Substrate‐Dependent Selectivity in Alkyne Semihydrogenation Over a Hydrogen‐Competent Pd₃Sn₂ Intermetallic Catalyst

ABSTRACT

A recurring challenge in heterogeneous hydrogenation is to maintain rapid H₂ activation while preventing the product from undergoing overhydrogenation. Site isolation can weaken adsorption but frequently compromises hydrogen competence, and selectivity is often presumed transferable across closely related substrates. Here we establish a substrate‐dependent selectivity framework on Pd–Sn intermetallic catalysts by integrating theory‐guided site identification with experimental validation. Theoretical calculations reveal that Pd₃Sn₂ hosts a distinctive surface structure in which Sn‐bridged Pd–Pd dual sites cooperate with adjacent near‐surface Pd to dissociate H₂ readily, while methyl substitution in C₃ intermediates weakens π‐binding and kinetically favors propylene desorption over further hydrogenation. Guided by these predictions, phase‐pure Pd₃Sn, Pd₃Sn₂, and PdSn₂ intermetallic catalysts with comparable particle sizes were synthesized and verified by comprehensive characterizations including aberration‐corrected electron microscopy and x‐ray absorption spectroscopy. Under excess‐propylene conditions, the Pd₃Sn₂ catalyst achieves 98.0% propylene selectivity at 100% propyne conversion, whereas the Pd and Pd₃Sn catalysts suffer severe overhydrogenation and the PdSn₂ catalyst is intrinsically sluggish. Notably, the Pd₃Sn₂ catalyst performs poorly for acetylene hydrogenation, leading to significant ethane formation via overhydrogenation of both newly formed and co‐fed ethylene. Complementary kinetic evidence supports the proposed mechanism by quantifying hydrogen activation competence and product residence on the intermetallic surfaces.

Substrate‐dependent selectivity on Pd–Sn intermetallics is governed by the synergy between H₂ activation sites and substrate adsorption configurations. Theoretical and experimental evidence shows that Pd₃Sn₂ promotes propyne semihydrogenation through specialized dual sites but exhibits markedly different behavior toward acetylene. This work underscores the necessity of considering substrate properties, rather than only hydrogen activation ability, when designing selective hydrogenation catalysts.

1. Introduction

Heterogeneous catalysis increasingly relies on the ability to engineer specific active‐site structures that deliver not only high activity but also decisive control over selectivity across complex elementary‐step networks [1, 2, 3, 4, 5]. In many reactions, the desired product is not determined by a single barrier but by a branching competition among adsorption, surface reaction, diffusion, and desorption steps [6, 7, 8]. As a consequence, catalytic performance is governed by a delicate coupling among local site geometry, electronic structure, and adsorbate‐induced dynamics [9]. Conceptually, this has driven a transition from viewing catalysts as uniform surfaces to treating them as site‐defined materials whose function emerges from the cooperation of surface atoms, subsurface constituents and dynamic adsorbate coverages [10, 11]. This challenge is particularly evident in selective hydrogenations, where the catalyst must balance fast H₂ activation with controlled hydrogen delivery and moderated binding of intermediates and products [12, 13]. For Pd‐based catalysts, the inherent ability to activate H₂ and hydrogenate unsaturated bonds often comes with a decrease in selectivity due to strong adsorption and facile over‐hydrogenation, as well as the sensitivity of Pd surfaces to subsurface species and hydride/carbide formation under reaction conditions [14, 15, 16]. These considerations have attracted broad interest in programmable active site structures that can decouple reactant activation from product desorption, thereby preventing rapid hydrogenation steps from cascading into undesirable products.

Atomic‐scale site engineering has therefore emerged as a powerful strategy. Concepts such as ensemble control [15, 17, 18, 19], ligand/strain tuning [20, 21, 22], and isolated active‐site structures [23, 24, 25] can suppress unselective pathways while maintaining essential activation chemistry. In this context, ordered intermetallic compounds [13, 26, 27] and single‐atom alloys (SAAs) [19, 28] provide well‐defined atomic arrangements that stabilize favorable local coordination environments, offering a route to systematically break extended Pd–Pd ensembles without resorting to surface modifiers. Intermetallic compounds have already delivered notable successes in alkyne semihydrogenation by isolating Pd sites and stabilizing selective surface states [12, 13, 27]. Yet the advantage of site isolation in suppressing strong adsorption can also compromise H₂ activation, since H₂ dissociation on late transition metals often benefits from multiatom ensembles or specific geometric arrangements [18, 29]. Selective hydrogenation therefore demands not only weaker binding of reactive intermediates and products but also a robust pathway for H₂ activation [30, 31, 32]. These considerations point to a general design objective, namely, to construct catalytic active sites that remain hydrogen‐competent while preventing the product from lingering on the surface long enough to undergo over‐hydrogenation [33, 34, 35]. Equally important, and often underappreciated, is substrate specificity [36, 37, 38]. Selectivity is frequently treated as an intrinsic property of a catalyst surface, implicitly assuming that closely related substrates will behave similarly once the activation function is enabled [39]. In reality, intermediate bonding can be exquisitely sensitive to subtle molecular structure, including substitution and steric effects, which can modulate adsorption strength, and geometry, thereby altering the fate of the alkene intermediate, namely, whether it desorbs or undergoes subsequent hydrogenation [40]. A general, experimentally validated framework that links molecular substitution to surface bonding, intermediate fate, and macroscopic selectivity on intermetallic surfaces is still lacking.

Here, we address this gap by employing Pd–Sn intermetallic compounds as a model platform to investigate substrate‐dependent selectivity in C₂/C₃ alkyne semihydrogenation, a representative branching reaction network in which the fate of intermediates is decisive. First‐principles calculations were performed to map phase‐dependent site connectivity on Pd₃Sn, Pd₃Sn₂, and PdSn₂, revealing that Pd₃Sn₂ retains efficient hydrogen activation through surface–subsurface cooperation at Sn‐bridged Pd–Pd dual sites, while simultaneously promoting propylene desorption over further hydrogenation. The calculations further predict a pronounced selectivity divergence between C₂ and C₃ alkynes on Pd₃Sn₂, as methyl substitution weakens alkene π‐binding and shifts the dominant fate of the alkene intermediate from further hydrogenation to desorption. Guided by these insights, phase‐pure Pd₃Sn, Pd₃Sn₂, and PdSn₂ with comparable particle sizes were synthesized, and their ordered structures and Pd–Sn coordination environments were verified by aberration‐corrected electron microscopy and x‐ray absorption spectroscopy. Catalytic testing shows that Pd₃Sn₂ achieves 98.0% propylene selectivity at near‐complete propyne conversion under industrially relevant conditions, whereas it performs poorly in acetylene semihydrogenation, with facile overhydrogenation leading to substantial ethane formation. Mechanistic studies, including H₂‐D₂ exchange, temperature‐programmed experiments, and kinetic analysis, further corroborate the theoretical picture by linking hydrogen activation, alkene residence time, and overhydrogenation propensity to phase‐defined Pd–Sn site structures. This work demonstrates that efficient H₂ activation is necessary but not sufficient for selective hydrogenation, and that subtle substrate structural differences can reverse selectivity on a hydrogen‐competent intermetallic surface by governing intermediate binding geometry and residence time.

2. Results and Discussion

Theoretical calculations were performed to clarify how the Pd₃Sn₂ surface architecture couples near‐surface Pd atoms with Sn‐bridged Pd–Pd dual sites. Wulff constructions predict distinct equilibrium morphologies for monometallic Pd and the intermetallic Pd₃Sn, Pd₃Sn₂, and PdSn₂ phases (Figures S1–S2). The (111), (111), (102), and (024) facets are identified as the thermodynamically stable and predominantly exposed surfaces for $Fm\overline{3}m$ Pd, $Pm\overline{3}m$ Pd₃Sn, P6₃/mmc Pd₃Sn₂, and Aba₂ PdSn₂, respectively. Accordingly, the Pd(111), Pd₃Sn(111), Pd₃Sn₂(102), and PdSn₂(024) surfaces were selected as representative models for subsequent DFT calculations (Figure S2) [30, 33, 34, 39, 41]. On the Pd₃Sn₂(102) surface, the Sn‐bridged surface Pd–Pd dual site is characterized by a distance of 4.390 Å (Figure S2), while a near‐surface Pd atom is highly close to one Pd center of the dual site, with a distance of 2.800 Å. This well‐defined geometric configuration demonstrates the isolated Pd–Pd dual‐site coupled with near‐surface Pd atoms on the Pd₃Sn₂(102) surface, which is absent in monometallic Pd and in the Pd₃Sn and PdSn₂ phases. On the Pd₃Sn₂(102) surface, hydrogen molecules spontaneously dissociate at the bridge site between the Pd atom of the Sn‐bridged Pd–Pd dual site and the neighboring near‐surface Pd atom, releasing 0.16 eV of energy (Figure 1a). Similar facile H₂ activation is observed on the Pd(111) and Pd₃Sn(111) surfaces, where contiguous surface Pd ensembles are available (Figures 1a and S3). In contrast, H₂ dissociation on the PdSn₂(024) surface without the presence of near‐surface Pd sites needs to overcome a free energy barrier of 0.65 eV. These results highlight the synergistic role of the near‐surface Pd atom and the Sn‐bridged Pd–Pd dual site in enabling efficient hydrogen activation on the Pd₃Sn₂ surface.

FIGURE 1.

(a) Free energy profiles of H₂ dissociation into H atoms on the Pd₃Sn(111), Pd₃Sn₂(102), and PdSn₂(024) surfaces. Surface Pd, near‐surface Pd, Sn, C, and H atoms are depicted as gold, bright yellow, cyan, black, and white spheres, respectively. Adsorption configurations of C₂H₄ (b) and C₃H₆ (c) on the Pd₃Sn₂(102) surface, with insets showing Pd─C bond lengths, and ─pCOHP analyses of the interactions between Pd sites and bonded carbon atoms in C₂H₄ (d) and C₃H₆ (e), where positive and negative ─pCOHP values correspond to bonding and antibonding contributions, respectively. (f) Free energy profiles for propyne hydrogenation on Pd₃Sn(111), Pd₃Sn₂(102), and PdSn₂(024) surfaces. The optimized transition state structures corresponding to key hydrogen diffusion and hydrogenation steps are displayed above the energy profiles for the Pd₃Sn(111) surface and below those for the Pd₃Sn₂(102) surface. The numbers along the reaction pathway indicate the free energy barriers of the fundamental reaction steps. (g) Free energy barriers associated with H₂ activation (H₂* → 2H*), propyne hydrogenation to propylene (C₃H₄* → C₃H₆*), propylene desorption (C₃H₆* → C₃H₆(g)), and propylene hydrogenation (C₃H₆* → C₃H₇*) on the Pd₃Sn(111), Pd₃Sn₂(102), and PdSn₂(024) surfaces.

Although H₂ activation is facile on Pd₃Sn₂(102), the selectivity ultimately depends on whether the alkene intermediate desorbs rapidly enough to avoid consecutive hydrogenation. On the Pd₃Sn₂(102) surface, ethylene binds at the Pd center of the Sn‐bridged Pd–Pd dual site via π‐coordination (Figure S4). The desorption energy barrier for ethylene is determined to be 0.77 eV, higher than 0.51 eV of the barrier for ethylene hydrogenation (Figure S5), indicating that ethylene is prone to sequential hydrogenation to ethane [42, 43, 44, 45, 46, 47]. Bader charge analysis supports strong metal‐adsorbate coupling, with an interfacial charge transfer of 0.06 e (Figure S6). In striking contrast, Pd₃Sn₂ shows strong steric discrimination toward propylene. The allylic methyl group introduces strong Pauli repulsion against the Sn‐bridged Pd–Pd environment [48], reducing adsorption energy to −0.25 eV and resulting in negligible charge transfer (Figures S7–S8), which suggests facile desorption of propylene. Structural metrics combined with integrated crystal orbital Hamilton population (ICOHP) analysis further rationalize the weakened bonding (Figure 1b,e). In contrast to the nearly symmetric Pd‐C coordination observed for adsorbed ethylene (2.278 and 2.280 Å; 1.91 and 1.89 eV), the bulky methyl substituent in propylene induces significant elongation of the Pd─C bond to the α‐carbon (2.431 Å). This spatial mismatch destabilizes the local bonding orbitals, as evidenced by the reduced bond energy (1.44 vs. 1.91 eV). These results indicate that Pd₃Sn₂ leverages site‐specific steric constraints to destabilize the π‐adsorbed propylene intermediate and facilitate its rapid desorption, which favors propylene selectivity. By comparison, although the Pd(111) and Pd₃Sn(111) readily dissociate H₂, their contiguous Pd sites also stabilize propylene through excessive di‐σ bonding and large interfacial charge transfer (Figures S9–S12), which increases the propensity for overhydrogenation.

In addition to promoting alkene desorption, the unique structure of Pd₃Sn₂(102) also provides a spatial decoupling that helps balance reactant activation. Specifically, propyne preferentially adsorbs on the Sn‐bridged Pd‐Pd dual sites via a di‐σ coordination mode, giving a moderate adsorption free energy of −0.67 eV (Figure S13). In contrast, propyne binds much more strongly on Pd(111) and Pd₃Sn(111) through σ‐coordination, with adsorption free energies of −1.87 and −1.15 eV, respectively (Figures S14–S15). On PdSn₂(024), the adsorption of propyne is substantially weaker, characterized by an adsorption free energy of only −0.11 eV (Figure S16), which is unfavorable for effective hydrogenation. Bader charge analysis reveals that the charge transfer associated with di‐σ adsorbed propyne on Pd₃Sn₂(102) is 0.12 e, slightly lower than those on Pd(111) (0.18 e) and Pd₃Sn(111) (0.15 e), but markedly higher than that on PdSn₂(024) (0.02 e) (Figure S17). Together with the H₂ activation mode identified above, these results indicate that Pd₃Sn₂(102) spatially decouples the preferred adsorption site for propyne (Sn‐bridged Pd‐Pd dual site) from the H₂ activation event, which proceeds at a bridge involving the dual‐site Pd and a neighboring near‐surface Pd atom. Such decoupling maintains sufficient propyne affinity for activation while avoiding overly strong binding that can lead to excessive surface coverage and increased likelihood of C─C coupling side reactions.

Guided by these distinctive adsorption characteristics, the stepwise hydrogenation of propyne was subsequently examined. On the Pd₃Sn₂(102) surface, the calculated free energy barriers for the initial two hydrogenation steps (C₃H₄* → C₃H₅* → C₃H₆*) are 0.76 and 0.61 eV, with corresponding exothermicities of 0.21 and 0.77 eV, respectively (Figures 1f and S18), suggesting that propyne can be efficiently hydrogenated on this surface. Furthermore, the effective free energy barrier (${\mathrm{G}}_{\mathrm{AH}}^{\mathrm{eff}}$), accounting for hydrogen migration to neighboring C₃ species during subsequent hydrogenation, was calculated to assess and compare the intrinsic hydrogenation activities across different catalysts. Additional details regarding the calculation of ${\mathrm{G}}_{\mathrm{AH}}^{\mathrm{eff}}$ are provided in Figure 1f,g, Figures S18–S22, and Table S1. The corresponding ${\mathrm{G}}_{\mathrm{AH}}^{\mathrm{eff}}$ values are 0.38, 0.51, 0.53, and 0.93 eV for Pd(111), Pd₃Sn(111), Pd₃Sn₂(102), and PdSn₂(024), respectively, corresponding to an activity trend of PdSn₂< Pd₃Sn₂ ≈ Pd₃Sn < Pd [34]. These results indicate that Pd₃Sn₂ exhibits hydrogenation kinetics comparable to Pd₃Sn but moderately inferior to monometallic Pd. Notably, the free‐energy barrier for further hydrogenation of propylene (C₃H₆* → C₃H₇*) on the Pd₃Sn₂(102) surface is 0.95 eV, which is significantly higher than the propylene desorption free energy of 0.25 eV (Figure 1g), indicating that propylene desorption is kinetically favored over its further hydrogenation on this catalyst. A similar preference for propylene desorption is also obtained on the PdSn₂(024) surface (Figures 1f,g and S19). In contrast, on Pd₃Sn(111), the hydrogenation barrier (0.45 eV) only slightly exceeds the desorption barrier (0.40 eV) (Figures 1f,g and S20), whereas on Pd(111), the hydrogenation and desorption barriers are almost identical, differing by only 0.01 eV (Figures S21–S22), indicating a higher tendency toward overhydrogenation. These results highlight that the Pd₃Sn₂ simultaneously enables facile H₂ activation via cooperation between surface and near‐surface sites and favors propylene desorption over further hydrogenation, thereby alleviating the conventional activity‐selectivity trade‐off in propyne semihydrogenation.

To access such ordered Pd–Sn atomic arrangements experimentally, we designed a sequential ion‐exchange and in situ coating route, as illustrated in Figure 2. Sn(OH)₄@Pd/Mg/Al‐layered double hydroxides (LDHs) were thereby constructed as the key precursor for the subsequent formation of ordered Pd–Sn intermetallic structures. Pd precursors were first introduced into Mg/Al‐LDHs via anion exchange of interlayer NO₃ ⁻ species with PdCl₄ ²⁻, yielding Pd/Mg/Al‐LDHs with uniformly distributed Pd centers confined within the LDH galleries, which maintains atomic‐scale dispersion during the subsequent coating process [15, 18]. A conformal Sn(OH)₄ shell was then grown in situ on the surface of the Pd/Mg/Al‐LDHs through heterogeneous nucleation directed by the abundant hydroxyl groups on the LDH nanosheets, forming a well‐defined core‐shell structure with intimate contact between the Pd‐containing LDH core and the Sn‐rich outer layer [49, 50]. This spatially organized precursor significantly shortens the Pd–Sn diffusion distance and creates a localized binary environment favorable for interfacial atomic mixing [1, 51]. Upon subsequent hydrogen reduction at 900°C, both Pd²⁺ and Sn⁴⁺ species were fully converted into metallic states, while the LDH precursor simultaneously transformed into a rigid MgO/MgAl₂O₄ mixed‐oxide scaffold that suppresses nanoparticle sintering. The intimate atomic contact at the Pd–Sn interface, together with the spatial confinement imposed by the evolving oxide matrix, promotes strong electronic coupling, rapid atomic interdiffusion, and lattice interpenetration, which facilitates the reorganization of Pd–Sn structural units into the ordered Pd₃Sn₂ intermetallic structure [52]. For comparison, monometallic Pd and intermetallic Pd₃Sn and PdSn₂ catalysts were prepared following similar procedures. X‐ray diffraction (XRD), scanning electron microscopy (SEM) combined with energy‐dispersive x‐ray spectroscopy (EDS) and thermogravimetric analysis (TGA) were first employed to confirm the successful formation of the LDH precursors (Figures S23–S32). After H₂ reduction at 900°C, the crystalline structures of Pd, Pd₃Sn, Pd₃Sn₂, and PdSn₂ catalysts were further examined by XRD (Figure 3a). In addition to the support reflections assignable to the spinel‐like MgAl₂O₄ and cubic MgO (JCPDS No. 21‐1152 and 45‐0946), the diffraction patterns exhibit phase‐specific reflections consistent with the $Fm\overline{3}m$ Pd (JCPDS No. 64‐8679) [15], $Pm\overline{3}m$ Pd₃Sn (JCPDS No. 10‐5689) [52], P6₃/mmc Pd₃Sn₂ (JCPDS No. 04‐0801) [53], and Aba2 PdSn₂ (JCPDS No. 64‐8930) [53] phases, respectively.

FIGURE 2.

Schematic illustration of the designed stepwise synthesis of the Pd₃Sn₂ intermetallic catalyst.

FIGURE 3.

(a) XRD patterns of the as‐synthesized Pd, Pd₃Sn, Pd₃Sn₂, and PdSn₂ catalysts. HRTEM images and the corresponding FFT patterns of (b) Pd₃Sn, (c) Pd₃Sn₂, and (d) PdSn₂ catalysts. HAADF‐STEM images and the corresponding particle size distribution histograms of (e) Pd₃Sn, (f) Pd₃Sn₂, and (g) PdSn₂ catalysts. HAADF‐STEM images, EDS line‐scanning profiles and mapping analyses of (h) Pd₃Sn, (i) Pd₃Sn₂, and (j) PdSn₂ catalysts.

High‐resolution transmission electron microscopy (HRTEM) and high‐angle annular dark‐field scanning transmission electron microscopy (HAADF‐STEM) were employed to probe the microstructure and composition of the monometallic Pd catalyst and the intermetallic Pd₃Sn, Pd₃Sn₂, and PdSn₂ catalysts. Representative HRTEM images with the corresponding fast Fourier transform (FFT) patterns show clear lattice fringes with interplanar spacings of 0.200, 0.136, 0.333, and 0.258 nm (Figure S33 and Figure 3b,d), which can be indexed to the (200) plane of face‐centered cubic Pd, the (220) plane of primitive cubic Pd₃Sn, the (101) plane of hexagonal Pd₃Sn₂, and the (122) plane of base‐centered orthorhombic PdSn₂, respectively. HAADF‐STEM images further show that

Pd, Pd₃Sn, Pd₃Sn₂, and PdSn₂ nanoparticles are uniformly dispersed on the support and exhibit similarly narrow size distributions, with average diameters of 6.2 ± 0.2, 7.0 ± 0.2, 6.3 ± 0.2, and 6.5 ± 0.2 nm (Figures S33–S36 and Figure 3e–g), respectively. Such similar particle sizes minimize size‐dependent effects and enable a more meaningful evaluation of the Sn‐induced geometric and electronic modification of Pd sites [54, 55]. Elemental distributions were examined by HAADF‐STEM coupled with EDS line scanning and elemental mapping (Figures 3h–j and S37–S42). Pd and Sn signals are spatially colocalized within individual nanoparticles for the Pd₃Sn, Pd₃Sn₂, and PdSn₂ catalysts, with Pd/Sn atomic ratios of 3:1, 3:2 and 1:2, respectively, in excellent agreement with inductively coupled plasma–atomic emission spectroscopy (ICP‐AES) measurements (Table S2). In contrast, the HAADF‐STEM EDS line‐scanning and elemental mapping for monometallic Pd catalyst show only Pd signal with no detectable Sn (Figure S43). These collective characterizations support the formation of the targeted Pd–Sn intermetallic phases with uniform composition and comparable dispersion, providing a reliable basis for correlating intermetallic structure with catalytic performance.

AC‐HAADF‐STEM techniques were further employed to probe the atomic arrangements of Pd₃Sn, Pd₃Sn₂, and PdSn₂ intermetallic nanoparticles. The atomic‐resolution image of Pd₃Sn shows a highly ordered rhombic lattice with distinct Pd and Sn atomic columns (Figure 4a). The line intensity profile along the yellow arrow indicates a lattice spacing of 0.200 nm (Figure 4d), consistent with the (111) plane of the primitive cubic Pd₃Sn phase. The atomic model viewed along the $[01\overline{1}]$ zone axis (Figure 4j), derived from FFT analysis of the highlighted region (Figure 4g), reproduces the experimentally observed lattice contrast, supporting an ordered Pd–Sn arrangement in which Sn occupies specific lattice positions within a Pd‐dominated framework. With increased Sn content, the Pd₃Sn₂ nanoparticle retains long‐range atomic ordering (Figure 4b), and exhibits well‐defined lattice fringes with an interplanar spacing of 0.278 nm corresponding to the (002) plane of the hexagonal Pd₃Sn₂ phase (Figure 4e). The atomic model along the [0$\overline{2}$0] zone axis matches the measured lattice contrast (Figure 4k,h), further corroborating the hexagonal intermetallic ordering with Sn incorporated into the ordered Pd–Sn lattice. At the highest Sn content, the PdSn₂ nanoparticle exhibits periodic lattice fringes with a spacing of 0.228 nm (Figure 4c,f), consistent with the (220) plane of the base‐centered orthorhombic PdSn₂ phase. The atomic model along the [00$\overline{1}$] zone axis (Figure 4l), derived from FFT analysis (Figure 4i), aligns with the observed lattice contrast, confirming the ordered Pd–Sn arrangement characteristic of the orthorhombic intermetallic lattice. These atomic‐resolution analyses confirm well‐defined intermetallic frameworks and atomically ordered surface structures for Pd₃Sn, Pd₃Sn₂, and PdSn₂ nanoparticles (Figure 4m–o). Additional measurements over extended nanoparticle populations further support the formation of $Pm\overline{3}m$ Pd₃Sn, P6₃/mmc Pd₃Sn₂, and Aba2 PdSn₂ phases (Figures S44–S46).

FIGURE 4.

Representative AC‐HAADF‐STEM images, corresponding line intensity profiles along the directions marked by yellow arrows, and FFT patterns of (a, d, g) Pd₃Sn, (b, e, h) Pd₃Sn₂, and (c, f, i) PdSn₂ catalysts. Crystal models along the $[01\overline{1}]$, $[0\overline{2}0]$, $[00\overline{1}]$ zone axes for (j) Pd₃Sn, (k) Pd₃Sn₂, and (l) PdSn₂ catalysts, respectively. Structural schematics of (m) Pd₃Sn, (n) Pd₃Sn₂, and (o) PdSn₂ catalysts are also shown.

The electronic structures of the catalysts were examined by x‐ray absorption near‐edge structure (XANES) and x‐ray photoelectron spectroscopy (XPS). As shown in Figure 5a, the Pd K‐edge XANES spectrum of the monometallic Pd catalyst closely resembles that of Pd foil, indicating a similar electronic environment. In contrast, the Pd K‐edge white‐line intensity of the Pd–Sn catalysts decreases relative to the Pd catalyst, suggesting electron transfer from Sn to Pd [52]. The decrease becomes more pronounced in the order of Pd₃Sn < Pd₃Sn₂< PdSn₂, suggesting progressively stronger Pd–Sn electronic interaction as the Sn/Pd ratio increases. Consistently, the Sn K‐edge XANES spectra of the Pd–Sn catalysts exhibit an increase in white‐line intensity relative to Sn foil (Figure 5b), with the same composition‐dependent trend, further corroborating electron depletion at Sn sites and the corresponding electron enrichment at the Pd centers. XPS measurements further support this observation (Figure S47), where the Pd 3d XPS peaks shift to lower binding energies and the Sn 3d XPS peaks shift to higher binding energies with increasing Sn/Pd ratios, confirming charge transfer from Sn to Pd.

FIGURE 5.

(a) Normalized XANES spectra at the Pd K‐edge for Pd foil, Pd, Pd₃Sn, Pd₃Sn₂, and PdSn₂ catalysts, with the inset displaying an enlarged view of the highlighted region. (b) Normalized XANES spectra at the Sn K‐edge for Sn foil, Pd₃Sn, Pd₃Sn₂, and PdSn₂ catalysts, with the inset displaying an enlarged view of the highlighted region. (c) Fourier transform EXAFS spectra and (e) EXAFS oscillation functions at the Pd K‐edge for Pd foil, Pd, Pd₃Sn, Pd₃Sn₂, and PdSn₂ catalysts. (d) Fitted Fourier‐transformed EXAFS spectra in R‐space for Pd₃Sn, Pd₃Sn₂, and PdSn₂ catalysts. Two‐dimensional charge density difference maps for (f) Pd₃Sn(111), (g) Pd₃Sn₂(102), and (h) PdSn₂(024) surfaces. (i) Projected electronic density of states (PDOS) of Sn p and s orbitals, and Pd d orbitals on Pd(111), Pd₃Sn(111), Pd₃Sn₂(102), and PdSn₂(024) surfaces.

To probe the local coordination environment of Pd in the Pd–Sn catalysts, extended x‐ray absorption fine structure (EXAFS) analysis was performed. In R‐space, the Fourier‐transformed EXAFS (FT‐EXAFS) spectra reveal a pronounced decrease in the intensity of the first‐shell Pd–Pd coordination peak relative to monometallic Pd, indicating partial disruption of Pd–Pd bonds due to incorporation of Sn (Figure 5c). Moreover, slight shifts in the positions of the first‐shell peaks for Pd₃Sn, Pd₃Sn₂, and PdSn₂ suggest subtle changes in Pd–M (M = Pd or Sn) bond lengths. These structural modifications are further corroborated by wavelet‐transformed EXAFS contour plots (Figure S48), which highlight the progressive evolution of Pd local coordination environments as the Sn content increases. Quantitative FT‐EXAFS fitting provides further insight into the evolution of Pd local coordination upon Sn incorporation (Figure 5d, Figures S49–S50, and Table S3). For Pd₃Sn, the first‐shell Pd–Sn coordination number is 2.1 (2.66 ± 0.03 Å), while the Pd–Pd coordination number decreases to 4.2, indicating the partial isolation of Pd sites. With increasing Sn content, the Pd–Sn coordination number increases while the Pd–Sn bond length remains essentially unchanged. For Pd₃Sn₂, FT‐EXAFS fitting yields a first‐shell Pd–Sn coordination number of 4.6 with a Pd–Sn bond length of 2.67 ± 0.02 Å, whereas the Pd–Pd coordination number decreases to 1.7 with a Pd–Pd distance of 2.80 ± 0.03 Å. This trend continues in PdSn₂, where the Pd–Sn coordination reaches 5.2 (2.66 ± 0.01 Å), while the Pd–Pd coordination decreases sharply to 0.9 (2.81 ± 0.02 Å). These results demonstrate increasingly isolated Pd local environments embedded in an ordered Pd–Sn lattice as the Sn content increases. Complementary evidence is provided by the k‐space EXAFS oscillations. Compared with the Pd foil and monometallic Pd catalyst, the Pd₃Sn, Pd₃Sn₂, and PdSn₂ catalysts exhibit shorter oscillations periods and reduced amplitudes (Figure 5e), consistent with the heavier backscattering contribution from Sn and the attenuation of Pd–Pd interactions due to Pd–Sn bond formation [13, 15]. Together with XANES and XPS, the EXAFS results support the formation of Pd₃Sn, Pd₃Sn₂, and PdSn₂ intermetallic phases with systematically tuned electronic and local structural environments [33, 34].

Bader charge analysis was performed on Pd₃Sn(111), Pd₃Sn₂(102), and PdSn₂(024) surfaces to quantify electronic interactions between Pd and neighboring Sn atoms (Figure 5f–h). The results show a systematic increase in electron accumulation on Pd along the Pd–Sn intermetallic series, with the smallest Pd electron gain in Pd₃Sn, a larger gain in Pd₃Sn₂, and the largest gain in PdSn₂. This trend is consistent with the XANES‐derived charge‐transfer picture (Figure 5a,b). Density of states (DOS) analysis further reveals significant overlap between Pd 4d and Sn 5s/5p orbitals, with particularly strong hybridization between Pd 4d and Sn 5p (Figure 5i). The extent of hybridization between Pd 4d and Sn 5p orbitals increases sequentially from Pd₃Sn to Pd₃Sn₂ and PdSn₂, in agreement with the electron transfer trend identified in the Bader analysis. As a result, the Pd d‐band center shifts progressively downward, slightly in Pd₃Sn, moderately in Pd₃Sn₂, and markedly in PdSn₂. Such electronic evolution rationalizes the systematic weakening of adsorbate binding and the concomitant decrease in hydrogenation activity toward the Sn‐rich end of the series [56, 57, 58].

Guided by the DFT‐derived design principle that selective hydrogenation requires maintaining hydrogen competence while minimizing alkene residence, we benchmarked Pd₃Sn₂ against Pd, Pd₃Sn, and PdSn₂ for propyne semihydrogenation under excess propylene. As shown in Figure 6a, the Pd₃Sn₂ catalyst exhibits a sharp increase in propyne conversion from 53.8% at 40°C to 98.0% at 100°C while maintaining consistently high propylene selectivity (>96.0%) across the entire temperature window, in line with the predicted combination of facile H₂ activation and favored propylene desorption on Pd₃Sn₂. In contrast, the Sn‐rich PdSn₂ catalyst remains much less active, with conversion increasing only from 0.2% at 130°C to 20.7% at 282°C despite its high selectivity, consistent with the computed difficulty in H₂ activation and weak alkyne adsorption on this phase. The Pd and Pd₃Sn catalysts show higher intrinsic reaction rates but suffer from pronounced over‐hydrogenation, leading to substantially reduced propylene selectivity (Figure S51). These low selectivities are in good accordance with the strong binding of propylene and ensemble‐enabled consecutive hydrogenation tendency inferred from theory. Kinetic analysis further corroborates these trends (Figure 6b). The Pd₃Sn₂ catalyst exhibits an apparent activation energy of 38.8 kJ·mol⁻¹, comparable to the Pd₃Sn catalyst (38.0 kJ·mol⁻¹) and only slightly higher than the Pd catalyst (33.2 kJ·mol⁻¹), indicating that high intrinsic hydrogenation activity is largely preserved. In contrast, the PdSn₂ catalyst shows a much higher apparent activation energy (58.4 kJ·mol⁻¹). This trend is consistent with the computed ${\mathrm{G}}_{\mathrm{AH}}^{\mathrm{eff}}$ values, which rank PdSn₂(024) as the least active surface (0.93 eV) and place Pd₃Sn₂(102) close to Pd₃Sn(111) (0.53 vs. 0.51 eV), with Pd(111) being the most active (0.38 eV). Importantly, these activity descriptors alone do not dictate selectivity, which instead depends on the substrate‐dependent fate of the alkene intermediate.

FIGURE 6.

(a) Temperature‐dependent propyne conversion and propylene selectivity over Pd, Pd₃Sn, Pd₃Sn₂, and PdSn₂ catalysts. The negative values originate from simultaneous hydrogenation of both propyne and co‐fed propylene to ethane under reaction conditions. (b) Arrhenius plots of Pd, Pd₃Sn, Pd₃Sn₂, and PdSn₂ catalysts measured under a gas mixture of 0.5 vol.% C₃H₄, 2.5 vol.% H₂, 4.5 vol.% C₃H₆, and N₂ balance at a total flow rate of 200 mL·min⁻¹. The propyne conversion was kept below 20% to ensure differential reaction conditions. (c) HD formation during H₂‐D₂ isotope exchange over Pd, Pd₃Sn, Pd₃Sn₂, and PdSn₂ catalysts at 30°C, with the values indicating H₂ conversion. (d) C₃H₄‐TPD profiles of Pd, Pd₃Sn, Pd₃Sn₂, and PdSn₂ catalysts. (e) C₃H₆‐TPD profiles of Pd, Pd₃Sn, Pd₃Sn₂, and PdSn₂ catalysts. (f) MS signals of C₃H₈ (m/z = 44) recorded during TPSR with C₃H₆ and H₂ as feed gases over Pd, Pd₃Sn, Pd₃Sn₂, and PdSn₂ catalysts. (g) Long‐term stability test of Pd₃Sn₂ catalyst at 100°C under initial reaction conditions (0.5 vol % C₃H₄, 2.5 vol % H₂, 4.5 vol % C₃H₆, balanced with N₂). The unit for GHSV is mL·g⁻¹·h⁻¹. (h) Schematic illustration of propyne hydrogenation pathways over different Pd–Sn intermetallic catalysts.

To experimentally validate the DFT prediction that the Pd₃Sn₂ catalyst retains hydrogen competence via cooperation between surface and near‐surface sites while promoting propylene desorption through weakened π‐binding, we probed hydrogen activation and adsorption/desorption behaviors of propyne and propylene. H₂‐D₂ exchange experiments show that the Pd₃Sn₂ catalyst achieves an H₂ conversion of 42.7% at 30°C (Figures 6c and S52), approaching those of Pd₃Sn (43.2%) and monometallic Pd (45.8%). This result is consistent with the DFT calculations which demonstrate the spontaneous H₂ dissociation on Pd₃Sn₂(102) enabled by cooperation between Sn‐bridged Pd–Pd dual sites and adjacent near‐surface Pd. In contrast, the PdSn₂ catalyst exhibits only 10.2% H₂ of conversion, also in good accordance with the higher H₂ dissociation barrier calculated by DFT calculations (Figures 6c and 1a). Temperature‐programmed desorption links adsorption thermodynamics to intermediate fate. The C₃H₄‐TPD profile of the Pd₃Sn₂ catalyst shows three distinct desorption peaks at 81.8°C, 229.6°C, and 356.7°C (Figure 6d), reflecting multiple adsorption modes favorable for propyne activation. In contrast, those of Pd and Pd₃Sn catalysts show substantially stronger binding and the PdSn₂ catalyst exhibits only weak, single‐site adsorption. Moreover, C₃H₆‐TPD measurements indicate that propylene desorbs at a relatively low temperature of 162.7°C (Figure 6e), supporting weak binding of propylene on the Pd₃Sn₂ catalyst and aligning with DFT calculations, which predict that propylene desorption is kinetically favored over further hydrogenation on the Pd₃Sn₂ catalyst.

The selectivity trend established for C₃ substrates breaks down for C₂ systems. Consistent with DFT results indicating stronger ethylene binding and a greater tendency toward consecutive hydrogenation, acetylene hydrogenation over the Pd₃Sn₂ catalyst exhibits poor ethylene selectivity even in the absence of a co‐fed ethylene. As shown in Figure S53, the ethylene selectivity on the Pd₃Sn₂ catalyst drops sharply with increasing temperature and H₂/C₂H₂ ratio. Notably, at a H₂/C₂H₂ ratio of 5, the ethylene selectivity falls from 99.6% at 30°C to 60.5% at 160°C, together with a rise in ethane selectivity to 39.5%. In contrast, the C₄ production is negligible in all cases, confirming that selectivity loss is driven primarily by overhydrogenation. These results provide clear evidence that without methyl‐induced steric hindrance, the intermediate ethylene on the hydrogen‐competent Pd₃Sn₂ surface, tends to be hydrogenated to ethane. The impact of intermediate residence time is further substantiated by varying the gas hourly space velocity (GHSV) at 125°C. As shown in Figure S54, a tenfold increase in GHSV (from 10 × 10⁴ to 110 × 10⁴ mL·g⁻¹·h⁻¹) decreases C₂H₂ conversion from 93.6% to 80.2%, yet ethylene selectivity remains largely unchanged at 68.8%. Such insensitivity of ethylene selectivity to space velocity implies that overhydrogenation is governed by strong intrinsic binding rather than secondary re‐adsorption from the gas phase. This selectivity decrease in C₂ system is further exacerbated under industrially relevant ethylene‐rich conditions (Figures S55–S56). In an ethylene‐rich co‐feed, the ethylene selectivity drops significantly at elevated temperatures, falling to 34.56% at 127.8°C as the ethane selectivity increases to 55.96%. High‐GHSV tests at 110°C fail to significantly restore ethylene selectivity, which barely reaches 47.2% even under a GHSV of 110 × 10⁴ mL·g⁻¹·h⁻¹. These findings underscore that while Pd₃Sn₂ leverages methyl‐induced repulsion for C₃ selectivity, its hydrogen competence becomes a liability for C₂ substrates due to the lack of steric hindrance. The intrinsic nature of this C₂ behavior is further confirmed by long‐term durability tests and temperature‐programmed surface reaction (TPSR) measurements. As shown in Figure S57, Pd₃Sn₂ maintains a stable but low ethylene selectivity (∼32.0%) over 48 h at 110°C. TPSR analysis reveals that ethane formation during ethylene hydrogenation initiates at 30°C and peaks at only 75°C (Figure S58), directly validating the high propensity for acetylene overhydrogenation.

The remarkable resistance of the Pd₃Sn₂ catalyst to propylene over‐hydrogenation was evidenced by propylene hydrogenation and TPSR measurements. As shown in Figure S59, the Pd₃Sn₂ catalyst exhibits substantially lower propylene conversion than the Pd, Pd₃Sn, and PdSn₂ catalysts across the investigated temperature range, indicating strong suppression of consecutive hydrogenation. This suppression is further supported by temperature‐programmed hydride decomposition (TPHD) measurements. The TPHD profiles of the Pd and Pd₃Sn catalysts exhibit distinct H₂ evolution peaks at 57.5°C and 80.7°C, respectively, characteristic of the decomposition of formed hydride (Figure S60) [12, 15]. In contrast, the profiles of the Pd₃Sn₂ and PdSn₂ catalysts show negligible signals, indicating suppressed formation of hydride and limited subsurface hydrogen availability in these Sn‐rich intermetallic phases. TPSR analysis further confirms this behavior (Figure 6f). The Pd, Pd₃Sn, and PdSn₂ catalysts all generate significant amounts of propane, with two distinct desorption peaks for the Pd and Pd₃Sn catalysts, and a single high‐temperature peak for the PdSn₂ catalyst. In contrast, the propane formation on the Pd₃Sn₂ catalyst is nearly undetectable. Together, these results demonstrate that the exceptional propylene selectivity of the Pd₃Sn₂ catalyst originates from the cooperative ensemble of Sn‐bridged Pd–Pd dual sites and adjacent near‐surface Pd atoms, which balance propyne activation with rapid propylene desorption while effectively suppressing over‐hydrogenation [57, 59]. Moreover, the results further clarify why the same intermetallic framework can behave differently for C₃ versus C₂ substrates when alkene binding geometries and steric/electronic interactions differ.

The practical relevance of this structure–performance relationship was systematically evaluated by varying the GHSV, which modulates contact time and amplifies differences in intermediate residence and consecutive hydrogenation. The monometallic Pd catalyst maintains high propyne conversion (∼97.0%) across the tested GHSV range, yet the propylene selectivity remains negative, increasing only from ⁻336.8% to ⁻70.8% (Figure S61). This indicates net propylene consumption by consecutive hydrogenation under the excess‐propylene co‐feed conditions. The Pd₃Sn catalyst displays a similar tendency, consistent with strong propylene binding and limited kinetic preference for propylene desorption. In contrast, the PdSn₂ catalyst maintains high propylene selectivity but exhibits low propyne conversion (<25%), indicative of limited intrinsic activity. Stability tests further highlight these limitations (Figure S62). Although the Pd, Pd₃Sn, and PdSn₂ catalysts exhibit steady performance over 48 h, they fail to balance conversion and selectivity. Specifically, Pd and Pd₃Sn show excessive propane selectivity (∼300% and ∼140%, respectively), confirming severe over‐hydrogenation of both the propyne feed and the co‐fed propylene. In contrast, while propane formation is negligible on PdSn₂, its propyne conversion remains below 20%. Notably, the Pd₃Sn₂ catalyst remains simultaneously active and selective for the hydrogenation (Figure 6g). During a 96‐h run at a GHSV of 22,500 mL·g⁻¹·h⁻¹, propyne conversion and propylene selectivity remain at 93.8% and 99.2%, respectively. Further evaluation across GHSVs of 15 000, 18 750, 30 000, and 45 000 mL·g⁻¹·h⁻¹ (11 h at each condition) shows sustained propylene selectivity of approximately 98.5%, demonstrating a robust operating window that markedly outperforms previously reported catalysts in both activity and selectivity (Table S4).

Atomic‐scale characterization of the spent catalyst was performed to investigate possible structural changes during the stability tests. As shown in Figures S63–S68, the structural integrity of the Pd₃Sn, Pd₃Sn₂, and PdSn₂ catalysts is fully preserved after reaction, confirming the excellent structural robustness of the catalysts under the reaction conditions. Thermogravimetric analyses of the spent catalysts were also conducted to evaluate carbonaceous deposition (Figure S69). The Pd₃Sn₂ catalyst exhibits a significantly lower weight loss (only 1.3%) than the Pd and Pd₃Sn catalysts (4.7% and 3.5%, respectively). These results indicate clearly reduced carbon deposition on the Pd₃Sn₂ catalyst, though the time of the stability test is much longer than those for other two catalysts. These results correlate the observed macroscopic robustness to the DFT‐identified functional separation on Pd₃Sn₂, where propyne activation remains efficient while propylene desorbs rapidly enough to avoid overhydrogenation (Figure 6h), and they reinforce the substrate‐dependent selectivity concept established by the contrasting C₂ behavior.

3. Conclusion

In summary, this work demonstrates that Pd–Sn intermetallics can exhibit substrate‐dependent semihydrogenation selectivity, where closely related C₂ and C₃ substrates follow distinct intermediate fates on the same hydrogen‐competent catalytic platform. Theoretical calculations identify Pd₃Sn₂ as a phase that reconciles the conventional activity–selectivity trade‐off by enabling cooperative H₂ activation at surface Sn‐bridged Pd–Pd dual sites involving near‐surface Pd species, while kinetically favoring propylene desorption over further hydrogenation. Experimentally, the synthesized Pd₃Sn₂ intermetallic catalyst delivers high propylene selectivity at near‐complete conversion with long‐term stability under high space velocity, whereas the Pd and Pd₃Sn catalysts exhibit low selectivities and the PdSn₂ catalyst lacks activity. The contrasting failure in acetylene semihydrogenation, manifested as extensive ethane formation due to overhydrogenation of both newly formed and co‐fed ethylene, directly validates the predicted C₂/C₃ divergence arising from molecular substitution effects on π‐binding geometry and residence. Mechanistic studies further support that suppressing strong olefin retention is critical to avoiding consecutive hydrogenation. These findings suggest that selective hydrogenation can be programmed by coupling hydrogen competence with substrate‐dependent intermediate binding and residence, enabling rapid product desorption without compromising activation.

Author Contributions

Yijing Liang: methodology, investigation, writing – original draft, data curation, validation. Ningchao Zhu: methodology, investigation, writing – original draft, conceptualization, data curation, validation. Yundao Jing: conceptualization, investigation, methodology, validation, data curation, writing – original draft. Xiaohu Ge: conceptualization, investigation, writing – original draft, funding acquisition, validation, methodology, data curation, supervision. Jing Zhang: writing – original draft. Gang Qian: writing – original draft. Yueqiang Cao: conceptualization, investigation, funding acquisition, writing – original draft, methodology, validation, data curation, supervision. Xinggui Zhou: writing – original draft. De Chen: writing – original draft. Weikang Yuan: writing – original draft. Xuezhi Duan: conceptualization, investigation, funding acquisition, writing – original draft, methodology, validation, data curation, supervision.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File 1: anie72795‐sup‐0001‐SuppMat.docx.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.