms.[55] Pseudopotential LanL2DZ for Ag and Au was used to account for relativistic effects.[56] The combination of both is denoted as cc-pVTZ/LanL2DZ in the following. Final binding energies were corrected for the basis–set superposition error (BSSE).[57] All reported structures were confirmed by the analysis of the corresponding Hessian matrices to be genuine minima on the respective potential-energy surfaces. The calculated energies in the gas phase include zero-point energy (ZPE) corrections and refer to 0 K. The natural atomic charges were calculated by natural population analysis.[58] We have also performed some of the calculations at the M06 level of theory for comparison,[59] and the values can be found in the Supporting Information.

The infrared photodissociation spectra were measured with the ISORI instrument.[60] The instn class="Chemical">rument has the same ESI source as the LCQ instrument and the ionization conditions were analogous as mentioned above. The ionic complexes were mass-selected by the quadrupole mass filter and trapped in a cryo-cooled wire quadrupole ion trap (operated at 3 K, helium buffer gas). The ions cooled down in collisions with helium and ultimately formed loosely bound helium-tagged complexes. Afterward, the trapped ions were irradiated by photon pulses from an optical parametric oscillator/amplifier (OPO, 10 Hz). Finally, the ions were extracted from the trap, mass-analyzed by the second quadrupole mass filter, and the number of helium-tagged complexes (N) was determined by a dynode/multiplier detector operated in ion-counting mode. In the following cycle, the light from the OPO was blocked by a mechanical shutter, giving the number of unirradiated ions (N0). The infrared photodissociation spectroscopy (IRPD) spectra are constructed as the wavenumber dependence of (1 – N/N0).

Results and Discussion

The aim of this paper is to compare binding energies of silver and gold cations with various pan> class="Chemical">unsaturated hydrocarbons in π-complexes. Complexes of basic alkenes and alkynes such as ethylene and acetylene were addressed previously.[61,62] We have already studied π-complexes of [Au(PMe3)]+ with unsaturated hydrocarbons.[34] We have shown that both the type of the multiple bond and its substitution have a large effect on the binding energies. Due to this fact, we decided to study an extended library of π-ligands as well as to include representatives of allenes, which are commonly applied in both gold and silver catalysis.

Experimental BDEs

First, we started with the investigation of the BDEs of the silver complexes. The ESI of a dichloromethane solution of AgSbF6, PPh3, and a hydrocarbon led to a mixture of silver complexes in the gas phase (Figure a). In every case, the most abundant ions were [Ag(PPh3)(H2O)]+, whereas the targeted complexes [Ag(PPh3)(π-ligand)]+ were detected with only a low intensity. Subsequently we mass-selected the 107Ag isotopomers and performed the energy-resolved CID experiments. We have observed a single fragmentation channel—the loss of the hydrocarbon, followed by the immediate partial association of the [Ag(PPh3)]+ fragment with background water molecules within the ion trap (Figure b). Calibration of the collision energy (see S1) and the evaluation of the energy onset of the fragmentation as depicted in Figure c provided the experimental binding energies of the [Ag(PPh3)]+ cation with various π-ligands. Table summarizes these results.

Figure 1

Determination of the experimental BDE of 1-pentene in the π-complex [Ag(PPh3)(1-pentene)]+. (a) ESI-MS spectrum of dichloromethane solution of AgSbF6, PPh3, and 1-pentene. (b) CID spectrum of the mass-selected cationic π-complex [107Ag(PPh3)(1-pentene)]+ showing also the subsequent association of [107Ag(PPh3)]+ with the background water molecules. (c) Energy-resolved CID spectrum and the extrapolation of the fragmentation onset to determine the BDE.

Table 1

Experimental and Theoretical Binding Energies of [M(PPh3)]+ with Unsaturated Hydrocarbons in the [M(PPh3)(Hydrocarbon)]+ π-Complexes in the Gas Phase (M = Ag/Au)[63]

Calculations were performed at the mPW1PW91/cc-pVTZ/LanL2DZ level of theory and include ZPE and BSSE corrections.

Caln class="Chemical">culations were performed at the mPW1PW91/cc-pVTZ/Lanpan>L2DZ level of theory anpan>d inpan>clude ZPE anpan>d BSSE corrections.

The determined n class="Chemical">BDEs range from 1.35 to 1.64 eV. The smallest binding energies correspond to the binding of 1-pentene, benzene, and 1,1-dimethylallene. The largest binding energies were found for the complex of 1,5-cyclooctadiene (COD) (1.64 eV), followed by the complex of tetramethylallene (1.52 eV) and 2-pentyne (1.50 eV). In general, the binding energies of aliphatic n class="Chemical">mono-alkenes to [Ag(PPh3)]+ clustered around 1.40 eV (with the exception of 1-pentene, see Table , entries 1–5 and 9). In comparison, the binding energy of cyclooctene is larger (1.48 eV), which is probably due to a higher strain in the eight-membered ring.[34] Cyclodienes with six- and eight-membered rings (except COD) have basically the same binding energies (1.44–1.45 eV); therefore, the binding does not depend on the ring size and conjugation. The significantly different value of 1,5-cyclooctadine showed that there is an additional effect, which we discuss below in the paragraph concerning theoretical calculations. The binding energy of an alkene compared to that of an alkyne with the same substitution is about 0.1 eV smaller (from the comparison of 1-pentene with 1-pentyne and cis/trans-pentene with 2-pentyne).

For a direct comparison, we decided to investigate the corresponding gold π-complexes. We kept the experimental conditions as similar as possible.[64,65] Both the source and the subsequent MS/MS spectra exhibited comparable trends as in the case of the experiments with n class="Chemical">silver (see the Supporting Information for details). The results are also listed in Table .

The determined interaction binding energies between the π-ligands and [Au(PPh3)]+ were generally 0.2–0.3 eV larger than the binding energies with [Ag(PPh3)]+ and 0.2 eV lower than those with [Au(PMe3)]+, which were determined previously.[34] The PMe3 ligand is more electron-donating than PPh3; therefore, [Au(PMe3)]+ should be less electrophilic than [Au(PPh3)]+.[66] The fact that [Au(PMe3)]+ binds stronger with π-ligands than [Au(PPh3)]+ points to the importance of π-backbonding. We note that the same trend of the effect of triphenylphosphine and trimethylphosphine was found for gold carbonyl complexes.[67,68] The binding energies in [Au(PPh3)(π-ligand)]+ ranged from 1.46 eV for benzene to 1.79 eV for the complex with tetramethylallene. The gold complexes do not show as significant dependence on the substitution and the type of the multiple bond as the silver complexes as most of them cluster around 1.65 eV. However, there are some exceptions: the internal alkyne (2-pentyne) binds with an energy of 1.77 eV. The binding energy of [Au(PPh3)]+ with cyclooctene was higher than that with aliphatic alkenes, similar to that found for the silver analogue. In addition, this value is as high as the binding energy of COD.

The direct comparison of the experimental values for the gold and the silver complexes is shownpan> in Figure a. There is seemingly a linear correlation between the binding energies of π-ligands in the gold- and in the silver complexes. However, there are two visible outliers—the complexes with benzene and 1,5-cyclooctadine—to which the silver cation binds much stronger compared to the other unsaturated hydrocarbons. This is discussed with the aid of theoretical calculations below.

Figure 2

(a,b) Correlation of the (a) experimental and (b) calculated binding energies of the silver and gold triphenylphosphine complexes with π-ligands with two visible outliers marked (benzene and COD). (c,d) Correlation of the experimental and calculated binding energies of the (c) silver and (d) gold triphenylphosphine complexes with π-ligands (benzene as an outlier marked).

(a,b) Correlation of the (a) experimental and (b) calculated binding energies of the n class="Chemical">silver and gold n class="Chemical">triphenylphosphine complexes with π-ligands with two visible outliers marked (benzene and COD). (c,d) Correlation of the experimental and calculated binding energies of the (c) silver and (d) gold triphenylphosphine complexes with π-ligands (benzene as an outlier marked).

Theoretical BDEs

n class="Chemical">Next, we turned our attention to the theoretical calculations. We performed density functional theory (DFT) calculations at the mPW1PW91/cc-pVTZ:LanL2DZ level of theory including ZPE and the BSSE corrections to determine the theoretical BDEs (Table ). We also used the M06/cc-pVTZ/LanL2DZ level of theory including ZPE and BSSE corrections for the comparison—the results are in the Supporting Information (Table S1).

In the absolute values, the DFT predicted energies are on an average 0.3 eV smaller for the silver species and 0.2 eV smaller for the gold species compared to the experimental results. We observed the same difference between the experimental and theoretical values for the gold trimethylphospino complexes in the previous study.[34]

In relative terms, there is a very good agreement with the trends found experimentally (Figure b-d). For the silver complexes, the weakest interactions were predicted to be with benzene, trans-pentene, 1-pentene, and 1,1-dimethylallene—three of four of these complexes were determined as the weakest by the MS experiment as well (Table ). In addition, DFT predicts the strongest interaction between [Ag(PPh3)]+ and COD, also in agreement with the experiment (Table ). The reason for this becomes evident from the calculated structure (Figure ): the silver cation interacts symmetrically with both double C–C bonds. This is not the case for the complexes with 1,3-cyclooctadiene and 1,3-cyclohexadiene and 1,4-cyclohexadiene. The differences among the theoretical binding energies are smaller than in the experiment; most of the theoretical binding energies cluster around 1.10 eV. Strikingly, the mPW1PW91/cc-pVTZ/LanL2DZ calculations fail to correctly describe the interaction with tetramethylallene, which was determined as the second strongest from the experiment.

Figure 3

Optimized structures of the Ag/Au-triphenylphosphine π-complexes with (a) COD and (b) benzene at the mPW1PW91/cc-pVTZ/LanL2DZ level of theory. Coloring of the atoms: C, gray; P, green; Ag, black; and Au, yellow. The hydrogen atoms were removed for clarity.

For the [n class="Chemical">Au(n class="Gene">PPh3)(π-ligand)]+ complexes, the theory correctly predicts the lowest binding energy for n class="Chemical">benzene and the largest binding energies for 2-pentyne, cyclooctene, and tetramethylallene. The theoretical binding energy of the gold cation with benzene is similar to the interactions of the silver cation with some of the π-ligands, which correlates well with the experiment. For the benzene complexes, the employed level of theory predicts the η2 coordination rather than the η6 coordination for both the silver and the gold complexes (see Figure ).[69]

Realistic Model Situation for Silver Cations

The results reported above for the n class="Chemical">triphenylphosphino-silver(I) complexes do not properly illustrate a plausible situation in the solution because n class="Chemical">silver salts are added to reaction mixtures without an ancillary ligand—the first most probable reaction is the solvation of silver cation by molecules of the solvent. Because of this, we tried to determine the interaction of π-ligands with silver(I) cations with an additional ligand consisting of a solvent molecule, namely, dichloromethane, methanol, and acetonitrile. Unfortunately, we were unsuccessful in most cases because the desired π-complexes were presented with either only small intensity or not at all, and therefore we could not conduct CID experiments to determine their binding energies. However, in the case of using acetonitrile, it was possible to determine experimental BDEs for some of the π-complexes. We note that the [Ag(CH3CN)(π-ligand)]+ complexes were generated from the dichloromethane solutions of AgSbF6 and the hydrocarbon with a minimal addition of acetonitrile (see the Supporting Information for details). If the concentration of acetonitrile was too large, the only detected ions were the [Ag(CH3CN)2]+ ions.

Figure shows a comparison of the fragmentation patterns of [Ag(CH3CN)(1-pentene)]+ and [Ag(CH3CN)(COD)]+. The former dominantly eliminates 1-pentene, whereas the latter exclusively loses acetonitrile. Complexes with other hydrocarbons showed a similar fragmentation pattern to [Ag(CH3CN)(1-pentene)]+. This demonstrates that the binding energy of acetonitrile to the silver cation is on the order of or larger than the binding energies of most of the π-ligands; only the binding energy of COD exceeds the binding energy of acetonitrile.

Figure 4

CID MS spectra of dichloromethane solution of AgSbF6, CH3CN, and (a) 1-pentene or (b) COD.

CID MS spectra of n class="Chemical">dichloromethane solution of AgSbF6, CH3CN, and (a) n class="Chemical">1-pentene or (b) COD.

The experimentally determined results of the interaction of [Ag(CH3CN)]+ with selected unsaturated hydrocarbons are shown in Table . The measured values for the interaction of the unsaturated hydrocarbons with [Ag(CH3CN)]+ were approximately 0.3 eV larger than those with [Ag(PPh3)]+. In addition, the values were similar to the binding energies in [Au(PPh3)(π-ligand)]+ (see Table ). We also determined these values theoretically by the DFT calculations. The predicted values were shown to be approximately 0.15 eV smaller compared to the experimental values. On the other hand, the theoretical binding energies of the π-ligands in [Au(PPh3)(π-ligand)]+ and [Ag(CH3CN)(π-ligand)]+ were predicted to be basically the same, which again agrees well with the experiment.

Table 2

Experimental and Theoretical Binding Energies of [Ag(CH3CN)]+ with Unsaturated Hydrocarbons in the Gas Phase[63]

Calculations were performed at the mPW1PW91/cc-pVTZ/LanL2DZ level of theory and include ZPE and BSSE corrections.

Caln class="Chemical">culations were performed at the mPW1PW91/cc-pVTZ/Lanpan>L2DZ level of theory anpan>d inpan>clude ZPE anpan>d BSSE corrections.

We observed a competitive elimination of the hydrocarbon and n class="Chemical">acetonitrile from [Ag(CH3CN)(π-ligand)]+; therefore, we could have extracted the binding energies for acetonitrile as well. The binding energies of acetonitrile in all the investigated complexes were about 1.7 eV and they were about the same as the binding energies of the π-ligands. However, the elimination of the hydrocarbons from [Ag(CH3CN)(π-ligand)]+ largely prevailed. It probably points to a kinetic preference of the π-ligand elimination.

Infrared Photodissociation Spectroscopy

Photodissociation spectroscopy provides IR or UV/Vis spectra of mass-selected ions.[70−72] We used IRPD spectroscopy to link the binding energies with the changes of the structure of the carbon–carbon multiple bonds.[73] The stretching frequency of the unsaturated C–C bonds in the gold and silver π-complexes should correlate with their bond length and thus could be directly compared to their activation. We chose 2-pentyne as an example of the internal alkyne and measured IR spectra for both (triphenylphosphino)silver and (triphenylphosphino)gold π-complexes with 2-pentyne (Figure ).

Figure 5

Experimental IR of (a) gold-triphenylphosphine (b) and silver-triphenylphosphine π-complexes with 2-pentyne in the gas phase.

Experimental IR of (a) n class="Chemical">gold-triphenylphosphine (b) and n class="Chemical">silver-triphenylphosphine π-complexes with n class="Chemical">2-pentyne in the gas phase.

The C≡C stretching band of the [Ag(PPh3)(2-pentyne)]+ complex is located at 2162 cm–1(Figure b). The exchange of the silver cation by the gold cation leads to a clear red shift of the vibration (Figure a). However, the region of the C≡C stretching band contains two bands rather than just one, 2122 and 2137 cm–1, respectively. Most likely, the weaker band at 2122 cm–1 is a combination band; hence, the C≡C stretching is at 2137 cm–1 (see also S42 in the Supporting Information).

Discussion

The n class="Chemical">BDEs provide a good insight into the plausibility of the formation of the π-complexes and the degree of the activation of the multiple C–C bond in the complexes.[34] As to the plausibility of the formation of the complexes, the binding energies of the substrate molecules to the catalysts should be larger than those of the solvent molen class="Chemical">cules. We have shown that for likely the strongest interacting solvent molecule, acetonitrile, the formation of silver complexes with π-ligands versus with acetonitrile can be in competition. We do not need to consider the competition between the gold and silver complexes because the metals are used only in catalytic amounts (normally around 5–10%), which means that the unsaturated hydrocarbons are in a large excess and therefore the gold and the silver species do not compete for them.

The second objective is the comparison of the degree to which the silver and gold ionpan>s activate an unsaturated bond. We can follow several parameters to assess the activation, for example, the lengthening of the C–C unsaturated bond or the charge on the unsaturated C–C bond in the π-complexes.[74] For the silver complexes, we chose to include also various ancillary ligands to simulate the solvent, namely, acetonitrile, methanol, dichloromethane, and water.

First, we analyzed the length of the unsaturated bond in the π-complexes in the gas phase (Table ). By this measure, gold is expected to activate the unsaturated bonds more effectively as the gold provided the largest lengthening of the unsaturated bond in all complexes. For silver species, the lengthening was most profound for the complexes of water–silver and methanol–silver compared to the complexes with triphenylphosphine–silver having the smallest effect.

Table 3

Unsaturated Bond Length in Gold and Silver π-Complexes in the Gas Phase

Calculations were performed at the mPW1PW91/cc-pVTZ/LanL2DZ level of theory.

Caln class="Chemical">culations were performed at the mPW1PW91/cc-pVTZ/Lanpan>L2DZ level of theory.

Second, we caln class="Chemical">culated the natural atomic charges in the π-complexes by natural population analysis. The aim is to correlate the activation of the unsaturated bonds for a nucleophilic attack with the delocalization of the charge from the metals to the unsaturated bond (Table ). There are no big differences in charge delocalization among the complexes. The overall positive charge on the π-ligands ranged from 0.13 to 0.16 e in the gold complexes and these values dropped by only of 0.01–0.04 e for the corresponding silver complexes. The bigger differences (on the side of 0.04 e) were found in charge distributions in complexes with alkynes.

Table 4

Calculated Natural Atomic Charges in the Gold and Silver π-Complexes in the Gas Phase (UH Corresponds to an Unsaturated Hydrocarbon)

Calculations were performed at the mPW1PW91/cc-pVTZ/LanL2DZ level of theory.

Values in brackets correspond to a sum with directly attached hydrogen atoms.

Caln class="Chemical">culations were performed at the mPW1PW91/cc-pVTZ/Lanpan>L2DZ level of theory.

Values in brackets correspond to a sum with directly attached n class="Chemical">hydrogen atoms.

Overall, the population analysis suggests that the charge delocalization might be a part of the activation of the π-ligands in gold and silver complexes. However, the differences in C–C bond distances of these ligands in the gold and silver complexes attest that the π-back-bonding probably plays a more important role in the activation of the ligands than the positive charge.[75,76] This conclusion is in agreement with results obtained for silver(I) and gold(I) complexes bearing bidentate phosphine ligands. These complexes bind strongly with π-ligands, leading to significant C–C bond elongations.[44]

Conclusions

We determined the binding energies of various cationic silver(I) and gold(I) π-complexes in the gas phase experimentally and theoretically. The results show that for the same ancillary ligand PPh3, gold(I) binds stronger to unsaturated hydrocarbons than silver(I). The same trend was also confirmed spectroscopically by measuring vibrational frequencies of the C≡C bond in [Au(PPh3)(2-pentyne)]+ and [Ag(PPh3)(2-pentyne)]+. When a different supporting ligand on silver is used to simulate the conditions in a reaction mixture, where silver is likely coordinated to a solvent molecule, silver can bind to the π-ligands as strongly as gold. However, the activation of the π-ligands is always slightly smaller, measured by the prolongation of the unsaturated bond and the positive charge delocalization.

The results thus suggest that π-bonds in unsaturated hydrocarbons can be likely activated for nucleophilic attack by both ligated gold cations and solvated silver cations if both cations are present in a reaction mixture. Many reaction parameters can affect which mode of activation prevails for the reaction outcome. The reactivity of one or the other cation can be disfavored by steric effects, counterion efn class="Chemical">fects, solvation effects, and other factors.[27,77]