Layered Pd oxide on PdSn nanowires for boosting direct H₂O₂ synthesis

Abstract

Hydrogen peroxide (H₂O₂) has the wide range of applications in industry and living life. However, the development of the efficient heterogeneous catalyst in the direct H₂O₂ synthesis (DHS) from H₂ and O₂ remains a formidable challenge because of the low H₂O₂ producibility. Herein, we develop a two-step approach to prepare PdSn nanowire catalysts, which comprises Pd oxide layered on PdSn nanowires (Pd_L/PdSn-NW). The Pd_L/PdSn-NW displays superior reactivity in the DHS at zero Celcius, presenting the H₂O₂ producibility of 528 mol kg_cat⁻¹·h⁻¹ and H₂O₂ selectivity of >95%. A layer of Pd oxide on the PdSn nanowire generates bi-coordinated Pd, leading to the different adsorption behaviors of O₂, H₂ and H₂O₂ on the Pd_L/PdSn-NW. Furthermore, the weak adsorption of H₂O₂ on the Pd_L/PdSn-NW contributes to the low activation energy and high H₂O₂ producibility. This surface engineering approach, depositing metal layer on metal nanowires, provides a new insight in the rational designing of efficient catalyst for DHS.

Introduction

Hydrogen peroxide (H₂O₂) is widely used as oxidant in many fields because it is environmentally friendly and the byproducts only involve H₂O and O₂, as compared to other oxidants (e.g. Cl-containing oxidants and nitric acid)^1,2. Currently, the anthraquinone oxidation process (AO process) is employed as the method to produce H₂O₂ in industry, whereas the AO process requires highly intense energy consumption and the addition of toxic organic solvents, such as alkylbenzene and so on. Therefore, it is desirable to develop sustainable process and more efficient strategy towards the production of H₂O₂^3,4,5,6,7.

Direct H₂O₂ synthesis (DHS) from H₂ and O₂ is an effective alternative for small-scale and on-site H₂O₂ production, which is usually performed using Pd-based catalyst^8,9,10,11. However, the catalytic activity was limited due to the high-side reactions such as hydrogenation and decomposition when employing a pure Pd-based catalyst, leading to a decreased net yield. Numerous efforts have been devoted to developing high-performance catalysts towards DHS, such as the use of bimetallic Pd-Au^12,13,14,15, trimetallic Pd-Au-Pt¹⁶, acid and halide additives^{17,18,19,20,21}. However, these Pd catalysts contain expensive gold and the H₂O₂ yield is still far from that achieved by the AO process^22,23,24. The addition of Sn to Pd catalysts has attracted intensive attention due to the high stability and inert hydrogenation with Sn component for the H₂O₂ synthesis³. In particular, the synthesis of PdSn nanocatalyst involved a multi-step protocol of oxidation-reduction-oxidation (O-R-O), and a tin oxide surface layer that encapsulates small Pd-rich particles was formed while leaving larger PdSn alloy particles exposed. The PdSn nanocatalyst prepared via O-R-O presented the H₂O₂ production with high selectivity of >95%, while only showing the H₂O₂ producibility of ~70 mmol·g_cat⁻¹·h⁻¹.

Metal nanowires (NWs) is one-dimensional (1D) structure material, which has been utilized in photonic^25,26, electrical^27,28, and plasmonic-related applications^28,29. As compared to nanoparticles or bulk material, these 1D nanowire structures can expedite orientable electronic or ions transfer and diffusion to promote catalytic kinetics. Therefore, nanowires can provide a unique platform to study catalysis^30,31. For example, in a nanowire-bacteria hybrid system, nanowires can capture electrons and deliver them to bacteria, allowing bacteria to ensure the conversion of CO₂²⁶. On the other hand, cobalt oxide (CoO) nanorods/nanowires were reported to create oxygen vacancies on the nanofacets²⁷. The catalyst exhibits excellent electrocatalytic ORR/OER performance due to the modulated electronic structure of cobalt oxide nanorods/nanowires revealed by simulation.

Herein, we develop an approach to prepare PdSn nanowires to directly produce H₂O₂ and found that a layer of Pd oxide on a Pd₄Sn alloy nanowire (NW) prepared via two-step synthesis (Fig. 1) present efficient reactivity in the direct production of H₂O₂. This approach involves the synthesis of the surface-rough Pd₄Sn nanowires (PdSn-NW, Sn:Pd molar ratio is 4) by a solvothermal method firstly (Supplementary Fig. 1)³². Then, metal Pd precursor was deposited onto the Pd₄Sn nanowires, followed by dispersing on the TiO₂ support. After rapid annealing of the material in air, an efficient catalyst for direct H₂O₂ synthesis was obtained and denoted as Pd_L/PdSn-NW catalyst (Supplementary Fig. 2). The unsupported Pd_L/PdSn-NW presents the morphology of worm-liked nanowires and some nanowires connect each other to form interconnected structures (Fig. 2a). The outmost surface of the structures contains Pd and Sn terminated with oxygen because both of the two metals were oxidized after the annealing in the air as shown in Fig. 1. We also synthesize and test the TiO₂-supported PdSn nanowire catalyst prepared via one-step synthesis (denoted as PdSn-NW). As can be seen, the two unsupported materials show similar nanowire morphology (Fig. 2a and Supplementary Fig. 1). After loading onto TiO₂, the nanowire morphologies do not change (see discussion below).

Fig. 1: Schematic illustration of the synthesis of the unsupported PdSn-NW and Pd_L/PdSn-NW.

The supported PdSn-NW and Pd_L/PdSn-NW catalysts are supported on TiO₂. The two-step protocol involving an annealing process was used for the synthesis of Pd_L/PdSn-NW. PdSn nanowire (NW) was prepared with a Pd: Sn molar ratio of 4 firstly and then, Pd precursor was deposited on the as-prepared PdSn nanowires again, followed by depositing on TiO₂ and a rapid annealing in air. A PdSn-NW catalyst with Pd:Sn ratio of 4 was also synthesized for the purpose of comparison via one-step method by mixing Sn²⁺, Pd²⁺, PVP, EG, and NH₄Br.

Fig. 2: Representative TEM image and the catalytic performances of the PdSn nanowire catalysts in the direct H₂O₂ synthesis (DHS).

a Representative TEM image of the unsupported PdSn nanowire catalyst prepared by two-step (unsupported Pd_L/PdSn-NW). b H₂O₂ producibility of the supported Pd_L/PdSn-NW and Sn_x/PdSn-NW catalysts with different Pd/Sn ratios after annealing in air (350 °C, 8 min), demonstrating the addition of Sn has negative effect on the producibility of H₂O₂. c H₂O₂ producibility, hydrogenation, and decomposition of the supported Pd_L/PdSn-NW catalyst annealing at different temperatures in air, showing that the supported Pd_L/PdSn-NW annealing at 400 °C did not catalyze the hydrogenation and decomposition. d The comparison of the H₂O₂ producibility, hydrogenation, and decomposition of supported Pd_L/PdSn-NW catalyst with other Pd catalysts (Table 1). The error bars in b–d show the standard deviation in the measurements. The standard deviation was achieved from the repeated runs of three to five times using fresh catalyst each time. The error bars refers to the standard deviation of multiple times measurements of hydrogen peroxide productivity.

Results and discussion

Materials synthesis and catalytic performance

The catalytic reactivity of the TiO₂-supported PdSn nanowire catalyst synthesized by the two-step (Pd_L/PdSn-NW, 4.1 wt.%Pd loading) was evaluated in the direct synthesis of H₂O₂ at zero Celcius. In comparison, we also synthesized PdSn-NW catalyst via one-step (PdSn-NW), PdSn nanoparticle (PdSn-NP), and SnO_x supported on PdSn-NW (SnO_x/PdSn-NW) via the two-step approach. The supports of all these reference catalysts are TiO₂ and all the activity data reported in the work are achieved from catalysts supported on TiO₂. Prior to dispersing on TiO₂, the morphologies of these reference nanowire/nanoparticles are well characterized using TEM (Supplementary Figs. 3, 4). It should be mentioned that both TEM-EDS and X-ray photoelectron spectroscopy (XPS) analysis indicate that there is no residual Br ion detected on the Pd catalyst (Supplementary Fig. 5) and therefore, the effect of the Br ion in the H₂O₂ synthesis can be excluded. The supported Pd_L/PdSn-NW catalyst synthesized by two-step shows the high H₂O₂ producibility of ~680 mol/kg h⁻¹ after annealing at 350 °C in the air (Fig. 2b). The addition of Sn onto the PdSn nanowires leads to the formation of SnO_x/PdSn-NW catalyst, which is less active than the supported Pd_L/PdSn-NW catalyst after annealing at 350 °C in air. Moreover, with the increase of Sn loading (4.5–13 wt.%), the reactivity of the SnO_x/PdSn-NW catalyst further decreases, indicating the addition of Sn on the PdSn nanowires has a negative effect on the reactivity of the PdSn-NW (Fig. 2b). However, the Pd_L/PdSn-NW catalyst annealing at 350 °C presents the high hydrogenation activity, which is undesirable for the DHS³. We investigated the effect of the annealing temperature on the reactivity of the Pd_L/PdSn-NW catalyst and found that annealing the Pd_L/PdSn-NW catalyst at 400 °C presented the H₂O₂ producibility of ~528 mol/kg h⁻¹ with the complete absence of hydrogenation (Fig. 2c). Therefore, 400 °C was employed as the annealing temperature for all supported catalysts in the following discussion, unless otherwise noted. Moreover, the Pd/Sn ratios and metal loadings of all the catalysts were measured by EDS and ICP-OES analysis, and a good agreement was obtained using the two techniques (Supplementary Figs. 6 and 7).

When using the two-step approach to deposit Pd precursor onto the as-synthesized PdSn nanowires to form Pd_L/PdSn-NW catalyst, the Pd loading is precisely controlled to cover the PdSn nanowires as monolayer (Supplementary Note 1) and two layers, which are denoted as Pd_L/PdSn-NW and Pd_2L/PdSn-NW, respectively. The annealed Pd_L/PdSn-NW catalyst shows superior reactivity than the PdSn-NW catalyst prepared via one-step (Fig. 2d and Supplementary Fig. 8). Although PdSn nanowire alone (PdSn-NW) shows no hydrogenation or decomposition activity (Fig. 2d), the H₂O₂ selectivity is only ~70% (Table 1). This is explained by the fact that the H₂O₂ selectivity is calculated from the first step in this two-step process (Supplementary Fig. 9). Furthermore, the annealed Pd_L/PdSn-NW catalyst presents outstanding catalytic activity for H₂O₂ production and higher H₂O₂ selectivity, as compared to Pd_2L/PdSn-NW catalyst (Fig. 2d). This indicates that the addition of the extra Pd onto the Pd_L/PdSn-NW has a negative effect on the H₂O₂ production. Besides, both the H₂O₂ hydrogenation and decomposition are completely inhibited on the annealed Pd_L/PdSn-NW catalyst, as compared to other active Pd catalysts reported. Furthermore, the annealed Pd_L/PdSn-NW catalyst is very stable in the reaction and no deactivation is found in multiple runs (Supplementary Fig. 10). The reactivity of the supported Pd_L/PdSn-NW catalyst was also compared with other Pd-based catalysts (Fig. 2d and Table 1, Supplementary Tables 1 and 2), such as Pd nanoparticles, PdSn nanoparticles, and nice PdSn nanowires prepared by the approach reported in the literature³². None of them showed comparable H₂O₂ producibility and low decomposition/hydrogenation activity (Supplementary Fig. 11) as the supported Pd_L/PdSn-NW catalyst exhibiting the H₂O₂ selectivity of >95% under the same conditions. It should be mentioned that the H₂O₂ productivity slightly slows down over time (after 30 min) and it is partially ascribed to the decrease of the reactant gas concentration and the accumulation of H₂O₂ in the autoclave according to Le Chatelier’s principle (Supplementary Fig. 12). Besides, it possibly suggests that the catalyst changes in the presence of high concentration H₂O₂ and either become less active or begins to degrade H₂O₂ at high H₂O₂ concentrations.

Table 1 Catalytic performances of various Pd catalysts supported on TiO₂ after pretreating under different conditions in the direct synthesis of H₂O₂ from H₂ and O₂

Materials characterization

The composition and morphology of the Pd_L/PdSn-NW catalyst are first characterized by X-ray diffraction (XRD) and electron microscopy. XRD patterns of the unsupported PdSn nanowire prepared by two-step before and after annealing show the diffraction peaks located at lower diffraction angles (i.e., larger d spacings) as compared to that of metal Pd (Fig. 3a), indicating the crystal expansion of metal Pd due to the alloying of Sn atom. Annealing this unsupported PdSn nanowire at 350 °C or 400 °C does not make significant changes on the XRD patterns, indicating that the nanowire is thermally stable. When dispersing the nanowire on TiO₂, XRD pattern of the supported Pd_L/PdSn-NW catalysts only presents the diffraction peaks of the support TiO₂ (Fig. 3a and Supplementary Fig. 13), indicating the high dispersion of PdSn nanowire on the support. After annealing the supported Pd_L/PdSn-NW catalyst at different temperatures (350 °C and 400 °C, Supplementary Fig. 13a) in air, we do not find the presence of other diffraction peak or phase separation, indicating that the PdSn-NW and Pd_L/PdSn-NW catalysts are quite stable after the pretreatment. The representative high-resolution transmission electron microscopy (TEM)/scanning transmission electron microscopy (STEM) images of the Pd_L/PdSn-NW are shown in Fig. 3b, c. The nanowires have the diameter ranging from 5 to 10 nm and STEM-EDS line scanning shows the presence of both Pd and Sn in the nanowire (Supplementary Fig. 14). The high-resolution TEM of the nanowire is shown in Fig. 3b and the image exhibits the lattice fringes of 0.23 and 0.20 nm, corresponding to the (111) and (200) plane of Pd₄Sn phase, respectively (more images are shown in Supplementary Fig. 15). STEM-EDS mapping demonstrated that both Pd and Sn signals are uniformly dispersed in the nanowires (Fig. 3c). Therefore, the electron microscopy characterization of the annealed Pd_L/PdSn-NW sample demonstrates that the nanowire morphology of PdSn-NW is well maintained after loading onto TiO₂ and annealing in air.

Fig. 3: Structural characterization of the unsupported and supported Pd_L/PdSn-NW catalysts before and after annealing in air at 400 °C.

a XRD patterns of the unsupported and supported Pd_L/PdSn-NW catalysts before and after annealing in air. b HRTEM image of the unsupported Pd_L/PdSn-NW sample showing the lattice fringes of the PdSn phase. c STEM image and STEM-EDS mapping of the supported Pd_L/PdSn-NW showing the close proximity of both Pd and Sn elements, indicating the successful synthesis of Pd₄Sn alloy. d XPS spectra of Pd 3d core level of the supported Pd_L/PdSn-NW catalyst before and after annealing in air. e Surface valence band photoemission spectra of the supported Pd_L/PdSn-NW catalysts annealing at different temperatures in air. f HAADF-STEM image of the supported Pd_L/PdSn-NW catalyst after annealing showing a layer of Pd oxide on the PdSn nanowires.

XPS is used to investigate the surface species of the supported Pd_L/PdSn-NW catalyst during the annealing process (Fig. 3d, e, Supplementary Figs. 16–19). As shown in Fig. 3d, the XPS spectrum of the unannealed Pd_L/PdSn-NW catalyst shows that the main Pd species is Pd metal and the presence of a very small amount of Pd²⁺ prior to annealing. After annealing at 350 °C in air, the Pd²⁺ species content increases and Pd metal is still the dominant species on the surface. After annealing the Pd_L/PdSn-NW at 400 °C in air, we observe that the main Pd on the catalyst is Pd²⁺ species with a small amount of Pd metal, demonstrating the oxidation of metal Pd after annealing. Furthermore, the XPS analysis indicates that the Sn species on the supported Pd_L/PdSn-NW catalyst is SnO_x (Supplementary Fig. 19) and the Pd/Sn ratio is ~2. Since Sn/PdSn-NW catalysts are less active and selective than PdSn-NW catalyst, the active site is not ascribed to the SnO_x. Therefore, the rapid annealing process generated Pd oxide on the PdSn nanowires, which is observed in the HAADF-STEM image (Fig. 3f and Supplementary Fig. 20). The d-band model has been widely used in order to understand activity trends in metal-surface-catalyzed reactions³³. Here, the d-band center obtained from XPS was used to study the interaction between gas adsorption/intermediates and metal surface during annealing in air. The surface valence band photoemission spectrum of the annealed Pd_L/PdSn-NW is more delocalized and widely distributed than that of the unannealed Pd_L/PdSn-NW (Fig. 3e), indicating the different activation ability of gas molecules or intermediates on the annealed Pd_L/PdSn-NW catalyst³⁴. The annealed Pd_L/PdSn-NW catalyst, therefore, provides Pd oxide to modulate the adsorption/desorption behaviors of the reactants on the catalyst surface, which was confirmed by the improved H₂O₂ producibility of Pd_L/PdSn-NW (Table 1) and density functional theory (DFT) calculation below.

The chemical environment of the Pd atoms in the supported Pd_L/PdSn-NW catalyst after annealing in the air was examined via X-ray adsorption spectroscopy (Fig. 4a and Supplementary Figs. 21 and 22). We also performed the XAS measurements of the Pd foil, PdO, PdSn-NW, and PdSn-NP catalysts for comparison. For the reference Pd foil, a major peak attributing to the Pd-Pd scattering was observed. The EXAFS spectra of PdSn-NW and PdSn-NP catalysts show a similar or slightly expanded scattering distance as compared to that of the metal Pd foil (Fig. 4a), which is attributed to the Pd-Pd and PdSn scattering. The Pd-K edge EXAFS of the supported Pd_L/PdSn-NW catalyst after annealing (air, 400 °C) also have a major peak corresponding to Pd-Pd and PdSn scattering (Fig. 4a), indicating the main Pd phase on the Pd_L/PdSn-NW catalyst is metal Pd and the rapid annealing process does not change the bulk phase of the PdSn nanowires. EXAFS fit gave a coordination number of 4.2 at a bond distance of 2.76 Å (Supplementary Table 3). The EXAFS is fitted to Pd-Pd and PdSn scattering features because the XRD results clearly indicate that the bulk phase of the PdSn nanowire after annealing in air is metal Pd (Fig. 3a) and both XRD and XAS are bulk techniques. Furthermore, the high-resolution TEM (HRTEM) image shows the lattice fringes corresponding to metallic PdSn nanowire (Supplementary Fig. 15a). Since XPS is more surface sensors for observing the changes under different conditions than XRD and XAS, We performed near-ambient pressure X-ray photoelectron spectroscopy (NAP-XPS) to in-situ investigate the surface properties of the PdSn catalysts (both PdSn nanoparticle and PdSn nanowire prepared via one-step and two-step approaches) in the flowing O₂, H₂ and O₂/H₂ mixture (Fig. 4b–d). Conventional PdSn nanoparticle catalyst showed the presence of Pd oxides in both vacumm and the flowing O₂, whereas metal Pd was formed after switching to H₂ or H₂/O₂ mixture (Fig. 4d), indicating that the conventional PdSn nanoparticles are not stable in H₂ and H₂/O₂ atmospheres. For the PdSn nanowire catalysts, the Pd_L/PdSn-NW prior to annealing mainly shows the metal Pd phase on the catalyst surface (Fig. 4b) and there is a very small change for the Pd 3d XPS spectra of the catalyst after flushing in H₂, O₂, and H₂/O₂, indicating that the unannealing Pd_L/PdSn-NW maintain the metal Pd surface in the flowing H₂/O₂. Likewise, the Pd 3d XPS spectrum of the Pd_L/PdSn-NW after annealing in air at 400 °C indicated that Pd oxide is the main Pd species on the surface at vacuum (Fig. 4c). The introduction of a flow of O₂ has no observed effect on the Pd 3d XPS spectrum of the annealed Pd_L/PdSn-NW. After introducing a flow of H₂, we observed the presence of metal Pd, as well as the presence of Pd oxide. However, in the presence of H₂/O₂ mixture, the Pd_L/PdSn-NW surface after annealing was maintained as Pd oxide. Therefore, in-situ NAP-XPS evidenced that the surface species of the Pd_L/PdSn-NW after annealing is Pd oxide in the reaction feed of H₂/O₂, while the major Pd species on both PdSn-NP and PdSn-NW are metal Pd in the presence of H₂/O₂.

Fig. 4: EXAFS spectra of the PdK edge of PdSn catalysts and reference samples, and NAP-XPS of Pd 3d spectra for PdSn catalysts under different treatment conditions.

a EXAFS spectra of the PdK edge of PdSn catalysts after annealing in air. b In situ NAP-XPS of Pd 3d spectra for PdSn-NW in the presence of O₂, H₂, and O₂/H₂. c In situ NAP-XPS of Pd 3d spectra for Pd_L/PdSn-NW after annealing in air in the presence of O₂, H₂ and O₂/H₂. d In situ NAP-XPS of Pd 3d spectra for PdSn-NP after annealing in air in the presence of O₂, H₂ and O₂/H₂.

Theoretical studies on reaction mechanism

Based on the above reactivity and characterization results, we propose that the excellent reactivity of the PdSn nanowires prepared by the two-step approach (Pd_L/PdSn-NW) is associated with the Pd oxide layer on PdSn nanowire which was detected via AC-STEM and NAP-XPS (Figs. 3 and 4). The stable Pd oxide layer on PdSn nanowire in the presence of H₂/O₂ exhibited enhanced reactivity as compared to the conventional PdSn nanoparticle and PdSn nanowire prepared by a one-step approach. To better understand the unique properties of the Pd oxide layer on the Pd_L/PdSn-NW in direct H₂O₂ synthesis, we performed DFT calculations based on three models of PdO(101), Pd₄Sn and PdO monolayer supported on Pd₄Sn(111) (PdO@Pd₄Sn) (Fig. 5a). The PdO(101) surface contains tetra-coordinated Pd (denoted as Pd_4c) and tri-coordinated Pd (denoted as Pd_3c) and the outmost surface atoms of the Pd₄Sn model are Pd atom, while the PdO@Pd₄Sn surface contains Pd_4c and bi-coordinated Pd (denoted as Pd_2c). Details for model construction can be found in the computational method section. Bader charge analysis shows that Pd_4c on PdO@Pd₄Sn and PdO(101) are charged similarly with +0.83|e| and +0.85|e| (Supplementary Fig. 23), respectively. Nevertheless, the charges of Pd_2c on PdO@Pd₄Sn and Pd_3c on PdO(101) are different, with the former being charged +0.46|e| and the latter having a charge of +0.67|e|. The surface Pd atoms in Pd₄Sn are slightly positive by +0.11|e|, which is therefore more metallic than those in PdO(101) and PdO@Pd₄Sn. Thus, these three catalysts with different Pd chemical states may lead to different interactions between the adsorbate and the catalyst.

Fig. 5: Adsorption energy of key species and proposed mechanism.

a DFT optimized structures of PdO(101), Pd₄Sn, and PdO@Pd₄Sn with b adsorption energies of H₂, O₂, H₂ + O₂, and H₂O₂, c free energy profiles for H₂ activation, and d free energy profiles for O₂ reduction by the surface hydrides on these three models. TS transition state. Color code: Pd, blue; O, red; H, white; Sn, yellow.

Subsequently, we investigated the adsorption behavior of the reaction molecules (O₂, H2, and H₂O₂) on PdO(101), Pd₄Sn, and PdO@Pd₄Sn. The adsorption configurations and corresponding adsorption energies are shown in Supplementary Fig. 24 and Fig. 5b, respectively. The calculation results show that all three molecules exhibit strong adsorption on the PdO(101) surface. H₂ prefers to adsorb on top of Pd_3c and O₂ tends to adsorb on the Pd_3c-Pd_3c bridge site, leading to adsorption energies of −0.62 and −1.56 eV, respectively. For H₂O₂, we found that it has a spontaneous dissociation to form two OHs on the two Pd_3c sites on PdO(101), suggesting that PdO(101) has poor selectivity for H₂O₂ synthesis. For PdO@Pd₄Sn, its binding to the three molecules is significantly weaker than that on PdO(101). H₂ is physically adsorbed on the Pd_2c top sites (2.76 Å away from Pd_2c) with an adsorption energy of −0.11 eV, the adsorption of O₂ on the Pd_2c-Pd_2c bridge sites leads to small adsorption energy of −0.42 eV. H₂O₂ adsorbs on Pd_2c-Pd_2c bridge sites, yielding moderate adsorption energy of −0.63 eV. Likewise, as for the Pd₄Sn surface, H₂ is adsorbed on the top of Pd while the O₂ molecule connects with four surface Pd atoms, having the adsorption strengths as those located on between PdO(101) and PdO@Pd₄Sn. To gain in-depth insight into the relation between electronic structure and catalytic behavior, the projected density of states (PDOSs) of the surface Pd- ${d}_{{z}^{2}}$ states on PdO(101) and PdO@Pd₄Sn were calculated and compared since the ${d}_{{z}^{2}}$ states are considered to play an important role in molecule adsorption^35,36,37. As can be seen in Supplementary Fig. 25, the distribution of ${d}_{{z}^{2}}$ of surface Pd_3c on PdO is more localized than that of surface Pd_2c on PdO@Pd₄Sn. Furthermore, a sharp peak of ${d}_{{z}^{2}}$ states exists across the Fermi level on PdO(101), implying strong interaction between surface Pd and the reaction species via donating/accepting electrons. In contrast, on PdO@Pd₄Sn, the distribution of Pd-${d}_{{z}^{2}}$ state is less significant around the Fermi level, indicating the weaker binding of reaction species to Pd. It is therefore proposed that different adsorption behaviors of O₂, H₂, and H₂O₂ on the three models lead to different catalytic performances in the direct synthesis of H₂O₂.

In the following, we investigated the reaction mechanism of the O₂ reduction on PdO(101), Pd₄Sn, and PdO@Pd₄Sn. First, we calculated the activation of the most stable adsorption state of H₂, whose reaction pathways on these three models are shown in Fig. 5c. On PdO(101), H₂ dissociation occurs readily with an energy barrier of 0.09 eV. In the final state, the two dissociated H atoms are adsorbed on the Pd_3c-Pd_3c bridge site and the Pd_3c top site, respectively. In addition, it is found that the migration of H between two adjacent Pd_3c sites is facile with a small energy barrier of 0.11 eV, while the migration of H from the Pd_3c site to an adjacent oxygen site is difficult, requiring an energy barrier of 0.87 eV(Supplementary Fig. 26). It indicates that Pd-H hydride species are stable on PdO(101). On Pd₄Sn, H₂ activation on Pd₄Sn is barrierless, similar to the previous DFT findings on Pd(111) surface³⁸ (Supplementary Fig. 27). On PdO@Pd₄Sn, the cleavage of H-H bond in the most stable H₂ adsorbed has to overcome an energy barrier of 0.35 eV. Similar to the case on PdO(101), the Pd-H species is also stable because the energy barrier for H migration between Pd_2c sites (0.69 eV) is lower than that (1.07 eV) of its migration to oxygen sites (Supplementary Fig. 28). Therefore, H₂ readily dissociates on all of PdO(101), Pd₄Sn, and PdO@Pd₄Sn to produce stable Pd-H hydride, which provides a hydrogen source for the subsequent O₂ reduction (Fig. 5d). We also calculated other possible H₂ dissociation pathways but they possess higher energy barriers (Supplementary Figs. 29 and 30). Then, we investigate the O₂ reduction to H₂O₂ (energy profile in Fig. 5d and configurations in Supplementary Figs. 31 and 32). On the Pd₄Sn, the energy barriers of the first (*O₂ + *H → *OOH) and second (*OOH + *H → *H₂O₂) hydrogenation of O₂ is 0.55 eV and 0.65 eV with the reaction energies of 0.09 eV and −0.01 eV, respectively. On PdO@Pd₄Sn, the *O₂ + *H → *OOH reaction encounters an energy barrier of 0.53 eV, while the subsequent *OOH + *H → *H₂O₂ process has a lower energy barrier of 0.27 eV. This result indicates that the reaction of O₂ reduction to produce H₂O₂ has a lower rate-determining energy barrier on PdO@Pd₄Sn than that on Pd₄Sn, in agreement with the experimental observations that PdO@Pd₄Sn has better catalytic performance than Pd₄Sn prepared by one-step. We did not infer the reaction pathway on PdO(101) because H₂O₂ was proved to be unstable on this surface (Fig. 5b and Supplementary Fig. 25), leading to low H₂O₂ production.

In addition, we simulated three main side reactions of *O₂, *OOH, and *H₂O₂ dissociation on Pd₄Sn and PdO@PdSn, and the corresponding structures of IS, TS, and FS are shown in Supplementary Figs. 33 and 34. On Pd₄Sn(111), the *O₂ and *OOH dissociations are barrierless while the *HOOH dissociation has a small barrier of 0.07 eV (Supplementary Fig. 33), suggesting the poor selectivity of H₂O₂ on Pd₄Sn(111). On PdO@Pd₄Sn, the energy barriers of the three side reactions are 1.74, 0.53, and 0.32 eV (Supplementary Fig. 34), respectively, which are larger than those of the corresponding competitive reactions (0.53, 0.27, and 0.14 eV for O₂ hydrogenation, OOH hydrogenation, and H₂O₂ desorption, respectively, Supplementary Figs. 32, 34). Therefore, these results suggest the high selectivity of H₂O₂ on PdO@Pd₄Sn, which is consistent with the experimental observations (Fig. 2d and Table 1).

Selective production of H₂O₂ is a challenge in the direct H₂O₂ synthesis (DHS) from H₂ and O₂. In this contribution, we develop a two-step approach to prepare PdSn nanowires to efficiently catalyze direct H₂O₂ synthesis (DHS). This approach involves the first synthesis of the surface-rough Pd₄Sn nanowire (PdSn-NW) by a solvothermal method. Then Pd precursor was deposited onto the PdSn alloy nanowire (NW), followed by annealing in air. The as-prepared Pd_L/PdSn-NW via the two-step approach presents efficient reactivity with H₂O₂ producibility of >520 mol kg_cat⁻¹ h⁻¹ and selectivity of >95% in the direct production of H₂O₂ at zero Celcius. For the PdSn nanowire prepared by one-step method and PdSn nanoparticle catalysts, the catalyst surfaces are prone to be reduced in the flowing H₂/O₂ and show low H₂O₂ reactivity.

The excellent H₂O₂ production over the Pd_L/PdSn-NW catalyst is attributed to the presence of Pd oxide layer on the PdSn nanowires. The Pd oxide layer is stable against reduction or oxidation in the flowing H₂/O₂, while other PdSn nanoparticles and PdSn nanowire catalysts undergo a reduction in the presence of H₂/O₂. The layered Pd oxide enables less adsorption of oxygen/hydrogen and decreases the rupture of both O-O and H-H bonds, as well as less adsorption of peroxide produced, leading to the complete inhibition of the H₂O₂ hydrogenation and decomposition. Therefore, engineering the surface of PdSn nanowires via two-step approach can generate a layer of Pd oxide on the PdSn nanowire, presenting excellent reactivity in the direct H₂O₂ synthesis, which provides a promising strategy to design and develop highly active DHS catalyst.

Methods

Chemicals

Bis (acetylacetonato) palladium (II) (Pd(acac)₂, 99%), tin(II) acetate (Sn(Ac)₂, 95%), methanol (CH₃OH, HPLC grade), titanium dioxide (TiO₂, P25, 99%) were purchased from Sigma-Aldrich. Palladium(II) nitrate dihydrate (Pd(NO₃)₂·2H₂O, 99%), polyvinylpyrrolidone (PVP, MW = 58000), Tin(IV) chloride pentahydrate (SnCl₄·5H₂O, 99%) were purchased from J&K Scientific Ltd. Ethylene glycol (EG, analytical grade), N,N-Dimethylacetamide (DMAC, analytical grade), ammonium bromide (NH₄Br, analytical grade) were purchase from Sinopharm Chemical Reagent Co. Ltd.(Shanghai, China). The de-ionized water (DI H₂O,18 MΩ/cm) used in all experiments was obtained by passing through an ultra-pure purification system. All the chemicals were used without further purification.

Synthesis of unsupported PdSn nanowires via one-step

In a typical synthesis, the PdSn NWs (Pd₄Sn, marked as PdSn) were synthesized according to our previous report³². Briefly, 7.6 mg of Pd(acac)₂, 1.5 mg of Sn(Ac)₂, 15 mg of NH₄Br, 100 mg of PVP, 2 mL of DMAC, and 8 mL of EG were added into a vial (30 mL). And then, the mixture was sonicated for 15 min to ensure form a homogeneous solution. The solution was heated and maintained at 180 °C for 2 h in an oil bath. After that, the mixture was cooled down to room temperature. Finally, the precipitate was centrifuged and washed with ethanol/acetone mixture.

Synthesis of unsupported Pd_x/PdSn-NW and Sn_y/PdSn-NW via two-step

In a typical synthesis, the Pd_x/PdSn or Sn_y/PdSn NWs were synthesized by the following steps. Briefly, 7.6 mg of Pd(acac)₂, 1.5 mg of Sn(Ac)₂, 15 mg of NH₄Br, 100 mg of PVP, 2 mL of DMAC, and 8 mL of EG were added into a vial (30 mL). Then, the mixture was sonicated for 15 min to ensure form a homogeneous solution. The solution was heated and maintained at 180 °C for 2 h in an oil bath. After that, the mixture was cooled down to room temperature to achieve the unsupported PdSn nanowires. The role of NH₄Br in the synthesis of these worm structures is to induce the formation of the PdSn nanowires. Subsequently, the desired amount of Pd(acac)₂ or Sn(Ac)₂ solution (including 320 μL of EG and 80 μL of DMAC, e.g. 1.9 mg Pd(acac)₂) was added into the above solution. Next, the mixture was continuously stirred and held for 30 min at room temperature, followed by heating to 150 °C (hold for 2 h) with vigorous stirring. Then, the mixture was cooled down to ambient temperature again. Finally, the precipitate was centrifuged and washed with ethanol/acetone mixture. The precipitate achieved with the addition of 1.9 mg Pd precursor in the second step is denoted as unsupported Pd_L/PdSn-NW and the sample achieved with the addition of Sn(Ac)₂ solution in the second step is denoted as unsupported SnO_x/PdSn-NW.

Synthesis of unsupported Pd nanoparticles

The unsupported Pd NPs were synthesized from a procedure similar to that of PdSn nanowires synthesized via one-step, but without the addition of Sn(Ac)₂.

Synthesis of unsupported PdSn nanoparticles (PdSn-NP)

The bimetallic PdSn NP catalysts were synthesized according to a previous report³. In brief, 62.5 mg of Pd(NO₃)₂·2H₂O, 2 mL of DI water were added into a vial (30 mL). The solution was heated to 80 °C with continuous stirring. Then, 1 mL of DI water solution (including 73.8 mg of SnCl₄·5H₂O) was added to the vial and held for 15 min under stirring. Next, 0.6 g of TiO₂ and 1 mL of DI water were added to the solution, and the mixture was heated to 110 °C with continuous stirring in the air to evaporate the water. Subsequently, the solid sample was placed in a tube furnace and heated to 500 °C for 3 h under static air (heating rate of 10 °C/min). After cooling down, the solid sample was further heated to 200 °C for 2 h in a flow of 5% H₂/Ar (heating rate of 10 °C/min). Then, the solid sample was annealed in static air at 400 °C for 8 min or 4 h, respectively (heating rate of 10 °C/min). The resultant was denoted as PdSn-NP.

Synthesis of catalysts supported on TiO₂

Typically, the commercial TiO₂ material (anatase) and 6 mL of CH₃Cl were added into a 30 mL vial. The PdSn-NW or nanoparticle materials were firstly dispersed in 6 mL of CH₃Cl and ultrasonicated for about 5 min. The resulting mixture was added to the TiO₂ vial. Next, the mixture was continuously ultrasonicated for 30 min. Finally, the products were collected by centrifugation and dried at 60 °C overnight.

Annealing treatment of the supported catalysts

The as-synthesized catalysts were treated via a rapid annealing process. Firstly, the tube furnace was heated to 400 °C and held for 30 min under static air. Next, the appropriate amount of catalyst was placed in the porcelain boat and pushed into the tube furnace. The catalyst left for several minutes by annealing and was taken out (8 min) for cooling. The final products were denoted as Pd_L/PdSn-NW (supported on TiO₂, the “NW” represents nanowires) and collected for further characterization.

Catalyst characterization

The crystallographic structure of all the samples was determined by Powder XRD (Rigaku Ultima IV, Japan) patterns using a Cu Kα X-ray source (λ = 1.54056 Å). The morphology of all the samples was imaged by transmission electron microscope (TEM, JEM-1400, JEOL Co., Japan). The HRTEM, line-scan analysis, and elemental mapping were carried out on FEI Tecnai F30 electron microscope with an acceleration voltage of 200 kV. STEM images were obtained on an FEI Titan Themis Z, operating at 60-300 kV to observe the surface morphology of the materials. Before microscopy examination, the catalyst powders were ultrasonically dispersed in ethanol and then a drop of the solution was put onto a carbon-coated copper grid. The elemental content of all the samples was examined by scanning electron microscopy energy dispersive spectrometer (SEM-EDS, ZEISS Sigma, Germany) and inductively coupled plasma-atomic emission spectrometry (ICP-AES, iCap 7000, USA). The chemical compositions and valence states of all the samples were analyzed by XPS (K-Alpha, USA). NAP-XPS (SPECS Surface Nano Analysis GmbH, Germany) measurements were carried out on a SPECS system equipped with a differentially pumped Phoibos hemispherical electron energy analyzer using monochromatic Al Kα radiation (1486.6 eV). The H₂ or O₂ (99.999%) flow was introduced into the NAP cell and the total pressure was kept constant at 0.2 mbar via the electronic back-pressure regulator. X-ray Absorption Fine Structure (XAFS) of all samples was analyzed by TLS-01C beamline of the National Synchrotron Radiation Research Center (NSRRC, Hsinchu, Taiwan), and data were processed according to standard procedures using the Demeter program package (Version 0.9.24)³⁹.

Direct H₂O₂ synthesis

All the catalysts supported on TiO₂ were tested in the direct H₂O₂ synthesis. Direct H₂O₂ synthesis experiment was carried out in a stainless-steel semibatch autoclave with a volume of 50 mL. Briefly, 5 mg of supported catalyst, 5.54 g of MeOH, and 3.0 g of H₂O (both HPLC grade) were added to the autoclave. Then purged three times with O₂ (0.4 MPa) to the autoclave, and with filled with O₂ (0.4 MPa) and 5% H₂/Ar (3.6 MPa) until the total pressure of 4.0 MPa at room temperature. All experiments were performed in an ice-water bath and kept continuous stirring (1200 rpm) of reaction time for 15 min. The reaction was carried out at zero Celsius because it is much safer and easier to carry out in the lab. We also investigated the reactivity of the catalyst at room temperature for comparison. The producibility of the Pd_L/PdSn-NW catalyst at room temperature is slightly lower than that performed at zero Celcius in the direct H₂O₂ synthesis (Supplementary Fig. 35). Control experiments are also performed by varying the stirring rate from 100-1200 rpm, and varying the catalyst mass from 3 to 10 mg using autoclaves having different volumes (50 mL and 100 mL). All these experiments show similar H₂O₂ producibility under these conditions. Therefore, it indicates that there are no interphase mass transfer constraints in the reaction under the present conditions (Supplementary Fig. 36).

Quantification of gas product

The gas products were detected via a gas chromatograph (GC, 8890, Agilent) equipped with a thermal conductivity detector (TCD) and a MolSieve 5 A packing column (G3591-80022). Typically, after the direct H₂O₂ synthesis process, the autoclave was directly connected into the GC. The gas products were purged into GC by Ar.

The H₂ conversion was calculated as follows:

$${{{{{{\rm{H}}}}}}}_{2}{{{{{\rm{conversion}}}}}}=\frac{n{({H}_{2})}_{{in}}\,-\,n{({H}_{2})}_{{out}}}{n{({H}_{2})}_{{in}}}\times 100\%$$

(1)

H₂O₂ productivity evaluation

The H₂O₂ productivity was analyzed by titrating with acidified Ce(SO₄)₂ (0.05 M) and using ferroin (about 100 µL) as indicator^3,40. The acidified Ce(SO₄)₂ solutions were standardized against (NH₄)₂Fe(SO₄)₂·6H₂O using ferroin as an indicator. The H₂O₂ productivity and H₂O₂ selectivity were calculated based on the equations below:

$${{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2}+{2{{{{{\rm{Ce}}}}}}}^{4+}\to {2{{{{{\rm{Ce}}}}}}}^{3+}+{2{{{{{\rm{H}}}}}}}^{+}+{{{{{{\rm{O}}}}}}}_{2}$$

$${{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2}{{{{{\rm{selectivity}}}}}}=\frac{n{({H}_{2}{O}_{2})}_{{out}}}{n{({H}_{2}{O}_{2})}_{{out}}+n{({H}_{2}O)}_{{out}}}\times 100\%$$

(2)

$${{{{{\rm{Where}}}}}}\,{{{{{{\rm{n}}}}}}({{{{{{\rm{H}}}}}}}_{2}{{{{{\rm{O}}}}}})}_{{{{{{\rm{out}}}}}}}={{{{{{\rm{n}}}}}}({{{{{{\rm{H}}}}}}}_{2})}_{{{{{{\rm{converted}}}}}}}-{{{{{{\rm{n}}}}}}({{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2})}_{{{{{{\rm{out}}}}}}}$$

(3)

H₂O₂ hydrogenation and decomposition

H₂O₂ hydrogenation and decomposition were carried out using similar procedures to direct H₂O₂ synthesis. Typically, for H₂O₂ hydrogenation test (H₂O₂ + H₂ → 2H₂O), 5 mg of supported catalyst (support is TiO₂), 5.54 g of MeOH, 2.0 g of H₂O and 1.1 g H₂O₂ (30%) were added in the autoclave. Then, 3.6 MPa of H₂/Ar (5% H₂) was filled into the autoclave. Afterward, the reaction time extended until 30 min. For the H₂O₂ decomposition test (H₂O₂ → H₂O + O₂), 5 mg of catalyst, 5.54 g of MeOH, 2.0 g of H₂O, and 1.1 g H₂O₂ (30%) were added to the autoclave. Then, 3.6 MPa of N₂ was filled into the autoclave. Subsequently, the reaction was carried out for 30 min. The H₂O₂ hydrogenation or decomposition rate was calculated based as follows:

$${{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2}{{{{{\rm{hydrogenation}}}}}}/{{{{{\rm{decomposition}}}}}}=\frac{n{({H}_{2}{O}_{2})}_{{in}}-n{({H}_{2}{O}_{2})}_{{out}}}{{W}_{{catalyst}}\times t}\times 100\%$$

(4)

Computational details

All spin-polarized DFT calculations were carried out by using the Vienna Ab initio Simulation Package. Interactions between ion cores and valence electrons were described by the projected augmented wave method and the exchange-correlation function was by the Perdew–Burke–Ernzerhof functional based on generalized gradient approximation⁴¹. Wavefunctions were expanded by the plane wave basis with a cutoff energy of 400 eV⁴². The van der Waals interactions were included via DFT-D2 in Grimme’s scheme⁴³. Transition states were determined by the Climbing Image Nudged Elastic Band and Dimer method^44,45, and charge states were obtained from Bader Charge analysis⁴⁶. For all models (crystallographic information files of Pd₄Sn, PdO@Pd₄Sn, and PdO(101) were provided in supplementary materials as Supplementary Data 1–3, respectively), Monkhorst–Pack mesh with 2 × 2 × 1 grid was adopted to sample the Brillouin zone⁴⁷. Convergence criteria for structural optimizations were set to 10⁻⁵ eV and 0.02 eV/Å for energy and force, respectively. And for transition states searching, the criterion for force was set to 0.05 eV/Å. The (2 × 3)-PdO(101) surface was cleaved from the optimized PdO unitcell (a = b = 3.056 Å, c = 5.381 Å, in agreement with experimental values (a = b = 3.043 Å and c = 5.335 Å)), containing 4 O-Pd-O tri-atomic layers. The bottom two layers were fixed while the other layers with surface species were fully relaxed in structural optimization. The lengths of a and b of supercell are 12.38 and 9.17 Å, respectively. Since the Pd unit is composed of 4 Pd atoms, uniform Pd_xSn_y can only be constructed as Pd₂Sn₂ and Pd₃Sn. In addition, our experiments show that nanowires expose more Pd on the surface, we, therefore, simulate the slab with 4 layers of uniform Pd₃Sn covered with one layer of Pd (Supplementary Fig. 37). Then, the total Pd/Sn ratio equals 4:1. For all slabs, a vacuum space of 15 Å is adopted along the z-direction to avoid the interaction between periodic images.

The Gibbs free energy (G) is obtained by VASPKIT code⁴⁸ via the following formula:

$$G=E+{E}_{{ZPE}}+{\varDelta U}_{0\to T}-T\times S$$

(5)

in which E is the electronic energy of DFT calculations, E_ZPE and S are the zero-point energy and entropy computed from vibrational frequency analysis. T was the temperature adopted in our experiment, 273.15 K. ΔU_0→T is the internal energy difference between 0 and T K.