Growth of Au-Pd 2 Sn Nanorods via Galvanic Replacement and their Catalytic Performance on Hydrogenation and Sonogashira Coupling Reactions

Colloidal Pd 2 Sn and Au-Pd 2 Sn nanorods (NRs) with tuned size were produced by the reduction of Pd and Sn salts in the presence of size- and shape-controlling agents and the posterior growth of Au tips through a galvanic replacement reaction. Pd 2 Sn and Au-Pd 2 Sn NRs exhibited high catalytic activity toward quasi-homogeneous hydrogenation of alkenes (styrene and 1-octene) and alkynes (phenylacetylene and 1-octyne) in dichloromethane (DCM). Au-Pd 2 Sn NRs showed higher activity than Pd 2 Sn for 1-octene, 1-octyne and phenylacetylene. In Au-Pd 2 Sn heterostructures, XPS evidenced an electron donation from the Pd 2 Sn NR to the Au tips. Such heterostructures showed distinct catalytic behaviour in the hydrogenation of compounds containing a triple bond such as tolan. This can be explained by the aurophilicity of triple bonds. To further study this effect, Pd 2 Sn and Au-Pd 2 Sn NRs were also tested in the Sonogashira coupling reaction between iodobenzene and phenylacetylene in DMF. At low concentration, this reaction provided the expected product, tolan. However, at high concentration, more reduced products such as stilbene and 1,2-diphenylethane were also obtained, even without the addition of H 2 . A mechanism for this unexpected reduction is proposed.


INTRODUCTION
Multimetallic catalysts have associated several potential advantages over elemental compositions, including: 1-4 i) Cost reduction associated with the utilization of lower amounts of noble metals; ii) Additional degrees of freedom to tune the electronic structure toward the creation of suitable adsorption/reaction sites; iii) Close location of different adsorption/reaction sites enabling tandem reactions that potentially reduce the number of synthetic steps toward a specific product; iv) Allow for alternative multisite reaction paths that may be faster, more selective and/or prevent poisoning species to be formed or to remain at the catalyst surface.
Multimetallic catalysts can be realized by several methods including the incorporation of multiple metallic centers within a molecule or the impregnation of the different metals on a high surface area support. However, a particularly interesting and at the same time underexploited class of catalysts is that of size-, shape-and compositional-engineered multimetallic colloidal nanoparticles (NPs). Being unsupported, essentially solution-dispersed, colloidal NPs combine the advantages of classic homogeneous and heterogeneous catalysts: [5][6][7][8][9][10][11][12] Like catalytic organic molecules, colloidal NPs can be produced with extraordinary control over chemical and structural parameters, potentially enabling the rational engineering of their catalytic activity and especially selectivity in sensitive reactions. [12][13][14][15][16] Colloidal NPs also have extremely high surfaceto-volume ratios, which makes them potentially very active. Additionally, unlike molecular catalysts, NPs are easily separated from solvent, reactants and products, preventing product contamination and allowing catalyst reutilization in multiple cycles.
Multimetallic colloidal NPs are typically prepared by the co-reduction or thermal decomposition of precursors of the different metals and/or the heteronucleation of a second or third compound at the surface of a pre-formed NP. Additionally, atomic substitution reactions can be used to partially or totally modify the stoichiometry of pre-formed NPs obtaining new and eventually much more complex compositions. In this direction, galvanic replacement reactions, involving the substitution of lattice atoms by ions in solution mediating a redox reaction, are particularly suitable. [17][18][19][20][21][22] The galvanic replacement is driven by a difference in reduction potential between the replacing and replaced elements, which allows the reaction to proceed at moderate temperatures and minimizing homonucleation of independent NPs. However, in spite of its high potential and versatility, very few examples exist on the modification of the composition of multimetallic nanostructures by a galvanic replacement reaction. 23,24 The development of catalysts for organic reactions is driven by the search for cost-effective and environmentally-friendly processes suitable for a sustainable society. From the myriad of currently exploited catalytic reactions, hydrogenation and cross-coupling are among the most heavily studied. Hydrogenation comprises an exceedingly important group of reactions, including the Haber-Bosch process as well as the reduction of alkenes, aldehydes, ketones and imines. 25 On the other hand, cross-coupling reactions comprise several essential mechanisticallyrelated reactions, including Suzuki, Stille, Heck and Sonogashira couplings among others. 26 While numerous homo-and heterogeneous catalysts have been successfully applied in these reactions, several performance, economic and impact parameters, such as activity, selectivity, substrate scope, durability/recyclability, cost-effectiveness, environmental friendliness and sustainability require further improvement, making the design of better hydrogenation and crosscoupling catalysts a worth endeavor.
Pd-based multimetallic catalysts and particularly Pd-Sn alloys have raised especial attention in these reactions owing to their reduced cost and improved performance compared to bare Pt or Pd catalysts. 16,[26][27][28][29][30][31] We recently described the synthesis of Pd 2 Sn nanorods (NRs) with narrow size distribution and geometry control. 16 In the present paper we report a procedure based on a galvanic replacement reaction for the growth of Au tips onto Pd 2 Sn NRs to produce Au-Pd 2 Sn heterostructured NRs. With both types of NRs in hand, we compare the performance of Pd 2 Sn and Au-Pd 2 Sn NPs in alkene and alkyne hydrogenations and in Sonogashira couplings. These reactions have been chosen because Pd-Sn systems have been previously shown excellent performances 28 and there is also literature precedents demonstrating that supported Au NPs are active both in hydrogenation 32 and Sonogashira 33-38 couplings due the aurophilicity of alkynes. 39 (MAHC), oleylamine (OAm, >70%), hydrochloric acid (37% in water), styrene, 1-octene, phenylacetylene (PhA), 1-octyne, tolan, iodobenzene (PhI), potassium carbonate and potassium hydroxide were purchased from Sigma Aldrich. Tri-n-octylphosphine (TOP, 97%) was acquired from Strem. Analytical grade hexane, chloroform, N,N-dimethylformamide (DMF), dichloromethane, toluene and ethanol were obtained from various sources. All chemicals were used as received, except OAm, which was purified by distillation.
Synthesis of Pd 2 Sn NRs: Pd 2 Sn NRs were produced following our previous report 16  The sample temperature was set to 298.2 K. One dimensional (1D) 1H and 2D NOESY (Nuclear Overhauser Effect Spectroscopy) spectra were acquired using standard pulse sequences from the Bruker library. For the quantitative 1D 1H measurements, 64k data points were sampled with the spectral width set to 20 ppm and a relaxation delay of 30 s. NOESY mixing time was set to 300 ms and 4096 data points in the direct dimension for 512 data points in the indirect dimension were typically sampled, with the spectral width set to 10 ppm. Diffusion measurements (2D DOSY) were performed using a double stimulated echo sequence for convection compensation and with monopolar gradient pulses. 44 Smoothed rectangle gradient pulse shapes were used throughout. The gradient strength was varied linearly from 2 to 95% of the probe's maximum value in 64 increments, with the gradient pulse duration and diffusion delay optimized to ensure a final attenuation of the signal in the final increment of less than 10% relative to the first increment. For 2D processing, the spectra were zero filled until a 4096−2048 real data matrix.
Before Fourier transformation, the 2D spectra were multiplied with a squared cosine bell function in both dimensions, the 1D spectra were multiplied with an exponential window function. The diffusion coefficients were obtained by fitting the appropriate Stejskal-Tanner (ST) equation to the signal intensity decay. 43 Diffusion measurements (2D DOSY) were performed using a double stimulated echo sequence for convection compensation and with monopolar gradient pulses. 44 Smoothed rectangle gradient pulse shapes were used throughout.
with the gyromagnetic ratio of the observed 1H nucleus γ, the gradient pulse length δ, the gradient strength g, the diffusion time ∆ and the diffusion coefficient D.

Structural and chemical properties of Pd 2 Sn and Au-Pd 2 Sn NRs
Representative TEM micrographs of rod-shaped Pd 2 Sn NPs with three different sizes (10  2 nm × 4  1 nm; 26  2 nm × 9  1 nm; 40  5 nm × 11  2 nm) produced following the methodology detailed in the experimental section are shown in Figure 1a, 1c, 1e. NRs showed narrow size distributions and no apparent aggregation.
Au-Pd 2 Sn heteronanostructures were produced by injecting, at room temperature, a solution containing 0.02 mmol of AuCl 3 in 50 µl of OAm and 2 ml of ODE into 5 mL of a toluene dispersion of Pd 2 Sn NRs (5 mg/mL) under strong stirring. After 60 min reaction, multiple Au dots with average size of ca. 2 nm were grown along the whole NR surface for the larger NRs and preferentially at the NR ends in the smallest ones (Figures 1b, 1d, 1f). Figure 1g shows    could be also further reduced to Au 0 through charge transfer from Pd atoms (Figure 3a). 23 Table 1. Atomic ratios of 10 nm Pd 2 Sn and 12 nm Au-Pd 2 Sn NRs as obtained by EDX and XPS analyses, and atomic percentage of each oxidation state as obtained from the fitting of the XPS spectra.
To further clarify the composition and oxidation states of the different elements within Pd 2 Sn and Au-Pd 2 Sn NRs, samples were analysed using XPS (Figure 3b- 42 In parallel, NOESY spectra presented negative cross peaks corroborating the interaction of OAm with the Pd 2 Sn surface ( Figure S4b). These data pointed toward the possibility of a highly dynamic surface stabilization, as previously reported for other systems. 47,48 Hence, we studied the diffusion coefficient of our system (NR plus the ligand shell) using DOSY experiments. From DOSY data ( Figure S5), a diffusion coefficient of 120 ± 12 µm 2 s -1 was obtained for 26 nm Pd 2 Sn NRs. The diffusion coefficient calculated using the theoretical model of Mansfield and Douglas 49 for 26 nm NRs with rounded ends was 26 µm 2 s -1 . The much larger diffusion coefficient experimentally measured was ascribed to a highly dynamic binding of OAm on the Pd 2 Sn NRs surface.

Hydrogenation reactions
The exploration of the catalytic potential of Pd 2 Sn and Au-Pd 2 Sn NRs was started with the hydrogenation of styrene to ethylbenzene (   Tables 3 and 4 show results obtained from the hydrogenation of 1-octene and alkynes, namely phenylacetylene and 1-octyne, under the optimized conditions set with styrene. In the hydrogenation of 1-octene, apart from the hydrogenated product, octane, isomerisation of the substrate to the two geometric isomers of 2-octene was observed. This isomerization depends on the sequence of insertion and β-hydride elimination on the NR surface (Scheme S1) Similar results and tendencies were obtained for the hydrogenation of 1-octyne. On the other hand, in the hydrogenation of phenylacetylene, very little amounts of ethylbenzene were formed, evidencing the higher reactivity of phenylacetylene compared to styrene in hydrogenation.

NRs
In terms of the Au effect on hydrogenation reactions, Pd 2 Sn NRs were considerably more active in the hydrogenation of styrene than Au-Pd 2 Sn NRs, whereas the opposite trend was observed for 1-octene. The non-conjugated nature of the double bond and the lesser steric shielding of 1octene compared to styrene could explain this result. In contrast, the differences in performance were less important for phenylacetylene and 1-octyne, with Au-Pd 2 Sn being more active for both substrates. These trends could be understood by the known alkynophilicity of Au. [39][40][41]51 The high electron rich character of the triple bond probably hid the effect of the substituent. It should be mentioned that all the comparisons carried out between Pd 2 Sn and Au-Pd 2 Sn NRs were based on results obtained from the exact same batch of Pd 2 Sn NRs.

Sonogashira coupling reactions
Pd 2 Sn and Au-Pd 2 Sn NR performance in the Sonogashira reaction between phenylacetylene (PhA) and iodobenzene (phenyl iodide, PhI) to give diphenylacetylene (tolan) are displayed in Table 5. Using Pd 2 Sn NRs and K 2 CO 3 as base ( Table 5, entries 1-4), very high conversions could 20 be attained at 2 h (Table 5, entry 2). The reaction yielded mainly tolan, although products 1 and 1' were also formed. Both products were analyzed by GC-MS, providing a peak at m/z of 280 units, what proved that they were isomeric. These compounds arose from the condensation of two molecules of PhA and one molecule of PhI, which gave a pair of geometric isomers of a 1,3enyne ( Figure 4). The decrease of the catalyst loading ( When the base was changed to KOH (Table 5,   The recyclability of the Pd 2 Sn NRs in the Sonogashira reactions was studied by precipitating the nanoparticles by centrifugation and using them in successive runs. To the recovered NRs, a new batch of fresh reagents was added to carry out a second catalytic run under the same conditions. A 10 % drop in the conversion in the first run, but no change in the selectivity ( Figure S6). We associate this drop to non-full recovery of the material in each precipitation step. A partial NR aggregation during the recovery was also observed ( Figure S7), which likely reduced the NR catalytic activity (see SI for details).   (Table 7). Under Sonogashira conditions, Pd 2 Sn and Au-Pd 2 Sn NRs catalysed the reduction of tolan to St, but not to DPE. The reduction was quite selective to the formation of cSt, as it happened when the Sonogashira coupling was carried out at high concentration ( Table 6). As no trace of 1/1' or 2 was detected, it was concluded that the formation of these molecules needed PhI and PhA. Au-Pd 2 Sn NRs were less active for the reduction of tolan, possibly due to the stronger adsorption of the triple bond to Au. [39][40][41]51  From the above data, a mechanism for the formation of the different products could be envisaged (Scheme 1). The formation of the expected tolan through Sonogashira coupling (left in Scheme 1) involved oxidative addition of PhI, transmetallation with PhA and reductive elimination. The reduction of tolan catalyzed by the Pd 2 Sn or Au-Pd 2 Sn NRs lead to the formation of mainly cSt as shown by the data of Table 6 at high concentration and also by experiments of Table 7. An alternative mechanism (right in Scheme 1) might start with an initial insertion of PhA into the interesting to note that by this mechanism, tSt, the most thermodynamically stable compound, is mainly formed as shown in Tables 5 and 6.
Although the reduction of both St isomers produces DPE, the selectivity to cSt observed when performing the reactions at high concentration (Table 6) suggested that most of the DPE was formed by the reduction of cSt, shown in the left-hand side of Scheme 1. Au-Pd 2 Sn NRs, when the reaction was performed at high dilution. Interestingly, St was produced at higher concentration either by reduction of tolan or by a mechanism involving insertion of PhA into the NR-Ph bond. In this regard, the production of St can be considered a cascade or tandem reaction, which was performed with a better selectivity using Au-Pd 2 Sn NRS as a catalyst. Additionally, the deuterated experiments confirmed that the solvent (DMF) played a role in this cascade reaction. We are currently investigating the implications of this methodology with other substrates.