Chemical activation of atom-precise Pd₃ nanoclusters on γ-Al₂O₃ supports for transfer hydrogenation reactions

Abstract

Deposition of atom-precise nanoclusters onto solid supports is a promising route to synthesize model heterogeneous catalysts. However, to enhance nanocluster-support interactions, activation of the nanoclusters by removal of surface ligands is necessary. Thermal treatment to remove surface ligands from supported metal nanoclusters can yield highly active heterogeneous catalysts, however, the high temperatures employed can lead to poor control over the final size and speciation of the nanoclusters. As an alternative to high-temperature thermal treatments, chemical activation of [Pd₃(μ-Cl)(μ-PPh₂)₂(PPh₃)₃]⁺ (Pd₃) nanoclusters on γ-Al₂O₃ supports under mild reaction conditions has been demonstrated in this work. Hydride-based reducing agents such as NaBH₄, LiBH₄, and LiAlH₄ have been examined for the activation of the Pd₃ nanoclusters. The structural evolution and speciation of the nanoclusters after activation have been monitored using a combination of XAS, XPS, STEM-EDX mapping, and solid-state NMR techniques. The results indicate that treatment with borohydride reducing agents successfully removed surface phosphine and chloride ligands, and the extent of size growth of the nanoclusters during activation is directly correlated with the amount of borohydride used. Borate side products remain on the γ-Al₂O₃ surface after activation; moreover, exposure to high amounts of NaBH₄ resulted in the incorporation of B atoms inside the lattice of the activated Pd nanoclusters. LiAlH₄ treatment, on the other hand, led to no significant size growth of the nanoclusters and resulted in a mixture of Pd single-atom sites and activated nanoclusters on the γ-Al₂O₃ surface. Finally, the catalytic potential of the activated nanoclusters has been tested in the transfer hydrogenation of trans-cinnamaldehyde, using sodium formate/formic acid as the hydrogen donor. The catalytic results showed that smaller Pd nanoclusters are much more selective for hydrogenating trans-cinnamaldehyde to hydrocinnamaldehyde, but overall have lower activity compared to larger Pd nanoparticles. Overall, this study showcases chemical activation routes as an alternative to traditional thermal activation routes for activating supported nanoclusters by offering improved speciation and size control.

---

Introduction

Solid-supported nanoclusters (NCs) of Pd are highly selective and active catalysts for essential catalytic processes such as the selective hydrogenation of alkynes (e.g., acetylene), CO oxidation, C–C coupling, and dehydrogenation reactions.^1–6 The high catalytic selectivity and activity of NCs can be attributed to their large surface area and the size-dependent arrangement of surface atoms.^3–5 One common strategy to obtain solid-supported NCs is the deposition of pre-synthesized ligand-protected atom-precise NCs onto a support material followed by activation (i.e., removal of surface ligands).^5,7,8 Bottom-up, surface ligand-assisted synthetic approaches have enabled structural control over atom-precise NCs which contain well-defined metal cores with surface ligands coating the metal core.⁹ However, upon immobilization of NCs on solid supports (i.e., solid-supported NCs) for heterogeneous catalytic processes, their surface ligands can prevent strong support–NC interactions and also act as a barrier for substrates to access catalytically active metal surfaces during catalytic processes.^8,10 Therefore, atom-precise NCs are first deposited on a suitable solid support followed by removing the surface ligands from the NCs in a subsequent activation step.^7,8,10 A central question in the field is how to effectively remove the ligands from the surface of pre-fabricated NCs with minimum distortion in the structure and size of the NCs.^10–12

[Pd₃(μ-Cl)(μ-PPh₂)₂(PPh₃)₃]⁺ (abbreviated hereafter as Pd₃) NCs have shown promising catalytic activity and selectivity in cross-coupling reactions when used without immobilization on solid supports.^13,14 However, a recent study by Fairlamb and coworkers suggested that non-immobilized Pd₃ NCs are less stable during cross-coupling reactions compared to resin-immobilized Pd₃ NCs.⁶ Yun et al. demonstrated that non-immobilized Pd₃ NCs are almost inactive catalysts for the oxidation of benzyl alcohol, however, upon immobilizing the Pd₃ NCs on titanate nanotubes, a remarkable increase in their activity was observed due to metal–support interactions.¹⁵ These studies suggest that immobilizing Pd₃ NCs onto solid supports not only stabilizes the NCs but can also significantly enhance their catalytic potential.^6,15 Our group previously reported the immobilization of Pd₃ NCs onto activated carbon supports for hydrogenation and cross-coupling reactions, and found that Pd₃ NCs adsorbed on activated carbon supports do not behave as true heterogeneous catalysts, as intact Pd₃ NCs leached into the reaction mixture during cross-coupling reactions.¹⁶ Thus the presence of phosphine ligands on the Pd₃ NCs hindered strong interactions between NCs and the support. However, thermal activation of Pd₃ NCs in the air led to the complete removal of surface ligands, resulting in catalysts that exhibited true heterogeneous behavior.¹⁶ While the thermal activation of NCs is a straightforward strategy, drawbacks associated with this strategy include uncontrolled sintering of NCs to larger nanoparticles (NPs), particularly at higher activation temperatures.^16–20 Additionally, many metals, including Pd, tend to oxidize and form metal oxides at moderate temperatures of thermal activation.^16,20 Therefore due to the poor speciation and morphology control over NCs during thermal activation, the catalytic potential of the NCs can be adversely affected.^10,16

An alternative method to activate NCs is through chemical activation, where ligands are removed from the surface of the NCs following chemical treatment.^10,21 One of the most promising strategies for chemically activating NCs involves treating them with reducing agents.¹⁰ This process includes the reduction of surface ligands, which are subsequently removed as free ligands.¹⁰ Collins et al. have shown a comparison between the thermal and reductive chemical activation of oleylamine, dodecanethiol, and polyvinylpyrrolidone-stabilized Pd NPs.²² They found that the chemical activation of Pd NPs using NaBH₄ was less effective in completely removing strongly binding ligands such as dodecanethiol from the surfaces of NPs than thermal activation routes.²² However, the chemically activated NPs were found to be more catalytically active in cross-coupling reactions compared to both non-activated and thermally activated NPs.²² Several groups, including our own, have previously shown that thiolate monolayer-protected Au NCs can be reductively activated in the presence of excessive borohydride reducing agents, and the post-activation size growth of NCs directly depends upon both the concentrations and strength of reducing agents used for the activation.^23–25 For example, the activation of γ-Al₂O₃-supported Au₂₅(SC₈H₉)₁₈ NCs with an excess of NaBH₄ led to the partial removal of thiolates from the Au NCs and resulted in minimal size growth of Au NCs.²⁵ Zhang and co-workers noted that hydrides exhibit a higher binding affinity to the surface of Au compared to organothiolate ligands, and consequently, an excess of hydride species can displace surface ligands from precious metals.²⁶ To the best of our knowledge, reductive chemical routes toward the activation of Pd NCs have not yet been explored and require further investigation. Chemical activation pathways have the potential to act as an efficient way to activate Pd NCs while avoiding the sintering and oxidation of Pd NCs seen during thermal activation in air.¹⁶

In this study, the chemical activation of γ-Al₂O₃-supported atomically precise Pd₃ NCs using hydride-based reducing agents i.e., NaBH₄, LiBH₄, and LiAlH₄, has been investigated. Various loadings of these reducing agents were examined to explore their impact on the activation process and subsequent structural changes of the NCs. The findings from this work indicate that all the selected reducing agents effectively removed phosphine and chloride ligands from the Pd₃ NC surfaces. In the case of NaBH₄ and LiBH₄, ligand removal was accompanied by very mild post-activation growth of the Pd₃ NCs at lower reducing agent amounts, while higher amounts of NaBH₄ during activation promoted the transformation of Pd₃ NCs into larger Pd NPs and led to the incorporation of B into the NPs. In contrast, LiAlH₄ treatment removed ligands from the Pd₃ NCs without any noticeable size growth, facilitating the formation of a mix of activated NCs and single-atom Pd sites on the γ-Al₂O₃ surface. The structural evolution of Pd₃ NCs during activation directly impacted the catalytic activity of the NCs, as demonstrated in their use as catalysts for the transfer hydrogenation of trans-cinnamaldehyde using sodium formate as the hydrogen donor. Smaller NCs exhibited high selectivity for hydrocinnamaldehyde formation, albeit with lower activity compared to larger Pd NPs. This work demonstrates the facile removal of phosphine and chloride ligands from Pd₃ NC surfaces via chemical means, highlighting how the choice and amount of the reducing agent used can lead to distinct structural transformations during activation.

Results and discussion

The synthesis, purification, and characterization of Pd₃ NCs were carried out following protocols established in the literature and our previous work.^16,27 Upon immobilization of the Pd₃ NCs on the surface of γ-Al₂O₃, the resulting composite has been denoted as Pd₃/Al₂O₃. Elemental analysis of the Pd₃/Al₂O₃ by atomic absorption spectroscopy showed a 0.5 wt% loading of the Pd on the γ-Al₂O₃ support. For chemical activations, three amounts of NaBH₄, LiBH₄, and LiAlH₄ have been used (as shown in Table S1†). Based on the reducing agents and their concentrations, the activated samples have been labelled as Pd/Al₂O₃-RA(x) (where RA is the used reducing agent) (i.e., NaBH₄, LiBH₄ or LiAlH₄) and x is the molar equivalents of the reducing agent used with respect to the Pd (labelled A, B, or C (Table S1†)). For the NaBH₄ reducing agent, 300 (A), 600 (B), or 1200 (C) mole equivalents were used for activation. Meanwhile, for LiBH₄ and LiAlH₄ reducing agents, 10 (A), 50 (B), and 100 (C) mole equivalents of the reducing agent were used for activation. The amounts of NaBH₄ used were larger because our previous work demonstrated that higher concentrations of aqueous NaBH₄ were necessary to activate thiolate-protected Au NCs compared to LiBH₄.²⁵

To investigate the speciation and bonding environment of the NCs after chemical activation, XAS analysis has been performed. Pd K edge XANES spectra of all the activated samples have been provided in Fig. 1(a–c), along with Pd, PdO, and Pd(OH)₂ standards. Upon treatment with reducing agents, small changes in the shape of the white lines compared to the initial Pd₃/Al₂O₃ were observed. These changes indicate a modification in the speciation and coordination environment of the supported Pd NCs due to activation. In the case of Pd₃/Al₂O₃–NaBH₄(A) and Pd₃/Al₂O₃–LiAlH₄(A/B/C), the feature above the white line appeared slightly more intense compared to Pd₃/Al₂O₃, suggesting an increase in Pd(II) species. In contrast, the intensity of the sharp feature above the white line in the remaining NaBH₄ (Pd₃/Al₂O₃–NaBH₄(B/C)) and LiBH₄ (Pd₃/Al₂O₃–LiBH₄(A/B/C)) treated samples were found to be relatively lower, suggesting more Pd in the zerovalent state. However, the small XANES changes observed at the edge after activation suggest that XANES data does not offer detailed insights into the speciation changes within these systems. In contrast, changes detected by EXAFS (discussed below) are more informative.

Fig. 1 Pd K-edge XANES spectra of the Pd₃/Al₂O₃ activated via treatment with different amounts of (a) NaBH₄, (b) LiBH₄ and (c) LiAlH₄ along with some standards. Pd K-edge Fourier transformed EXAFS spectra in R-space for Pd₃/Al₂O₃ activated via treatment with different amounts of (d) NaBH₄, (e) LiBH₄ and (f) LiAlH₄. All EXAFS spectra are shown without phase shift corrections.

To investigate the coordination environment of the activated Pd₃ NCs, EXAFS analysis at the Pd K edge was performed. The Fourier-transformed (FT) EXAFS spectra of activated samples along with the spectra of Pd₃/Al₂O₃ are provided in Fig. 1(d–f). In the spectra of Pd₃/Al₂O₃, a broad peak at around 1.6–1.8 Å can be seen due to the Pd–P and Pd–Cl interactions of the Pd core with the phosphine and chloride ligands. The small feature at around 2.7 Å can be assigned to Pd–Pd bonding in the triangular metal core of Pd₃ NCs. Notably, the Pd–Pd bond distance in Pd₃/Al₂O₃ was found to be slightly longer than that in fcc Pd, which is also the case in the reported single-crystal structure of Pd₃ NCs.²⁷ In the EXAFS spectra of all the activated NCs, an intense peak can be observed at 1.5 Å, which is shifted to the left in comparison to the Pd–P and Pd–Cl peak of the non-activated Pd₃/Al₂O₃ (at 1.6–1.8 Å). This suggests a complete removal of phosphines and chloride surface ligands from the metal core of the Pd₃ NCs upon activation; the peak at 1.5 Å in the Pd K edge EXAFS spectra is likely indicative of new Pd–O interaction. The appearance of this peak in activated samples indicates the interaction of Pd NCs with the basic surface oxygen sites of γ-Al₂O₃ after the removal of surface ligands from the Pd₃ NCs.

In the case of Pd₃/Al₂O₃–NaBH₄(A), the peak pertaining to the Pd–Pd interaction (originally at 2.7 Å) became less prominent in comparison to the Pd–Pd peak of Pd₃/Al₂O₃. This is likely due to the partial collapse of the triangular core of the Pd₃ NCs after the removal of surface ligands. Similar behaviour was previously observed with the carbon-supported Pd₃ NCs during their mild thermal activation at 150 °C.¹⁶ With the use of higher amounts of NaBH₄ (i.e., Pd₃/Al₂O₃–NaBH₄(B/C)) for activation, the peak for the Pd–Pd interactions around 2.6 Å became higher in intensity than seen for un-activated Pd₃/Al₂O₃, indicating an increase in the size of the Pd NCs during the activation. Meanwhile, the intensity of the Pd–O peak at around 1.5 Å in the spectra of Pd₃/Al₂O₃–NaBH₄(B/C) became lower in intensity compared to the corresponding Pd₃/Al₂O₃–NaBH₄(A). This implies the reduction in the Pd–O interaction between Pd and the support due to the growth in the size of the Pd NPs. Furthermore, the peak shapes in the 1.5–2.0 Å region in the FT EXAFS spectra of the Pd₃/Al₂O₃–NaBH₄(B/C) samples differ significantly from those observed in the Pd₃/Al₂O₃–NaBH₄(A) sample. However, confirming whether this change in peak shape is attributable to the presence of a second scatterer cannot be achieved by solely examining the FT EXAFS spectra. To further investigate the possibility of the presence of a second weak scatterer, wavelet transformed (WT) EXAFS spectra were generated.²⁸ In the WT EXAFS spectra of Pd₃/Al₂O₃–NaBH₄(A), a single peak can be seen at R ∼ 1.5 Å (Fig. S1(c and d)†) across all the k range, which reflects the presence of a single scatter i.e., O of the γ-Al₂O₃ support. However, in the cases of Pd₃/Al₂O₃–NaBH₄(B/C), significantly intense peaks around R ∼ 2.6 Å from Pd–Pd interactions can be seen in the WT EXAFS data (Fig. S1(e and g)†). Furthermore, in the region between R-values of 1.5 to 2.0 Å in the FT EXAFS spectra, two distinct peaks can be observed in different k-regions (0.5 to 8.5 and 8.5 to 12 Å⁻¹) of the WT EXAFS spectra, suggesting the probable existence of two scattering atoms (Fig. S1(f and h)†). Based on STEM-EDX mapping, solid-state NMR, and XPS analyses indicating the presence of B (with the B source being the borohydride-reducing agent) in these samples (vide infra), it is likely that the two scatterers are O and B.

In the FT EXAFS spectra of LiBH₄-treated samples (Pd₃/Al₂O₃–LiBH₄(A/B/C)) (Fig. 1e), the peaks of Pd–Pd interaction around 2.6 Å appeared to be higher in the intensity in comparison to Pd₃/Al₂O₃, reflecting a growth in the size of the Pd₃ NCs upon activation. The growth in the Pd–Pd peak became larger with an increase in the amount of LiBH₄ for activation (shown in Fig. 1e) suggesting the formation of larger Pd NCs with higher amounts of LiBH₄. All the WT EXAFS spectra of the LiBH₄-activated NCs (i.e., Pd₃/Al₂O₃–LiBH₄(A/B/C)) (Fig. S2†) showed the presence of a single scatter for the first peak of the FT EXAFS around R ∼ 1.5 Å. This affirms that Pd atoms in these systems are attached only to the O atoms of the surface of γ-Al₂O₃ after the activation, and no secondary scatterers are present. Moreover, another peak for the Pd–Pd interactions at around R ∼ 2.6 Å in the WT EXAFS spectra can also be seen, which became more prominent with a further increase in the amount of the reducing agent. Finally, for Pd₃/Al₂O₃–LiAlH₄(A/B/C) (Fig. 1f), the FT EXAFS spectra showed that the peak corresponding to Pd–Pd interactions at 2.6–2.7 Å is nearly completely absent. This indicates that the triangular metal core of Pd₃ NCs has been broken after activation using LiAlH₄ without any significant size growth. Similar to the LiBH₄ treated samples, the LiAlH₄ treated samples have shown the presence of only one scatterer for the first peak at R ∼ 1.5 Å in the WT EXAFS spectra (Fig. S3†), indicating the presence of Pd–O interactions. In addition, a very weak peak can be seen at around R ∼ 2.6 Å for the Pd–Pd bonds.

EXAFS data modelling for these samples have been performed to get further insight into the structural evolution of the Pd₃ NCs upon activation. The k-space EXAFS fits for Pd₃/Al₂O₃ activated via different NaBH₄, LiBH₄, and LiAlH₄ amounts are shown in Fig. S4–S6,† respectively. The corresponding fitting parameters from the EXAFS fits can be found in Table 1. The fitting of the non-activated Pd₃/Al₂O₃ sample was carried out using Pd–P, Pd–Cl, and Pd–Pd interactions based on the structure of [Pd₃(μ-Cl)(μ-PPh₂)₂(PPh₃)₃]⁺.²⁷ EXAFS fitting of the Pd₃/Al₂O₃ data provided coordination number (CN) values of 2.9 ± 0.3, 0.7 ± 0.3, and 1.8 ± 0.1 for Pd–P, Pd–Cl, and Pd–Pd bonds, respectively, which is quite consistent with the crystal structure of Pd₃ NCs.²⁷ The fitting of Pd₃/Al₂O₃–NaBH₄(A) suggested the presence of Pd–O contributions at a bond distance of 2.009 ± 0.001 Å with a CN value of 3.3 ± 0.1, likely due to metal–support interactions. Compared to Pd₃/Al₂O₃, a drop in the CN of Pd–Pd for this sample can be due to the distortion in the triangular core of the Pd₃ NCs upon activation. Since the WT EXAFS spectra of Pd₃/Al₂O₃–NaBH₄(B) and Pd₃/Al₂O₃–NaBH₄(C) (Fig. S1†) indicated the presence of two scatterers in the first coordination shell of Pd, and solid-state NMR and XPS also confirmed the presence of B in these two samples (vide infra), a Pd–B model has been chosen for fitting along with the Pd–O model for the peak present at R ∼ 1.5 Å in the FT EXAFS of these two samples. B atoms are small enough to be intercalated inside the interstitial sites of the Pd fcc lattice.^29–31 In the literature, the use of borohydride reducing agents such as (CH₃)₂NH·BH₃,²⁹ NaBH₄,³⁰ and C₄H₈O·BH₃ ³¹ during the synthesis of Pd NPs has been reported to show the incorporation of B in the lattice of Pd NPs. Chen and co-workers have found that the B content in the Pd lattice can be directly correlated with the concentration of NaBH₄ used during the synthesis of Pd NPs.³⁰ From the EXAFS modelling, the CNs for Pd–O, Pd–B, and Pd–Pd interactions in Pd₃/Al₂O₃–NaBH₄(B) were 2.8 ± 0.1, 2.3 ± 0.6, and 2.4 ± 0.3, respectively. Bond lengths for the new Pd–B interactions were fitted at 1.87(3) Å and 1.98(2) Å, respectively, for the Pd₃/Al₂O₃–NaBH₄(B) and Pd₃/Al₂O₃–NaBH₄(C) samples. These Pd–B bond lengths are consistent with Pd–B bond lengths seen for PdB_x solid-state solutions, molecular complexes, and B-doped Pd NPs in the literature (ca. 1.9–2.2 Å).^32–34 In Pd₃/Al₂O₃–NaBH₄(C), the fitted CN values for Pd–B (3.3 ± 0.1) and Pd–Pd (4.0 ± 0.5) interactions are slightly higher than in the previous sample, while the CN for Pd–O has decreased significantly to 1.4 ± 0.3. This implies that as the amount of NaBH₄ used for activation increased, the size of the final Pd NPs also increased, along with the incorporation of B in the lattice of Pd NPs. The decrease in Pd–O interactions can be attributed to reduced metal–support interactions as the size of the supported NPs increased.

Table 1 EXAFS fitting parameters of Pd₃/Al₂O₃ samples at the Pd K edge upon activation

Catalyst	Component	CN	R (Å)	σ ^2/Å^2	E 0 shift (eV)	R-factor
Pd3/Al2O3	Pd–P	2.9 (3)	2.250 (2)	0.0063 (1)	1.1 (2)	0.019
	Pd–Cl	0.7 (3)	2.312 (3)	0.0058 (3)
	Pd–Pd	1.8 (1)	2.98 (1)	0.0101 (5)
Pd3/Al2O3–NaBH4(A)	Pd–O	3.3 (1)	2.009 (1)	0.0028 (2)	5.3 (2)	0.004
	Pd–Pd	1.1 (1)	2.65 (1)	0.010 (2)
Pd3/Al2O3–NaBH4(B)	Pd–O	2.8 (1)	1.99 (1)	0.0043 (2)	9.2 (7)	0.003
	Pd–B	2.3 (6)	1.87 (3)	0.005 (4)
	Pd–Pd	2.4 (3)	2.762 (7)	0.010 (1)
Pd3/Al2O3–NaBH4(C)	Pd–O	1.4 (3)	2.01 (3)	0.0048 (2)	9.6 (5)	0.014
	Pd–B	3.3 (1)	1.98 (2)	0.006 (2)
	Pd–Pd	4.0 (5)	2.781 (3)	0.009 (3)
Pd3/Al2O3–LiBH4(A)	Pd–O	2.3 (1)	2.014 (4)	0.0021 (6)	2.4 (8)	0.016
	Pd–Pd	2.0 (5)	2.74 (1)	0.012 (9)
Pd3/Al2O3–LiBH4(B)	Pd–O	2.7 (1)	2.006 (2)	0.0022 (3)	−2.5 (2)	0.007
	Pd–Pd	2.3 (1)	2.727 (4)	0.0075 (4)
Pd3/Al2O3–LiBH4(C)	Pd–O	2.4 (1)	2.000 (3)	0.0027 (5)	3.2 (3)	0.012
	Pd–Pd	2.5 (1)	2.725 (5)	0.0073 (6)
Pd3/Al2O3–LiAlH4(A)	Pd–O	2.5 (1)	2.017 (3)	0.0025 (5)	4.7 (4)	0.017
	Pd–Pd	0.5 (2)	2.68 (3)	0.011 (4)
Pd3/Al2O3–LiAlH4(B)	Pd–O	2.8 (1)	2.008 (3)	0.0024 (4)	4.6 (4)	0.019
	Pd–Pd	0.6 (1)	2.69 (2)	0.007 (2)
Pd3/Al2O3–LiAlH4(C)	Pd–O	2.5 (1)	2.013 (2)	0.0024 (3)	4.1 (4)	0.022
	Pd–Pd	0.5 (1)	2.69 (2)	0.008 (2)

---

For the LiBH₄-activated samples (Pd₃/Al₂O₃–LiBH₄(A/B/C)), the CN for the Pd–O interaction was in the 2.3–2.7 range for the three samples, while the CN for Pd–Pd interactions was 2.0 ± 0.5, 2.3 ± 0.1, and 2.5 ± 0.1 for Pd₃/Al₂O₃–LiBH₄(A), Pd₃/Al₂O₃–LiBH₄(B), and Pd₃/Al₂O₃–LiBH₄(C), respectively. This suggests a very slight growth of NCs after LiBH₄ treatment. Similar to LiBH₄-activated samples, the CN of Pd–O in samples activated by LiAlH₄ (Pd₃/Al₂O₃–LiAlH₄(A/B/C)) was found to be around 2.5–2.8 for the three samples. However, the Pd–Pd interactions showed a significantly lower CN of approximately 0.5 in the fitting of these samples. This indicates a substantial disruption of Pd–Pd bonds in the triangular metal core of Pd₃ NCs upon LiAlH₄ activation. This is consistent with the possibility of a mixture of Pd NCs and single-atom sites in the LiAlH₄-activated samples. Unlike the NaBH₄ and LiBH₄-treated samples, no correlation between the loadings of the reducing agent and the resulting Pd–Pd CN of the activated NCs has been observed in the case of samples activated via LiAlH₄. Overall, the XAS results showed that all the reducing agents can activate Pd₃ NCs by removing phosphine and chloride ligands from their surfaces. After activation, the Pd atoms interact with oxygen atoms on the surface of γ-Al₂O₃. In the case of NaBH₄ and LiBH₄-activated samples, the structure and morphology of the NCs upon activation depend on the amount of reducing agent used.

XPS analysis was conducted on activated Pd₃/Al₂O₃ samples to investigate the speciation of Pd and validate the presence of B atoms in the samples. The Pd 3d region of XPS spectra for Pd₃/Al₂O₃, Pd₃/Al₂O₃–NaBH₄(A/B/C), and Pd₃/Al₂O₃–LiAlH₄(A/B/C), along with the Cl 2p, B 1s, and P 2s regions of Pd₃/Al₂O₃–NaBH₄(A/B/C) have been shown in Fig. 2. The Pd 3d XPS spectra of the Pd₃/Al₂O₃ showed peaks corresponding to two components of Pd, with 3d_5/2 peaks at approximately 336.1 and 338.0 eV and an area ratio close to 1 : 2, consistent with XPS results reported in the literature (Fig. 2a).¹⁶ In the case of the samples activated using NaBH₄ (Pd₃/Al₂O₃–NaBH₄(A/B/C)), two species of Pd can be seen in the XPS spectra i.e., Pd(0) and Pd(II) with Pd 3d_5/2 peaks at around 335.3 and 337.2 eV respectively (Fig. 2b). The fraction of Pd(0) relative to Pd(II) can be seen increasing with the amount of NaBH₄ used for the activation progressing from Pd₃/Al₂O₃–NaBH₄(A) to Pd₃/Al₂O₃–NaBH₄(C), which can be due to reduction of Pd(II) to Pd(0) upon exposure to NaBH₄. For Pd₃/Al₂O₃–NaBH₄(C), the XPS spectra only showed Pd(0), with the Pd(II) fraction completely absent. This observation is consistent with the XAS analysis, which indicated that the size of the Pd NPs increased at higher NaBH₄ loadings, while the interaction of Pd atoms with the O atoms of the γ-Al₂O₃ decreased. The interaction of Pd with surface O of γ-Al₂O₃ likely accounts for the presence of the high-valent Pd(II) species. The absence of high valent Pd(II) species in the XPS data for the Pd₃/Al₂O₃–NaBH₄(C) sample may be due to the presence of buried Pd–O interfaces for the larger Pd NPs (as EXAFS data shows a Pd–O CN of 1.4(3) for this sample).

Fig. 2 (a) XPS Pd 3d spectra of Pd₃/Al₂O₃, XPS Pd 3d spectra of Pd₃/Al₂O₃ catalysts treated with (b) NaBH₄, and (d) LiAlH₄; black dots and orange solid lines represent the experiment data and fit respectively; red lines represent the 3d_5/2 (right) and 3d_3/2 (left) regions for Pd(0); blue lines represent the 3d_5/2 (right) and 3d_3/2 (left) regions for Pd(II). (c) XPS Cl 2p + B 1s + P 2s spectra of Pd₃/Al₂O₃ catalysts treated with NaBH₄; black dots and orange solid lines represent the experiment data and fit respectively; red lines represent the 2p_3/2 (right) and 2p_1/2 (left) regions for the first Cl component; blue lines represent the 2p_3/2 (right) and 2p_1/2 (left) regions for the second Cl component; purple lines represent the P 2s region and pink lines represent the B 1s region.

While Pd–P and Pd–Cl interactions were not present in the EXAFS spectra of the Pd₃/Al₂O₃–NaBH₄(A/B/C), in the XPS analysis, peaks for phosphine and chloride species can be observed for these samples (Fig. 2c). Most likely, these ligands were adsorbed onto the surface of γ-Al₂O₃ after being removed from the metal core of Pd₃ NCs during the activation. In the literature, thiolate-protected Au₂₅(SR)₁₈ and Ag₂₅(SR)₁₈ NCs have been reported to exhibit similar behavior during thermal activation on a solid support.^18,35 A very weak feature in the Cl 2p region of Pd₃/Al₂O₃–NaBH₄(A) can be seen with the Cl 2p_3/2 peak at 198.3 eV. With further increase in the concentration of the NaBH₄ for the activation i.e., Pd₃/Al₂O₃–NaBH₄(B/C), peaks for two distinct components of Cl can be observed with Cl 2p_3/2 peak position at 197.6 and 200.0 eV. The peak at around 197.6 eV can be assigned to NaCl and the peak at around 200.0 eV is likely due to the interaction between Al and Cl.^36,37 A strong P 2s signal in the samples can be seen at ca. 190 eV. The B 1s region in the XPS is very close to the P 2s region. In the case of Pd₃/Al₂O₃–NaBH₄(A/B/C), a B 1s component in the XPS spectra at around 192.3 eV (Fig. 2c) has been fitted. The hydrolysis of NaBH₄ typically produces borates as side products,³⁸ and the location of the B 1s peak in the spectra of the activated samples is close to that of the reported shift of sodium tetraborate (i.e., 192.6 eV).³⁶ Therefore this peak can be assigned to tetraborate side products adsorbed on the γ-Al₂O₃ support. The relative peak area of this 192.3 eV B 1s peak increased with higher amounts of NaBH₄ used for activation. This suggests that higher amounts of the reducing agent resulted in greater adsorption of tetraborate species on the surface of the γ-Al₂O₃. While XAS studies indicated that B also begins to intercalate into Pd NPs at large NaBH₄ excesses, a separate B 1s peak for this species can not be assigned, which is likely due to overlap with the large P 2s signal present at 190 eV.

Similar to Pd₃/Al₂O₃–NaBH₄(A/B), the Pd 3d XPS spectrum of Pd₃/Al₂O₃–LiBH₄(B) (Fig. S7a†) showed the presence of two Pd components with Pd 3d_5/2 peaks at 335.3 eV for Pd(0) and 337.1 eV for Pd(II). The peak area for the Pd(II) component was found to be slightly lower than that of Pd(0). Conversely, in the case of the LiAlH₄-treated samples i.e., Pd₃/Al₂O₃–LiAlH₄(A/B/C), the peak area for the Pd(II) component has been found higher than that of Pd(0), indicating that much of the Pd is present in a higher-valent state after activation (Fig. 2d). Moreover, the relative peak areas of Pd 3d regions for Pd₃/Al₂O₃–LiAlH₄(A/B/C) did not vary much from each other, suggesting identical speciation of Pd, which is consistent with EXAFS analysis. The Al 2p XPS spectra for the Pd₃/Al₂O₃–NaBH₄(A/B/C) and Pd₃/Al₂O₃–LiAlH₄(A/B/C) are shown in Fig. S7b and S7c,† respectively. The peak in the Al 2p region at approximately 74.1 eV is consistent with the reported binding energy value of γ-Al₂O₃.³⁹ The Al 2p region did not show the appearance of any new peaks after treatment with the reducing agents, suggesting that the surface γ-Al₂O₃ remained unchanged during the reducing agent treatments.

To validate the presence of adsorbed side products of B in the NaBH₄-and LiBH₄-activated samples, solid-state ¹¹B magic angle spinning (MAS) NMR analysis was performed on Pd₃/Al₂O₃–NaBH₄(A/B/C) and Pd₃/Al₂O₃–LiBH₄(B) samples (Fig. 3). The ¹¹B MAS NMR spectra of all the activated NCs showed three peaks at 15.6, 10.5, and 1.3 ppm. The positions and shapes of these peaks are in good agreement with those of sodium tetraborate reported in the literature.⁴⁰ This indicates that during the activation of Pd₃ NCs, NaBH₄ and LiBH₄ formed tetraborate (or related species) as a side product and remained adsorbed onto the γ-Al₂O₃ surface even after several washing cycles. Similar peaks were seen for a control sample in which γ-Al₂O₃ was treated with NaBH₄ in the absence of Pd₃ NCs (Fig. 3a). For the Pd₃/Al₂O₃–NaBH₄(B) sample, a new small peak was found around 5.8–5.9 ppm in ¹¹B MAS NMR, which increased in intensity further for the Pd₃/Al₂O₃–NaBH₄(C) sample. Chan et al. reported a similar shift for the peaks of intercalated or surface B in fcc Pd.³¹ Therefore, the peak at 5.9 ppm in ¹¹B MAS NMR was attributed to the intercalated B atoms inside the Pd lattice, which is consistent with EXAFS data showing Pd–B interactions in these samples. However, the new peak for intercalated B was not seen for the LiBH₄-treated samples, showing that there was no B incorporation in the Pd NCs, likely because of the smaller concentrations of LiBH₄ used for the activation compared to NaBH₄. These results are consistent with the absence of a Pd–B interaction in the EXAFS data for the LiBH₄-activated samples. Finally, ³¹P MAS NMR of these samples also indicated the complete removal of phosphine ligands from the Pd₃ NCs, however, peaks for remaining phosphine impurities on the surface of the γ-Al₂O₃ after activation were present (Fig. S8†).

Fig. 3 ¹¹B MAS NMR of (a) Al₂O₃–NaBH₄(C) control, (b) Pd₃/Al₂O₃–NaBH₄(A), (c) Pd₃/Al₂O₃–NaBH₄(B), (d) Pd₃/Al₂O₃–NaBH₄(C) and (e) Pd₃/Al₂O₃–LiBH₄(B).

HAADF-STEM analyses of the activated samples were performed to investigate the size, shape, and distribution of the NCs upon activation (Fig. 4). The Pd₃ NCs appeared to be well-dispersed on the surface of the γ-Al₂O₃ (Fig. 4a), with an average NC size of 0.9 ± 0.3 nm. Upon activation with low amounts of NaBH₄, the average NC size was 0.8 ± 0.4 nm for Pd₃/Al₂O₃–NaBH₄(A) (Fig. 4b). In contrast, for Pd₃/Al₂O₃–NaBH₄(B), an increase in Pd particle size was observed after activation, and Pd NPs with an average size of 3.4 ± 0.8 nm are apparent in both HR-TEM (Fig. S9a†) and HAADF-STEM (Fig. 4c) images. The HAADF-STEM image of Pd₃/Al₂O₃–LiBH₄(A) revealed the formation of NCs with an average size of 1.7 ± 0.5 nm (Fig. 4d), while activation with a higher concentration of LiBH₄ in Pd₃/Al₂O₃–LiBH₄(B) resulted in slightly larger NCs (2.5 ± 0.6 nm) (Fig. S9b† and Fig. 4e). In the HAADF-STEM images of Pd₃/Al₂O₃–LiAlH₄(A) (Fig. 4f) and Pd₃/Al₂O₃–LiAlH₄(B) (Fig. S10†), Pd NCs with average sizes of approximately 1.1 ± 0.5 nm and 1.5 ± 0.4 nm are apparent. However, it should be noted that the EXAFS fitting revealed a very low Pd–Pd CN of around 0.5 for these samples, which would suggest that the sample contains both NCs and single-atom sites γ-Al₂O₃ surface, the latter of which could not be imaged.

Fig. 4 HAADF-STEM images of (a) Pd₃/Al₂O₃ (avg. Pd size: 0.9 ± 0.3 nm), (b) Pd₃/Al₂O₃–NaBH₄(A) (avg. Pd size: 0.8 ± 0.4 nm), (c) Pd₃/Al₂O₃–NaBH₄(B) (avg. Pd size: 3.4 ± 0.8 nm), (d) Pd₃/Al₂O₃–LiBH₄(A) (avg. Pd size: 1.7 ± 0.5 nm), (e) Pd₃/Al₂O₃–LiBH₄(B) (avg. Pd size: 2.5 ± 0.6 nm) and (f) Pd₃/Al₂O₃–LiAlH₄(A) (avg. Pd size: 1.1 ± 0.5 nm).

To visualize the dispersion of the Pd particles and borates on the γ-Al₂O₃ support, STEM-EDX elemental mapping analysis of the activated samples was performed. STEM-EDX images of Pd₃/Al₂O₃–NaBH₄(A) (Fig. S11†) showed a homogenous distribution of Pd over the surface of the γ-Al₂O₃ support. In the case of Pd₃/Al₂O₃–NaBH₄(B), some larger Pd particles can be observed on the γ-Al₂O₃ support in the STEM-EDX images (Fig. S12†). However, in both of these cases (i.e., for Pd₃/Al₂O₃–NaBH₄(A/B)), the B density did not seem to be localized near the position of Pd (Fig. S11 and S12†). In the case of Pd₃/Al₂O₃–NaBH₄(B), the atomic fraction of B in the provided STEM-EDX image was found to be around 0.23% (Table S2†). We believe the large amount of borates still present on the γ-Al₂O₃ support makes it very difficult to visualize small amounts of B incorporated into the Pd NPs (as evidence by EXAFS and NMR). For Pd₃/Al₂O₃–LiBH₄(B), the STEM-EDX maps (Fig. S13†) indicated the random distribution of large Pd NCs with an average size of around 2.6 nm on the γ-Al₂O₃ surface. However, no B signal was detected from this LiBH₄-treated sample, likely due to the low concentration of tetraborate species on the γ-Al₂O₃ support (Table S2†). In the case of Pd₃/Al₂O₃–LiAlH₄(B), the STEM-EDX maps (Fig. S14†) showed a fairly homogeneous distribution of Pd over the γ-Al₂O₃ surface with a 0.35% Pd loading (Table S2†).

Overall, the results from the combination of the aforementioned techniques indicated that all the used reducing agents are capable of activating the Pd₃ NCs on the γ-Al₂O₃ surface by removing phosphine and chloride ligands from the NC surfaces. Treatment with a low amount of NaBH₄ in Pd₃/Al₂O₃–NaBH₄(A) led to the activation of Pd₃ NCs without noticeable size growth, while treatment with higher amounts of NaBH₄ in Pd₃/Al₂O₃–NaBH₄(A/B) resulted in the formation of larger Pd NPs with intercalated B atoms. In the case of LiBH₄-treated samples, i.e., Pd₃/Al₂O₃–LiBH₄(A/B/C), slightly larger NCs were formed depending on the amount used for activation, and no intercalated B species were seen in the resulting Pd NCs. Finally, activation with LiAlH₄ in Pd₃/Al₂O₃–LiAlH₄ (A/B/C) resulted in the formation of a mixture of Pd single-atom sites and very small NCs, and the size and distribution of the Pd particles were found to be independent of the amounts of the LiAlH₄ used for the activation.

To measure the effect of activation of the Pd₃ NCs on their catalytic activity, the transfer hydrogenation of an α,β-unsaturated aldehyde, trans-cinnamaldehyde (CAL), was performed. Selective partial hydrogenation of CAL can yield hydrocinnamaldehyde (HCAL) or cinnamic alcohol (COL), depending on whether the C=C or C=O bonds are hydrogenated, respectively.⁴¹ However, upon complete hydrogenation of both C=C and C=O bonds, hydrocinnamic alcohol (HCOL) can be obtained.⁴¹ The products of this reaction, i.e., HCAL and COL have various applications in the pharmaceutical and fragrance fields.⁴¹ Thermodynamically, hydrogenation of the C=C bond is favoured over the C=O bond due to its lower bond energy.⁴² For this transfer hydrogenation reaction, formic acid and its formates have been selected as the hydrogen donors. Due to increased sustainability and cost-effectiveness, biomass-derived hydrogen donors such as formic acid and related formates have attracted research attention in the field of transfer hydrogenation.⁴³ Formic acid has a very high volumetric capacity for hydrogen storage which can be released by either dehydrogenation or dehydration pathways.⁴⁴ However, one major challenge in using formic acid in high concentrations at the industrial level is the corrosion of the reactors; thus, using neutralized acids (e.g., formates) instead of formic acid is more favourable.⁴³

The reaction conditions for the transfer hydrogenation of CAL were optimized by using Pd₃/Al₂O₃–NaBH₄(A) as a model catalyst at 70 °C for 3 hours (Table 2). The reaction was first examined with formic acid as a hydrogen source and in the absence of a base, which resulted in a very low conversion, i.e., 12.2% (Table 2, entry 1). Upon adding 5 mmol of KOH to the reaction, the conversion increased to 63.4% (Table 2, entry 2). With a decrease in the amount of KOH from 5 mmol to 1 mmol, the conversion dropped by a factor of 2 (Table 2, entry 3). Previously, Lee et al. have shown that the decomposition of formic acid over Pd involves the formation and interaction of HCOO⁻ with the metal center followed by C–H activation via β-hydride elimination reaction (Scheme S1†).⁴⁵ The use of strong bases as the additive can make this process thermodynamically more favourable as the base can facilitate the formation of HCOO⁻ species.⁴⁴ Changing the solvent from MeOH to EtOH did not have much effect on the conversion (Table 2, entry 4). However, upon using water as a solvent, the conversion dropped significantly to 28.9% (Table 2, entry 5). The main reason behind this large drop is likely the poor solubility of CAL in water. With the absence of the base, the catalysts did not show any significant activity in water (Table 2, entry 6). Upon using the NaOH as a base in water, the catalyst showed 25.7% conversion (Table 2, entry 7). However, upon using a comparatively milder base, i.e., K₂CO₃, the conversion further dropped to 11.4% (Table 2, entry 8). At room temperature, no conversion was obtained in water with formic acid as the hydrogen donor and KOH as the base (Table 2, entry 9). Conversely, sodium formate as a hydrogen donor in the absence of a base provided 35.4% conversion in water (Table 2, entry 10), which further increased to 62.8% and 60.8% upon using EtOH and MeOH as the solvent, respectively (Table 2, entries 11 and 12). The reaction did not show any conversion when toluene was used as the solvent, probably due to the poor solubility of sodium formate in toluene (Table 2, entry 13). The best result, i.e., 68.7% conversion, was obtained using a mixture of water and EtOH as the solvent (Table 2, entry 14). However, at room temperature, the same conditions provided only a 3.3% conversion (Table 2, entry 15). This is likely due to the unfavorable conditions for C–H activation via β-hydride elimination at low temperatures (Scheme S1†). Doubling the amount of sodium formate resulted in a slight increase in conversion (Table 2, entry 16). Conversely, reducing the sodium formate to 1 mmol led to a conversion of only 21% (Table 2, entry 17). Finally, ammonium formate as the hydrogen donor gave a conversion rate similar to sodium formate (Table 2, entry 18). Notably, in all of these optimization reactions, no formation of COL was observed, and the selectivity towards HCAL remained consistently high, ranging from 85% to 93% (with HCOL as the other minor product). This suggests that the reaction selectivity remains largely unaffected by varying reaction conditions and is primarily governed by catalyst properties.

Table 2 Optimization of the reaction conditions for the transfer hydrogenation of CAL using Pd₃/Al₂O₃–NaBH₄(A)

Entry	Hydrogen source/amount (mmol)	Solvent	Additive/amount (mmol)	Conversion%
Reaction conditions: 1.0 mmol of CAL, 4.0 ml of solvent, 25.0 mg of Pd3/Al2O3–NaBH4(A) (Pd mol% = 0.11%) at 70 °C for 3 hours.a Reaction was carried out at room temperature. The selectivity for HCAL in all these reactions is between 85–93% with no traceable formation of COL.
1	HCOOH/5	MeOH	—	12.2%
2	HCOOH/5	MeOH	KOH/5	63.4%
3	HCOOH/5	MeOH	KOH/1	31.1%
4	HCOOH/5	EtOH	KOH/5	68.3%
5	HCOOH/5	H2O	KOH/5	28.9%
6	HCOOH/5	H2O	—	5.3%
7	HCOOH/5	H2O	NaOH/5	25.7%
8	HCOOH/5	H2O	K2CO3/5	11.4%
9^a	HCOOH/5	H2O	KOH/5	0%
10	HCOONa/5	H2O	—	35.4%
11	HCOONa/5	EtOH	—	62.8%
12	HCOONa/5	MeOH	—	60.8%
13	HCOONa/5	Toluene	—	0%
14	HCOONa/5	EtOH : H2O (1 : 1)	—	68.7%
15^a	HCOONa/5	EtOH : H2O (1 : 1)	—	3.3%
16	HCOONa/10	EtOH : H2O (1 : 1)	—	73.2%
17	HCOONa/1	EtOH : H2O (1 : 1)	—	21.0%
18	HCOONH4/5	EtOH : H2O (1 : 1)	—	63.9%

---

After identifying the optimum conditions, non-activated Pd₃/Al₂O₃ and all the activated NCs were employed as catalysts in the transfer hydrogenation of CAL using sodium formate as the hydrogen donor in a 1 : 1 H₂O : EtOH solvent (Table 3). Non-activated Pd₃/Al₂O₃ showed a low conversion of 38.3% with a selectivity of around 78.1% towards HCAL. The activated Pd₃/Al₂O₃–NaBH₄(A) catalysts showed improved activity, achieving a conversion rate of 68.7% and an improved selectivity towards HCAL of 90.5%. In the case of Pd₃/Al₂O₃–NaBH₄(B) and Pd₃/Al₂O₃–NaBH₄(C), the obtained conversions were very high (98.3% and 95.0%, respectively), however, these catalysts showed much lower selectivity towards HCAL and a higher proportion of the completely hydrogenated products, i.e., HCOL, formed. This result is consistent with the larger size of the Pd NPs under these activation conditions, and the higher activity may also be enhanced due to B-incorporation into the Pd NPs. Previously, it has been shown in the literature that the intercalation of B inside the Pd lattice can boost their efficiency in hydrogen generation from formic acid.^46–48 As these samples also contained the B intercalated Pd particles, their higher catalytic activity for the transfer hydrogenation of CAL is likely associated with their enhanced capability of extracting hydrogen from formate species during catalysis. Pd₃/Al₂O₃–LiBH₄(A/B/C) catalysts demonstrated reasonable conversions between 63–73% coupled with high selectivity of around 90% towards HCAL. This is consistent with the moderate NC sizes seen after LiBH₄ activation (1.7–2.5 nm) as evidenced by EXAFS and HAADF-STEM analyses. In contrast, Pd₃/Al₂O₃–LiAlH₄(A/B/C) catalysts showed slightly lower conversions, ranging from 49.6–61.1%, with selectivities towards HCAL exceeding 90%. Based on EXAFS and HAADF-STEM results, these catalysts consist of a combination of single-atom sites and small NCs ranging in size from 1.1 to 1.5 nm. Consequently, due to the small size of the Pd NCs, these catalysts exhibited lower catalytic activity but higher selectivity towards HCAL. To compare the catalytic performance of the chemically activated NCs with thermally activated NCs, Pd₃/Al₂O₃ was treated at 150 °C for 3 hours to obtain Pd₃/Al₂O₃-150.¹⁶ Previously, it has been found that after thermal activation at 150 °C, phosphines on the Pd₃ NCs can be effectively removed without any increase in NC size on activated carbon supports; but the resulting Pd₃ NCs have significant Pd–Cl interactions.¹⁶ In the hydrogenation of CAL, Pd₃/Al₂O₃-150 exhibited a relatively lower conversion of 37.2% along with a very high selectivity towards HCAL. The lower activity of the thermally activated sample may be due to the presence of Cl ligands on the surface of NCs after thermal treatment.¹⁶

Table 3 Catalytic performance of the activated catalysts for the transfer hydrogenation of CAL

Catalyst	Conversion, %	Selectivity, %
		HCAL	COL	HCOL
Reaction conditions: 1.0 mmol of CAL, 5.0 mmol of HCOONa, 4 ml of solvent (1 : 1 H2O : EtOH), 25.0 mg of catalyst (Pd mol% = 0.11%) at 70 °C for 3 hours.
Pd3/Al2O3	38.3%	78.1%	≤0.1%	21.9%
Pd3/Al2O3–NaBH4(A)	68.7%	90.5%	≤0.1%	9.4%
Pd3/Al2O3–NaBH4(B)	98.3%	71.2%	≤0.1%	28.7%
Pd3/Al2O3–NaBH4(C)	95.0%	67.1%	≤0.1%	32.9%
Pd3/Al2O3–LiBH4(A)	69.3%	91.2%	≤0.1%	8.7%
Pd3/Al2O3–LiBH4(B)	72.3%	89.9%	≤0.1%	9.9%
Pd3/Al2O3–LiBH4(C)	63.3%	93.2%	≤0.1%	6.8%
Pd3/Al2O3–LiAlH4(A)	57.3%	91.3%	≤0.1%	8.6%
Pd3/Al2O3–LiAlH4(B)	49.6%	90.0%	≤0.1%	9.9%
Pd3/Al2O3–LiAlH4(C)	61.1%	94.0%	≤0.1%	5.8%
Pd3/Al2O3-150	37.2%	96.0%	≤0.1%	4.0%

---

None of the catalysts exhibited significant formation of COL, consistent with previous work on Pd-based nano-sized catalysts.^41,49,50 In the hydrogenation of CAL, the Pd-based catalysts are mainly selective towards HCAL, however, other noble metals such as Pt, Au, and Ir are more selective towards the formation of COL.^41,49,50 This implies that the η⁴ adsorption mode of CAL is preferred over the Pd surface of these catalysts, which leads to the hydrogenation of the C=C bond followed by the hydrogenation of the C=O bond.⁴¹ Previously, Liu and co-workers demonstrated through DFT calculations that the adsorption energy of the C=C bond on Pd₄ NCs is lower than that of the C=O bond, whereas, larger Pd particles with Pd(111) surfaces exhibit stronger C=O bond adsorption.⁵¹ Thus, smaller Pd NCs are more selective for the conversion of CAL to HCAL. In recent work on di-1-adamantylphosphine-stablized Pd_n (n = 2–5) NCs supported on carbon,⁵² Tang et al. noted that the Pd₅ NCs exhibited high selectivity (95.3%) towards HCAL formation at a high conversion in the hydrogenation of CAL. Conversely, Pd₂ NCs analogs showed significantly lower activity but maintained a high selectivity of 99.9% towards HCAL, while Pd₃ NCs achieved complete conversion but a lower selectivity (80.5%) towards HCAL.⁵² These results indicated that changes in both the particle size and electronic structure of the Pd_n NCs significantly influenced their activity and selectivity.^51,52 Recent work by Li et al. showed that the carbon-nitride-nanosheet-supported Pd NCs along with Pd single-atom sites led to a complete conversion in the hydrogenation of CAL with a 97.3% selectivity towards HCAL.⁵³ Using DFT calculations, a synergistic effect between single atom sites and NCs was found, where the CAL is preferentially adsorbed on Pd single-atom sites and hydrogen adsorbed on NCs.⁵³ In the case of Pd₃/Al₂O₃–LiAlH₄(A/B/C), a mix of NCs and atomically dispersed Pd sites were observed, however, the conversions for these catalysts were somewhat lower compared to the reported results for this similar system.⁵³ This may be due to a non-uniform distribution of the Pd sites on the γ-Al₂O₃ support in comparison to the homogenous distribution of Pd sites on g-C₃N₄.⁵³ Overall the catalytic results in this manuscript showed that the post-activation evolution of NCs has a direct effect on the activity and selectivity of the catalysts. Larger Pd NPs have shown higher activity for the transfer hydrogenation reactions, likely due to the presence of well-faceted Pd(111) surfaces and the presence of intercalated B species in the Pd NPs. However, the presence of Pd(111) surfaces also led to strong adsorption of the substrate on the metal surface resulting in the increased formation of a completely hydrogenated product.⁵⁴ However, in the case of small-size NCs and single-atom sites, the activity has been found lower due to weaker interactions of the catalyst with the substrates, which leads to higher selectivity toward the partially hydrogenated product, HCAL.

Conclusions

This work demonstrates the chemical activation of the γ-Al₂O₃-supported Pd₃ NCs using various hydride-based reducing agents for transfer hydrogenation reactions. Different amounts of NaBH₄, LiBH₄, and LiAlH₄ were used for activation, and the post-activation structural evolution of the NCs was monitored using a combination of various techniques. XAS results indicate that all reducing agents effectively removed surface phosphine and chloride ligands from Pd₃ NCs. Both XAS and HAADF-STEM analyses showed that lower amounts of NaBH₄ activated the Pd₃ NCs without any noticeable size growth, while higher NaBH₄ amounts caused a transformation of Pd NCs into Pd NPs. XPS, solid-state NMR, and STEM-EDX analyses revealed the presence of tetraborate and phosphine impurities on the surface of γ-Al₂O₃, even after post-activation washings in the NaBH₄-treated samples. Additionally, solid-state NMR and EXAFS analysis suggested the intercalation of B atoms within the lattice of formed Pd NPs for samples activated with high NaBH₄ loadings. Activation with LiBH₄ resulted in a slight increase in the size of the Pd NCs, producing NCs ranging from 1.7 to 2.5 nm in size. Activation with LiAlH₄ converted Pd₃ NCs into a mixture of single-atom sites and small NCs around 1.4 nm in size. The post-activation evolution of Pd₃ NCs directly impacted their catalytic behavior in the transfer hydrogenation of CAL. Smaller Pd NCs were highly selective towards the formation of HCAL, the partial hydrogenation product, whereas larger Pd NPs were more active catalysts, but also increased the complete hydrogenation of CAL to HCOL. Overall, our results demonstrate that immobilizing atom-precise Pd₃ NCs onto an γ-Al₂O₃ support followed by chemical activation is an effective method to obtain supported Pd NCs for heterogeneous catalysis. This approach offers greater control over the structure and speciation of the Pd NCs compared to traditional thermal activation methods.

Author contributions

All authors have approved the final version of the manuscript. S. S. and R. W. J. S. contributed to the initial concept development, experimental design, characterization of catalysts, analysis of results, and writing of the manuscript. D. W. carried out catalytic studies under the supervision of S. S. HR-TEM, HAADF-STEM and STEM-EDX analysis has been carried out by K. E. S. under the supervision of D. H.

Data availability

The data supporting this article have been included as part of the ESI.†

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

We acknowledge financial assistance from the National Sciences and Engineering Research Council of Canada. We thank Ning Chen and Weifeng Chen at the HXMA beamline (Canadian Light Source) for their assistance in XAS measurements. XAS experiments were performed at the Canadian Light Source, a national research facility of the University of Saskatchewan, which is supported by the Canada Foundation for Innovation, the National Sciences and Engineering Research Council of Canada, the National Research Council, the Canadian Institutes of Health Research, the Government of Saskatchewan, and the University of Saskatchewan. The Saskatchewan Structural Sciences Centre (SSSC) is acknowledged for providing NMR and XPS facilities to conduct this research. Sarah Purdy and Jianfeng Zhu at SSSC are acknowledged for their assistance with XPS and solid-state NMR analysis, respectively. Funding from the Canada Foundation for Innovation, Natural Sciences and Engineering Research Council of Canada, and the University of Saskatchewan support research at the SSSC. Electron microscopy was performed at the Canadian Centre for Electron Microscopy at McMaster University.

References

1. C. Dong, Z. Gao, Y. Li, M. Peng, M. Wang, Y. Xu, C. Li, M. Xu, Y. Deng, X. Qin and F. Huang, Nat. Catal., 2022, 5, 485–493.
2. F. Huang, Y. Deng, Y. Chen, X. Cai, M. Peng, Z. Jia, P. Ren, D. Xiao, X. Wen, N. Wang and H. Liu, J. Am. Chem. Soc., 2018, 140, 13142–13146.
3. C. Dong, Y. Li, D. Cheng, M. Zhang, J. Liu, Y. G. Wang, D. Xiao and D. Ma, ACS Catal., 2020, 10, 11011–11045.
4. I. Cano, A. Weilhard, C. Martin, J. Pinto, R. W. Lodge, A. R. Santos, G. A. Rance, E. H. Åhlgren, E. Jónsson, J. Yuan and Z. Y. Li, Nat. Commun., 2021, 12, 4965.
5. X. Li, S. Mitchell, Y. Fang, J. Li, J. Perez-Ramirez and J. Lu, Nat. Rev. Chem., 2023, 7, 754–767.
6. N. Jeddi, N. W. J. Scott, T. Tanner, S. K. Beaumont and I. J. S. Fairlamb, Chem. Sci., 2024, 15, 2763–2777.
7. T. I. Levchenko, H. Yi, M. D. Aloisio, N. K. Dang, G. Gao, S. Sharma, C. Dinh and C. M. Crudden, ACS Catal., 2024, 14, 4155–4163.
8. D. Yazaki, T. Kawawaki, D. Hirayama, M. Kawachi, K. Kato, S. Oguchi, Y. Yamaguchi, S. Kikkawa, Y. Ueki, S. Hossain, D. J. Osborn, F. Ozaki, S. Tanaka, J. Yoshinobu, G. F. Metha, S. Yamazoe, A. Kudo, A. Yamakata and Y. Negishi, Small, 2023, 19, 2208287.
9. R. Jin, C. Zeng, M. Zhou and Y. Chen, Chem. Rev., 2016, 116, 10346–10413.
10. V. Sudheeshkumar, K. O. Sulaiman and R. W. J. Scott, Nanoscale Adv., 2020, 2, 55–69.
11. K. K. Li, C. Wu, T. Yang, D. Qin and Y. Xia, Acc. Chem. Res., 2023, 56, 1517–1527.
12. T. Kawawaki, Y. Kataoka, M. Hirata, Y. Iwamatsu, S. Hossain and Y. Negishi, Nanoscale Horiz., 2021, 6, 409–448.
13. N. W. J. Scott, M. J. Ford, N. Jeddi, A. Eyles, L. Simon, A. C. Whitwood, T. Tanner, C. E. Willans and I. J. S. Fairlamb, J. Am. Chem. Soc., 2021, 143, 9682–9693.
14. N. W. J. Scott, M. J. Ford, C. Schotes, R. R. Parker, A. C. Whitwood and I. J. S. Fairlamb, Chem. Sci., 2019, 10, 7898–7906.
15. Y. Yun, H. Sheng, J. Yu, L. Bao, Y. Du, F. Xu, H. Yu, P. Li and M. Zhu, Adv. Synth. Catal., 2018, 360, 4731–4743.
16. S. Singh, K. O. Sulaiman, Mahwar and R. W. J. Scott, Catal. Sci. Technol., 2024, 14, 322–333.
17. K. O. Sulaiman, M. Zubair, G. King, N. Bedford and R. W. J. Scott, Phys. Chem. Chem. Phys., 2022, 24, 24834–24844.
18. K. O. Sulaiman, V. Sudheeshkumar and R. W. J. Scott, RSC Adv., 2019, 9, 28019–28027.
19. Y. Liu, H. Tsunoyama, T. Akitab and T. Tsukuda, Chem. Commun., 2010, 46, 550–552.
20. V. Sudheeshkumar, M. Lavier, B. Lukan, J. Shen, N. Semagina and R. W. J. Scott, Catal. Today, 2023, 407, 223–229.
21. J. Kilmartin, R. Sarip, R. Grau-Crespo, D. D. Tommaso, G. Hogarth, C. Prestipino and G. Sankar, ACS Catal., 2012, 2, 957–963.
22. G. Collins, F. Davitt, C. Dwyer and J. D. Holmes, ACS Appl. Nano Mater., 2018, 1, 7129–7138.
23. S. Das, A. Goswami, M. Hesari, J. F. Al-Sharab, E. Mikmeková, F. Maran and T. Asefa, Small, 2014, 10, 1473–1478.
24. M. Dasog, W. Hou and R. W. J. Scott, Chem. Commun., 2011, 47, 8569–8571.
25. A. Shivhare and R. W. J. Scott, RSC Adv., 2016, 6, 62579–62584.
26. S. M. Ansar, F. S. Ameer, W. Hu, S. Zou, C. U. Pittman and D. Zhang, Nano Lett., 2013, 13, 1226–1229.
27. F. Fu, J. Xiang, H. Cheng, L. Cheng, H. Chong, S. Wang, P. Li, S. Wei, M. Zhu and Y. Li, ACS Catal., 2017, 7, 1860–1867.
28. T. J. Penfold, I. Tavernelli, C. J. Milne, M. Reinhard, A. E. Nahhas, R. Abela, U. Rothlisberger and M. Chergui, J. Chem. Phys., 2013, 138, 014104.
29. J. Wang, Y. Kang, H. Yang and W. Cai, J. Phys. Chem. C, 2009, 113, 8366–8372.
30. J. Li, J. Chen, Q. Wang, W. Cai and S. Chen, Chem. Mater., 2017, 29(23), 10060–10067.
31. C. W. A. Chan, A. H. Mahadi, M. M. Li, E. C. Corbos, C. Tang, G. Jones, W. C. H. Kuo, J. Cookson, C. M. Brown, P. T. Bishop and S. C. E. Tsang, Nat. Commun., 2014, 5, 5787.
32. A. Leineweber, T. G. Berger, A. Udyansky and V. N. Bugaev, Z. Kristallogr. – Cryst. Mater., 2014, 229, 353–367.
33. T. T. Zhuang, D. H. Nam, Z. Wang, H. H. Li, C. M. Gabardo, Y. Li, Z. Q. Liang, J. Li, X. J. Liu, B. Chen and W. R. Leow, Nat. Commun., 2019, 10, 4807.
34. A. P. Deziel, S. Gahlawat, N. Hazari, K. H. Hopmann and B. Q. Mercado, Chem. Sci., 2023, 14, 8164–8179.
35. B. Zhang, A. Sels, G. Salassa, S. Pollitt, V. Truttmann, C. Rameshan, J. Llorca, W. Olszewski, G. Rupprechter, T. Bürgi and N. Barrabés, ChemCatChem, 2018, 10, 5372–5376.
36. C. D. Wagner, A. V. Naumkin, A. Kraut-Vass, J. W. Allison, C. J. Powell and J. R. Rumble Jr., NIST Standard Reference Database 20, Version 3.4 (web version), 2003 (https://srdata.nist.gov/xps/).
37. X. Yang, D. Kim, D. Um, G. Kim and C. Kim, J. Vac. Sci. Technol., A, 2009, 27, 821–825.
38. C. Su, M. Lu, S. Wang and Y. Huang, RSC Adv., 2012, 2, 2073–2079.
39. B. R. Strohmeier, Surf. Sci. Spectra, 1994, 3, 135–140.
40. W. R. Gunther, Y. Wang, Y. Ji, V. K. Michaelis, S. T. Hunt, R. G. Griffin and Y. Román-Leshkov, Nat. Commun., 2012, 3, 1109.
41. X. Wang, X. Liang, P. Geng and Q. Li, ACS Catal., 2020, 10, 2395–2412.
42. P. Gallezot and D. Richard, Catal. Rev.: Sci. Eng., 1998, 40, 81–126.
43. R. Nie, Y. Tao, Y. Nie, T. Lu, J. Wang, Y. Zhang, X. Lu and C. C. Xu, ACS Catal., 2021, 11, 1071–1095.
44. A. K. Singh, S. Singh and A. Kumar, Catal. Sci. Technol., 2016, 6, 12–40.
45. W. J. Lee, Y. J. Hwang, J. Kim, H. Jeong and C. W. Yoon, ChemPhysChem, 2019, 20, 1382–1391.
46. Z. Li, J. Xu, F. Meng, K. Yang and D. Lin, ACS Appl. Mater. Interfaces, 2022, 14, 30735–30745.
47. J. S. Yoo, Z. Zhao, J. K. Nørskov and F. Studt, ACS Catal., 2015, 5, 6579–6586.
48. K. Jiang, K. Xu, S. Zou and W. Cai, J. Am. Chem. Soc., 2014, 136, 4861–4864.
49. S. Fujiwara, N. Takanashi, R. Nishiyabua and Y. Kubo, Green Chem., 2014, 16, 3230–3236.
50. Y. Zhao, M. Liu, B. Fan, Y. Chen, W. Lv, N. Lu and R. Li, Catal. Commun., 2014, 57, 119–123.
51. F. Jiang, J. Cai, B. Liu, Y. Xu and X. Liu, RSC Adv., 2016, 6, 75541–75551.
52. J. Tang, T. Ge, W. Wang, C. Liu and J. Huang, Chem. Commun., 2023, 59, 8854–8857.
53. X. Li, J. Liu, J. Wu, L. Zhang, D. Cao and D. Cheng, ACS Catal., 2024, 14, 2369–2379.
54. X. Zhao, Y. Chang, W. Chen, Q. Wu, X. Pan, K. Chen and B. Weng, ACS Omega, 2022, 7, 17–31.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4nr03364g

---