Pd@PdBO_x heterostructure with strong electronic interaction to promote C─H bond activation for glycerol oxidation reaction

Abstract

Glycerol oxidation reaction (GOR) represents an economical pathway for transforming renewable feedstock to value-added chemicals. However, the inertness of C(sp³)─H bonds of glycerol and intermediates results in the high energy barrier of the dehydrogenation step, relating to poor product selectivity at high glycerol conversion. Here, a carbon nanotube–supported PdBO_x@Pd heterostructure catalyst (PdBO_x@Pd/CNTs) was synthesized in which in situ–exsoluted PdBO_x clusters covalently covered Pd nanoparticles, thus yielding strong electronic interaction between Pd nanoparticles and PdBO_x clusters. The strong electronic interaction in PdBO_x@Pd/CNTs induces the hybridization between Pd(d), B(s, p), and O(s, p) atom orbits, optimizing the adsorption of reactants and intermediates, thus enhancing the activity for the GOR. The density functional theory calculation result reveals that the strong electronic interaction in PdBO_x@Pd/CNTs facilitates the hydrogen transfer in the primary C─H bond of the CH₂OHCHOHCH₂O* intermediate, thus reducing the energy barrier of the rate-determining step and improving glyceric acid selectivity toward the GOR.

PdBO_x@Pd with strong electronic interaction shows excellent catalytic performance in converting glycerol to glyceric acid.

INTRODUCTION

Glycerol, a representative polyol, is a major by-product of the biofuel industry while being the raw material for hundreds of biomass-derived intermediates (1, 2). The global output of glycerol has reached several million tonnes at present and continues to grow every year (3, 4). According to the US Department of Energy, the price of glycerol is relatively low, only 0.24 to 0.6 USD/kg, because the production of glycerol exceeds demand (5). Glycerol oxidation reaction (GOR) as a promising method produces several value-added chemicals, including formic acid (FA), glyceric acid (GLYA), glyceraldehyde (GLYHD), lactic acid (LA), and dihydroxyacetone (6–8). As a high value-added glycerol derivative, GLYA is a crucial raw material in the applications of the textile, food, and pharmaceutical industry. However, the industrialization of the GOR for producing GLYA has been obstructed because of the poor glycerol conversion rate (≤70%) and low single product selectivity (≤50%), which is difficult to reach the threshold of economic feasibility predicted by economic assessment (7, 9, 10).

The GOR typically requires the accurate site activation of C(sp³)─H bonds for transforming glycerol to a GLYA product. However, the inertness of C(sp³)─H bonds in glycerol/intermediates leads to the high energy barrier for the dehydrogenation step during the GOR (11, 12). Furthermore, the similar electronegativity of carbon and hydrogen results in small polarity in the C(sp³)─H bonds of glycerol/intermediates; thus, it is challenging to achieve C(sp³)─H bond activation under mild conditions. Recent advances have identified that metal oxide–supported transition metal–based catalysts (transition metal/metal oxide) play a crucial role in yielding strong electronic interaction and transferring the spillover hydrogen of C(sp³)─H bonds for reactants and intermediates (13–15) when the metal oxide and transition metal act as a cocatalyst and active site, respectively. Once the transition metal/metal oxide–based catalysts with a suitable structure are constructed, the strong electronic interaction between two components may induce the orbital hybridization for the interfacial atoms, optimize the adsorption behaviors of reactants/intermediates, and thus improve catalytic performance (16). Yan and co-workers (17) reported that the electron is transferred from Cu₂O to h-BN (hexagonal boron nitride) in the h-BN-Cu₂O catalyst with strong electronic interaction, which stabilizes the Cu⁺ species and promotes the adsorption of the CO* intermediate, thus enhancing the ratio of C₂H₄/CO products toward the CO₂ electroreduction reaction. However, it is difficult to obtain the compact contact between the transition metal as an active site and the metal oxide as a cocatalyst for transition metal/metal oxide–based catalysts (18). Hence, metal nanoparticles are easily detached from the surface of metal oxide in metal oxide–supported metal-based catalysts during the process of catalyzing the GOR. Even if the transition metal is used as a substrate to stabilize the metal oxide cocatalyst to form transition metal–supported metal oxide–based catalysts (metal oxide/transition metal), its surface finds it hard to directly hold the metal oxide during the synthesis process; thus, it also faces huge challenges in constructing strong electronic interaction between the metal and metal oxide.

In this work, we design and synthesize a PdBO_x@Pd/CNTs catalyst by encapsulating Pd nanoparticles with in situ–exsoluted PdBO_x clusters derived from the oxidation of interstitial boron atoms in the Pd lattice. The PdBO_x@Pd/CNTs catalyst achieved excellent catalytic performance with a high GLYA selectivity reaching 73.5% at a glycerol conversion of 99.1% toward the GOR. The PdBO_x@Pd/CNTs catalyst has the strong electronic interaction between PdBO_x and Pd because in situ–exsoluted PdBO_x clusters covalently covered the surface of Pd nanoparticles. The strong electronic interaction between PdBO_x clusters and Pd nanoparticles in PdBO_x@Pd/CNTs promotes the hydrogen transfer from the primary C─H bond of CH₂OHCHOHCH₂O* and CH₂OHCHOHCHOOH* intermediates to the absorbed hydroxyl on the catalyst surface, thus facilitating the activation of the C─H bond, lowering the energy barrier of the rate-determining step (RDS), and enhancing the GLYA selectivity toward the GOR. Besides, the strong electronic interaction between PdBO_x and Pd induces the hybridization between Pd(d), B(s, p), and O(s, p) atomic orbitals, thus leading to the upshift of the Pd d-band center for the PdBO_x@Pd/CNTs catalyst. The upshifted d-band center enhances the adsorption of multiple reactants (glycerol, H₂O, and O₂) and intermediates (GLYHD and hydroxyl), thus promoting the activity toward the GOR. These findings provide a guidance direction for designing a highly efficient catalyst and deepening insight into the catalytic mechanism of supported catalysts toward the GOR.

RESULTS

Synthesis and characterizations of the PdBO_x@Pd/CNTs catalyst

To form a metal oxide@transition metal catalyst with strong electronic interaction, we first consider the boron in p-block elements, which has a small atomic radius and is extremely easy to oxidize (19). The boron element can form an interstitial alloy with a transition metal because of the nature of the small atomic radius of boron atoms, which enables it to occupy the lattice interstitial sites of the transition metal. The interstitial boron atoms, in situ diffused from interstitial sites, transform into BO_x clusters onto the transition metal surface, which covalently bond with surface metal atoms to form the ABO_x clusters (A represents a transition metal element) (20, 21). However, the precise synthesis of transition metal-boron alloy is a great challenge. Figure 1A shows the comparison of the boron-doping energy in the octahedral interstice sites of 23 transition metals. It is worth noting that the boron-doping energy in Pd shows a larger negative value compared with those of other 22 transition metals. This result proves that a more negative formation energy is favorable, and it provides a possibility for synthesizing Pd─B alloy.

Fig. 1. Synthesis and structural characterizations of PdBO_x@Pd/CNTs.

(A) Comparison of boron-doping energy in 23 kinds of transition metals. (B) Schematic illustration for the synthetic procedure of the PdBO_x@Pd/CNTs catalyst. (C) HRTEM image of the PdBO_x@Pd/CNTs catalyst. (D and E) Normalized Pd K-edge XANES spectra and FT-EXAFS spectra of Pd foil, PdO, Pd/CNTs, PdBO_x-Pd/CNTs, and PdBO_x@Pd/CNTs catalysts. a.u., arbitrary units. (F and G) PDF patterns and valence band photoemission spectra of Pd/CNTs, PdBO_x-Pd/CNTs, and PdBO_x@Pd/CNTs catalysts.

On the basis of the above consideration, we first developed a solvothermal method referring to the previous report to synthesize Pd─B alloy/CNTs by using carbon nanotubes, sodium tetrachloropalladate (Na₂PdCl₄), and borane tetrahydrofuran (BH₃-THF) as support, Pd, and boron sources, respectively (19–21). A carbon nanotube has high specific surface area, which is beneficial to avoid the aggregation problem, thus improving the catalytic performance for the GOR. The PdBO_x@Pd/CNTs catalyst with strong electronic interaction was successfully obtained when boron atoms were in situ exsoluted from the lattice interstice of Pd in Pd─B alloy/CNTs calcined at 150°C under air atmosphere (Fig. 1B). Simultaneously, the exsoluted boron atoms were transformed into BO_x clusters on the Pd surface, which further covalently bond with the surface Pd atom to form PdBO_x, as demonstrated by the high-resolution transmission electron microscopy (HRTEM) image (Fig. 1C). A PdBO_x-Pd/CNTs catalyst with weak electronic interaction can also be successfully obtained by replacing BH₃-THF with boric acid (H₃BO₃) as the boron source. When PdBO_x is formed by in situ melting out the boron atoms from the lattice of interstitial Pd─B alloy, the Pd nanoparticles are closely and partly coated by ultrasmall PdBO_x clusters in the PdBO_x@Pd/CNTs catalyst (Fig. 1C), thus leading to the strong covalent interaction between PdBO_x clusters and Pd nanoparticles. However, for the PdBO_x-Pd/CNTs catalyst, PdBO_x is obtained from the decomposition of H₃BO₃, so the PdBO_x clusters are weakly bound on the surface region of Pd nanoparticles (fig. S1). This induces the weak covalent interaction between PdBO_x clusters and Pd nanoparticles in the PdBO_x-Pd/CNTs catalyst. This covalent link between PdBO_x clusters and Pd nanoparticles induces the strong electronic interaction in the PdBO_x@Pd/CNTs catalyst, while the PdBO_x-Pd/CNTs catalyst exhibits a weak electronic interaction. HRTEM images of PdBO_x@Pd/CNTs and PdBO_x-Pd/CNTs clearly show that their lattice spaces for Pd nanoparticles are 0.231 nm, which corresponds to the Pd (111) facet (the inset of Fig. 1C and figs. S1 and S2). Scanning transmission electron microscopy and the corresponding maps reveal that Pd, O, and B elements are collectively distributed throughout clusters on the surface of Pd nanoparticles, suggesting the formation of PdBO_x in PdBO_x@Pd/CNTs and PdBO_x-Pd/CNTs catalysts (figs. S3 and S4). Combining scanning transmission electron microscopy maps and in situ HRTEM results further illustrates the exsolution of interstitial B atoms from the interior of initial PdB alloy onto the Pd surface, which simultaneously transforms into BO_x clusters and further covalently bonds with the surface Pd atom to form PdBO_x (figs. S3 and S5).

The inductively coupled plasma atomic emission spectroscopy result demonstrates that the molar ratio of B to Pd in PdBO_x-Pd/CNT (~1.09) and PdBO_x@Pd/CNT (~1.01) catalysts is similar (table S2). Pd nanoparticles for Pd/CNTs, PdBO_x@Pd/CNTs, and PdBO_x-Pd/CNTs catalysts all display similar size distributions in the range of 3 to 8 nm (fig. S6). Besides, the size distribution of PdBO_x nanoclusters in PdBO_x@Pd/CNTs and PdBO_x-Pd/CNTs catalysts is similar in the range of 0.2 to 0.9 nm (fig. S7). Therefore, the size effect of Pd and PdBO_x in Pd/CNTs, PdBO_x@Pd/CNTs, and PdBO_x-Pd/CNTs catalysts can be excluded (figs. S6 and S7). Powder x-ray diffraction (XRD) patterns exhibit the diffraction peaks of Pd in Pd/CNTs, PdBO_x-Pd/CNTs, and PdBO_x@Pd/CNT catalysts (fig. S8), while that of PdBO_x is not observed, suggesting that PdBO_x is in the form of small clusters. The synchrotron XRD patterns, obtained from their atomic pair distribution function (PDF), show that the peak intensity of metallic Pd exhibits a decreased trend in the order of Pd/CNTs > PdBO_x-Pd/CNTs > PdBO_x@Pd/CNT (fig. S9). This implies that PdBO_x clusters likely partially mask the signal of Pd in PdBO_x@Pd/CNT because PdBO_x closely covers the surface of Pd.

Next, we explored the chemical state of boron to further determine the electronic structure of the PdBO_x@Pd/CNTs catalyst. High-resolution solid-state nuclear magnetic resonance analysis results show a typical characteristic peak at 30 ppm (parts per million) for PdBO_x@Pd/CNTs and PdBO_x-Pd/CNTs catalysts, which is attributed to BO_x (fig. S10) (21). This result demonstrates that the boron element in PdBO_x@Pd/CNTs is in an oxidized state rather than a metallic state. In the B 1s x-ray photoelectron spectroscopy (XPS) spectrum of PdBO_x-Pd/CNTs and PdBO_x@Pd/CNTs catalysts, a distinct characteristic peak at 191.7 eV is clearly observed, further proving that boron exists primarily in an oxidized state (fig. S11) (22). Figure S12 shows the Pd 3d XPS spectra of Pd/CNTs, PdBO_x-Pd/CNTs, and PdBO_x@Pd/CNTs catalysts. The binding energy of Pd 3d_5/2 for the PdBO_x@Pd/CNTs catalyst is higher than those of Pd/CNTs and PdBO_x-Pd/CNTs catalysts, implying the strong electronic interaction between PdBO_x clusters and Pd nanoparticles in the PdBO_x@Pd/CNTs catalyst (23). The XPS spectra illustrate that the Pd⁰ content exhibits a decreased trend in the order of Pd/CNTs (~67.4%), PdBO_x-Pd/CNTs (~47.4%), and PdBO_x@Pd/CNTs (~15.4%) catalysts, while the Pd^δ+ (0 < δ < 2) content in PdBO_x@Pd/CNTs (74.9%) is much higher than that in Pd/CNTs (19.1%) and PdBO_x-Pd/CNTs (42.7%) (table S3). This result suggests that the PdBO_x@Pd/CNTs catalyst has strong electronic interaction between PdBO_x and Pd.

We used x-ray absorption near-edge structure (XANES) spectroscopy to study the structure of PdBO_x and the corresponding effect on the regulation of the electronic structure for the metallic Pd in the PdBO_x@Pd/CNTs catalyst. The normalized Pd K-edge XANES spectra of Pd foil, PdO standard, Pd/CNTs, PdBO_x-Pd/CNTs, and PdBO_x@Pd/CNTs catalysts are shown in Fig. 1D. The energy of the adsorption edge (E₀) for the PdBO_x@Pd/CNTs catalyst is lower than that for PdO and higher than that for Pd foil, implying the formation of Pd^δ+ species. Compared with Pd/CNTs and PdBO_x-Pd/CNTs catalysts, the E₀ position for PdBO_x@Pd/CNTs shifts toward a higher energy direction, suggesting the stronger electronic interaction between PdBO_x clusters and Pd nanoparticles. PdBO_x-Pd/CNTs and PdBO_x@Pd/CNTs catalysts exhibit a peak of Pd─O/Pd─B bond at ~1.7 Å and another peak of Pd─Pd at ~2.5 Å, suggesting that they consist of a mixture of Pd─Pd, Pd─B, and Pd─O atom coordination (Fig. 1E and fig. S13) (24). Besides, this also indicates that the B in the PdBO_x@Pd/CNTs catalyst is simultaneously bonded with Pd and O atoms, thus suggesting the formation of PdBO_x. It is observed that there is a gradual increase in the Pd─O/Pd─B bond contribution as well as a corresponding decrease in the Pd─Pd bond contribution in the order of Pd/CNTs, PdBO_x-Pd/CNTs, and PdBO_x@Pd/CNTs catalysts (Fig. 1E and table S4). The coordination number of the Pd─O bond for the PdBO_x@Pd/CNTs catalyst (2.7 ± 0.3) is higher than those of Pd/CNTs (0.6 ± 0.2) and PdBO_x-Pd/CNTs (1.4 ± 0.4) catalysts, while it is lower than that of the PdO (4.0) reference (fig. S14 and table S4). It is worth noting that the intensity of the Pd─O/Pd─B bond as well as the corresponding coordinate number for the PdBO_x@Pd/CNTs catalyst is higher than those of Pd/CNTs and PdBO_x-Pd/CNTs catalysts, demonstrating the strong electronic interaction between PdBO_x and Pd in PdBO_x@Pd/CNTs (Fig. 1E) (17). Besides, Fourier-transformed k-space and corresponding wavelet-transformed k-space results of Pd K-edge XANES for Pd/CNTs, PdBO_x-Pd/CNTs, and PdBO_x@Pd/CNTs catalysts exhibit clear visualization for the presence of Pd─Pd and Pd─O/Pd─B bonds (figs. S14 and S15).

PDF patterns (Fig. 1F) show that the additional peak around ~2.38 Å is observed for PdBO_x@Pd/CNTs and PdBO_x-Pd/CNTs catalysts compared with the Pd/CNTs catalyst, and this peak was attributed to the Pd─B/Pd─O bond. Besides, the peak intensity for the second shell of the Pd─Pd bond exhibits a decreased trend that correlated well with the magnitude of strong electronic interaction as the capped effect of PdBO_x on Pd in PdBO_x@Pd/CNTs. Valence band photoemission spectra show that the d-band centers of Pd/CNTs, PdBO_x-Pd/CNTs, and PdBO_x@Pd/CNTs catalysts are −6.52, −6.37, and −6.23 eV, respectively (Fig. 1G). The d-band center of the PdBO_x@Pd/CNTs catalyst moved up compared with those of PdBO_x-Pd/CNTs and Pd/CNTs catalysts. This confirms that the strong electronic interaction between PdBO_x clusters and Pd nanoparticles leads to the upshift of the d-band center for the PdBO_x@Pd/CNTs catalyst, thus enhancing the binding strength between active sites and reactants/intermediates toward the GOR. These results further identify that PdBO_x@Pd/CNTs as a GOR catalyst was efficient for the enhancement of the catalytic activity (25, 26).

GOR performance on PdBO_x@Pd/CNTs with strong electronic interaction

We next assessed the catalytic activity and product selectivity for the PdBO_x@Pd/CNTs catalyst with strong electronic interaction toward the GOR compared with Pd/CNTs and PdBO_x-Pd/CNTs catalysts. As shown in Fig. 2A, Pd/CNTs and PdBO_x-Pd/CNTs catalysts obtain the glycerol conversion values of 47.5 and 42.1% under 60°C at 2 hours, respectively. The increased conversion of glycerol for PdBO_x@Pd/CNTs (94.9%) suggests that the strong electronic interaction between Pd nanoparticles and PdBO_x clusters greatly promotes the enhancement of catalytic performance.

Fig. 2. Catalytic performance of PdBO_x@Pd/CNTs toward the GOR.

(A) Glycerol conversion profiles of Pd/CNTs, PdBO_x-Pd/CNTs, and PdBO_x@Pd/CNTs catalysts. Reaction conditions: glycerol aqueous solution (200 mM), n(GLY)/n(Pd) = 200, 200 mg of NaOH, 60°C, and O₂ flow at a rate of 150 ml/min. (B and C) Product selectivity at ~75% glycerol conversion and TOF values at 15 min of Pd/CNTs, PdBO_x-Pd/CNTs, and PdBO_x@Pd/CNTs catalysts. (D) Comparison of GLYA selectivity and glycerol conversion for the PdBO_x@Pd/CNTs catalyst and some representative Pd-based catalysts. (E) Catalytic stability of Pd/CNTs, Pd-PdBO_x/CNTs, and PdBO_x@Pd/CNTs catalysts with five cycles.

High-performance liquid chromatography (HPLC) analysis was performed to investigate the product selectivity of all catalysts toward the GOR. As shown in Fig. 2B, the product distribution for three catalysts is GLYA, LA, TA (tartaric acid), OA (oxalic acid), GA (glycolic acid), and FA, and the main product is GLYA. The FA selectivity for the PdBO_x@Pd/CNTs catalyst (1.6%) is lower than those of Pd/CNTs (2.6%) and PdBO_x-Pd/CNTs (14.5%) catalysts at a similar glycerol conversion of ~75%, suggesting that the strong electronic interaction between PdBO_x clusters and Pd nanoparticles efficiently reduces the breaking of the C─C bond of glycerol and further promotes the generation of C3 products. Meanwhile, the carbon balance of PdBO_x@Pd/CNTs (99%) is higher than those of Pd/CNTs (64%) and PdBO_x-Pd/CNTs (85%), demonstrating that the strong electronic interaction between PdBO_x and Pd inhibits the breakage of the C─C bond in glycerol reactant/intermediates during the GOR and thus improves the selectivity of C3 products (fig. S16). Such a low carbon balance for the Pd/CNTs catalyst (64%) demonstrates that the formed FA product was oxidized into CO₂ during the GOR. Compared with the Pd/CNTs catalyst, the higher carbon balance for PdBO_x-Pd/CNTs (85%) means that the formed FA product was not further oxidized into CO₂. This result demonstrates that although PdBO_x-Pd/CNTs with weak electronic interaction between PdBO_x and Pd does not suppress the breakage of the C─C bond in glycerol/intermediates during the GOR, it plays a crucial role in suppressing the further oxidation of the FA product. Compared with Pd/CNTs (60.0%) and PdBO_x-Pd/CNTs (57.8%) catalysts, the PdBO_x@Pd/CNTs catalyst (75.0%) has higher GLYA selectivity (Fig. 2B and fig. S17) at a similar glycerol conversion of ~75%. Therefore, the GLYA yield of the PdBO_x-Pd/CNTs catalyst (36.8%) is higher than that of Pd/CNTs (28.8%) when we consider the carbon balance for the GOR (fig. S18). Although the glycerol conversion is close to 100% (99.1%) for 4 hours, the high GLYA product selectivity of 73.5% can also be obtained for the PdBO_x@Pd/CNTs catalyst toward the GOR (fig. S19). Besides, PdBO_x@Pd/CNTs exhibits a lower selectivity of LA compared with Pd/CNTs and PdBO_x-Pd/CNTs. Typically, when the key intermediate of GLYHD is stably adsorbed on the surface of the catalyst, the selectivity for GLYA would be enhanced (fig. S20). On the contrary, when the key intermediate of GLYHD* is easily removed from the catalyst surface, the selectivity of LA would be enhanced (27). These results further proved that the strong electronic interaction between PdBO_x clusters and Pd nanoparticles in PdBO_x@Pd/CNTs most likely facilitates the adsorption and dehydrogenation of the key GLYHD intermediate on the catalyst surface, thus promoting the catalytic pathway from glycerol to GLYA.

To further assess the role of PdBO_x in PdBO_x-Pd/CNTs for the enhanced catalytic performance toward the GOR, a series of control experiments was conducted. The glycerol conversion and GLYA product selectivity for the PdBO_x@Pd/CNTs catalyst are higher than those of the original PdB/CNTs, suggesting the enhanced effect of PdBO_x clusters on the catalytic performance toward the GOR (fig. S21). Figure S22 shows the product selectivity and glycerol conversion of PdBO_x@Pd/CNTs catalysts with different PdBO_x contents. The PdBO_x@Pd/CNTs (Pd/B = ~1/1) catalyst with a Pd:B ratio of 1:1 exhibits similar glycerol conversion but much higher GLYA selectivity compared with PdBO_x@Pd/CNTs-1 (Pd/B = ~1/0.75) and PdBO_x@Pd/CNTs-2 (Pd/B = ~1/1.25) catalysts.

As shown in table S5, there was little difference in the active site density and metal dispersion for the PdBO_x@Pd/CNTs, PdBO_x-Pd/CNTs, and Pd/CNTs catalysts, indicating that the increase in catalytic activity was due to the adjustment of the electronic structure of the catalyst rather than the variation in the number of active sites. As shown in Fig. 2C, under the same conditions, the time-of-flight (TOF) value of the PdBO_x@Pd/CNTs catalyst (796.2 hour⁻¹) is higher than those of PdBO_x-Pd/CNTs (260.9 hour⁻¹) and Pd/CNTs (280.3 hour⁻¹) catalysts, further demonstrating the excellent intrinsic activity of the PdBO_x@Pd/CNTs catalyst for catalyzing the GOR. From the above results, the PdBO_x@Pd/CNTs catalyst exhibited superior catalytic activity and a high yield (70.5%) of GLYA product for 4 hours toward the GOR, surpassing most of the representative catalysts reported so far (Fig. 2D and table S6) (28).

As shown in Fig. 2E, the PdBO_x@Pd/CNTs catalyst presents superior cycle stability in the GOR, measured by referring to the previous work (29). The catalytic performance of the PdBO_x@Pd/CNTs catalyst has no obvious change under the condition approaching 100% glycerol conversion. In comparison, the glycerol conversion for Pd/CNTs and PdBO_x-Pd/CNTs catalysts toward the GOR decreased by 10% after five catalytic cycles. Besides, the PdBO_x@Pd/CNTs catalyst also exhibits more excellent catalytic stability than those of Pd/CNTs and PdBO_x-Pd/CNTs catalysts under the condition of low glycerol conversion (fig. S23). The inductively coupled plasma atomic emission spectroscopy result indicates that the PdBO_x@Pd/CNTs catalyst exhibits little Pd and B loss after GOR, suggesting superior structural stability (table S7). At the same time, after the stability test, the XRD pattern, Pd 3d XPS spectrum, and HRTEM image of the PdBO_x@Pd/CNTs catalyst show no obvious change, suggesting that the PdBO_x@Pd/CNTs catalyst has excellent structural stability (figs. S24 to S26).

Investigating adsorption behaviors of the reactant, intermediate, and product toward the GOR

Because the GOR is a complex catalytic reaction involving many reactants and intermediates, multiple catalytic paths proceed concurrently (Fig. 3A and fig. S27). In this process, oxygen can promote H₂O dissociation to form OH* species (step 1). The resultant OH* species on the catalyst surface facilitates the deprotonation of the primary O─H bond for adsorbed glycerol (step 2) and the subsequent activation of the primary C─H bond to form the GLYHD* intermediate (step 3). Subsequently, the adsorbed OH* bonds with GLYHD* to form CH₂OHCHOHCHOOH* (step 4). Last, the adsorbed OH* facilitates the dehydrogenation of the formed CH₂OHCHOHCHOOH* to produce the GLYA product (step 5). Therefore, it is essential to design a highly efficient catalyst for ensuring the robust binding of reactants (glycerol, water, and O₂) and key intermediate species (OH* and GLYHD*). Simultaneously, the catalyst should exhibit weak adsorption of GLYA*, which is conducive to the generation of the GLYA product and the reexposure of active sites.

Fig. 3. Adsorption behavior of reactants, intermediate, and product on the PdBO_x@Pd/CNTs catalyst toward the GOR.

(A) Reaction pathway for glycerol oxidation to GLYA. (B) Kinetic isotope profiles of Pd/CNTs, PdBO_x-Pd/CNTs, and PdBO_x@Pd/CNTs catalysts. (C) Schematic diagram of adsorption behavior for glycerol, O₂, H₂O, hydroxyl intermediate, GLYHD intermediate, and GLYA product on the surface of the PdBO_x@Pd/CNTs catalyst. (D) Arrhenius plots of Pd/CNTs, PdBO_x-Pd/CNTs, and PdBO_x@Pd/CNTs catalysts.

The contact angle measurement results of glycerol, water, GLYHD, and GLYA on Pd/CNTs, PdBO_x-Pd/CNTs, and PdBO_x@Pd/CNTs catalysts prove that the PdBO_x@Pd/CNTs catalyst not only easily adsorbs glycerol, GLYHD intermediate, and water but also is efficient to remove the GLYA product compared with Pd/CNTs and PdBO_x-Pd/CNTs catalysts (figs. S28 to S31) (30, 31). The adsorbed water measurement result reveals that the PdBO_x@Pd/CNTs catalyst has faster water adsorption kinetics than Pd/CNTs and PdBO_x-Pd/CNTs catalysts (fig. S32). O₂ temperature-programmed desorption data reveal that the PdBO_x@Pd/CNTs catalyst has an obvious O₂ desorption peak, while Pd/CNTs and PdBO_x-Pd/CNTs catalysts have no obvious O₂ desorption peak (fig. S33). This result proves that PdBO_x@Pd/CNTs as a GOR catalyst can facilitate the adsorption and activation of O₂ (fig. S34) (32). The electron paramagnetic resonance spectroscopy test result reveals that the signal intensity of the PdBO_x@Pd/CNTs catalyst is higher than those of Pd/CNTs and PdBO_x-Pd/CNTs catalysts, suggesting that the PdBO_x@Pd/CNTs catalyst has a strong adsorption capacity for hydroxyl radicals (fig. S34) (33).

Figure 3B shows the isotopic kinetic data of PdBO_x@Pd/CNTs, PdBO_x-Pd/CNTs, and Pd/CNTs catalysts. The K_H/K_D value of the PdBO_x@Pd/CNTs catalyst (1.27) is higher than those of Pd/CNTs (1.05) and PdBO_x-Pd/CNTs (1.16), which proves that the introduction of PdBO_x clusters in the PdBO_x@Pd/CNTs catalyst promotes the breakage of the O─H bond in H₂O, thus facilitating the oxidation of glycerol and the formation of GLYA (34). We also tested CO stripping curves of Pd/CNTs, PdBO_x-Pd/CNTs, and PdBO_x@Pd/CNTs catalysts under alkaline conditions (fig. S35). Compared with Pd/CNTs and PdBO_x-Pd/CNTs catalysts, the PdBO_x@Pd/CNTs catalyst displays more negative onset potential and larger peak area for CO oxidation, suggesting that the existence of strong electronic interaction between PdBO_x and Pd promotes the production of OH* by splitting H₂O, which further reacts with the adsorbed CO on the surface of Pd (35). The above results demonstrated that the strong electronic interaction between PdBO_x and Pd in the PdBO_x@Pd/CNTs catalyst promotes the adsorption of reactants (glycerol, water, and O₂) and intermediates (OH* and GLYHD*) on the catalyst surface during the GOR while easily removing the GLYA product, thus improving the catalytic performance of the GOR (Fig. 3C).

To further understand the effect of strong electronic interaction between Pd nanoparticles and PdBO_x clusters in the PdBO_x@Pd/CNTs catalyst on the GOR activity and selectivity, the apparent activation energy of PdBO_x@Pd/CNTs, PdBO_x-Pd/CNTs, and Pd/CNTs catalysts was investigated. As shown in Fig. 3D, the activation energy toward the GOR for the PdBO_x@Pd/CNTs catalyst (20.1 ± 1.8 kJ/mol) is lower than those of Pd/CNTs (42.2 ± 2.9 kJ/mol) and PdBO_x-Pd/CNTs (29.7 ± 1.2 kJ/mol). This result indicated that PdBO_x@Pd/CNTs with strong electronic interaction decreased the apparent activation energy, suggesting that the GOR could be more easily carried out (11).

To further explore the underlying catalytic mechanism, we investigate the dynamic adsorption behavior of intermediates by conducting in situ attenuated total reflection Fourier transform infrared (ATR-FTIR) spectroscopy on Pd/CNTs, PdBO_x-Pd/CNTs, and PdBO_x@Pd/CNTs catalysts during the GOR. First, we used 1-propanol (PrOH) to simulate the activation of glycerol during the catalytic process of the GOR because PrOH only has the primary O─H bonds. Figure 4 (A to C) shows the in situ ATR-FTIR spectra with temperature-resolved intermediate variation on the Pd/CNTs, PdBO_x-Pd/CNTs, and PdBO_x@Pd/CNTs catalysts. Two characteristic adsorption bands are observed around 1066 and 1056 cm⁻¹, which belong to the absorbed υ(C─OH) and υ(C─O) bonds, respectively (36). With the in situ ATR-FTIR measures on Pd/CNTs and PdBO_x-Pd/CNTs catalysts, the intensity ratio of υ(C─OH) and υ(C─O) is unchanged as the temperature increased (Fig. 4, A and B, and fig. S36). Compared with Pd/CNTs and PdBO_x-Pd/CNTs catalysts, the intensity ratio of υ(C─O) and υ(C─OH) for the PdBO_x@Pd/CNTs catalyst increases with increased temperature (Fig. 4C and fig. S36). This result suggests that the dehydrogenation of C─OH in PrOH to form C─O can be promoted on the PdBO_x@Pd/CNTs catalyst. This result also indicated that the PdBO_x@Pd/CNTs catalyst had an excellent capacity for activating the O─H bond in PrOH than Pd/CNTs and PdBO_x-Pd/CNTs catalysts, which was ascribed to the strong electronic interaction between PdBO_x clusters and Pd nanoparticles.

Fig. 4. In situ dynamic information during the GOR.

(A to C) In situ ATR-FTIR spectra on PdBO_x@Pd/CNTs, PdBO_x-Pd/CNTs, and Pd/CNTs during PrOH oxidation at different temperatures of 20°, 40°, and 60°C. (D to F) In situ ATR-FTIR spectra on PdBO_x@Pd/CNTs, PdBO_x-Pd/CNTs, and Pd/CNTs during the GOR at different temperatures of 20°, 40°, and 80°C.

We further investigate the activation of the glycerol reactant and intermediates during the GOR using in situ ATR-FTIR. The Pd/CNTs, PdBO_x-Pd/CNTs, and PdBO_x@Pd/CNTs catalysts were dispersed in the mixed solution of glycerol and KOH, which then underwent in situ ATR-FTIR research at different reaction temperatures. Two distinctive adsorption peaks are observed at 1116 and 1050 cm⁻¹ for Pd/CNTs, PdBO_x@Pd/CNTs, and PdBO_x@Pd/CNTs catalysts, which are assigned to the secondary υ(C─OH) and primary υ(C─OH), respectively (Fig. 4, D to F, and fig. S37) (36). Compared with Pd/CNTs and PdBO_x-Pd/CNTs catalysts (Fig. 4, D and E), in addition to the above two peaks, an additional peak around 1037 cm⁻¹ is observed in the magnified image of in situ ATR-FTIR for the PdBO_x@Pd/CNTs catalyst during the GOR (fig. S37), which is attributed to the absorption band of primary υ(C─O). When the temperature increases, the intensity of primary υ(C─OH) becomes weak and the intensity of υ(C─O) increases for the PdBO_x@Pd/CNTs catalyst, suggesting the dehydrogenation of primary υ(C─OH) and generation of C─O during the GOR (Fig. 4F and fig. S37) (37–39). Different from PdBO_x@Pd/CNTs, the adsorption band of υ(C─O) is not observed for Pd/CNTs and PdBO_x-Pd/CNTs catalysts with the increase in temperature. This hints that the PdBO_x@Pd/CNTs catalyst exhibits a stronger activation capacity for the primary C─OH bond of glycerol than Pd/CNTs and PdBO_x-Pd/CNTs toward the GOR.

The H₂ temperature-programmed reduction result shows that the H₂ reduction peak for the PdBO_x@Pd/CNTs catalyst shifts to a lower reduction temperature compared with those for Pd/CNTs and PdBO_x-Pd/CNTs catalysts (fig. S38). This finding proves that the strong electronic interaction between PdBO_x clusters and Pd nanoparticles promotes the activation of H adsorbed on the PdBO_x@Pd/CNTs catalyst, namely, the active hydrogen first forms on Pd nanoparticles and then spills over to PdBO_x, which may also be an important reason for the improvement of GOR performance (40).

Understanding the catalytic mechanism on PdBO_x@Pd toward the GOR by density functional theory calculation

To demonstrate the effect of PdBO_x on the electronic structure and the mechanism for the enhanced catalytic performance of PdBO_x@Pd/CNTs toward the GOR, we performed first principle calculations based on density functional theory (DFT). We constructed a Pd@BO₃ model to simulate the PdBO_x@Pd catalyst (fig. S39). The BO₃ cluster, placed on the surface of the metallic Pd, can bond with the surface Pd atoms and then form PdBO_x in the local region for the Pd@BO₃ model. For comparison, we also constructed the Pd@BO₂ and Pd models. It is worth noting that Pd@BO₂ is transformed into Pd@BO₃ when O₂ is adsorbed on the surface of the Pd@BO₂ model (fig. S40). Thus, the Pd@BO₃ structure could be used as an effective model to simulate the PdBO_x@Pd/CNTs catalyst. In addition, the calculated charge density difference of Pd@BO₃ indicates that 1.05 electrons are transferred from Pd to the BO₃ cluster (fig. S41). Such a large electron transfer number indicated the strong electronic interaction between Pd and the BO₃ cluster in Pd@BO₃.

Next, the adsorption energies of glycerol, H₂O, O₂ reactants, OH intermediate, and GLYA product on Pd and Pd@BO₃ were investigated (fig. S42). Calculation results reveal that the adsorption of glycerol*, H₂O*, O₂*, and OH* on Pd@BO₃ is stronger than that on Pd, while the adsorption of GLYA* on Pd@BO₃ is weaker than that on Pd, which is consistent with the experimental results (Fig. 5A and table S8). The high adsorption strength of glycerol*, H₂O*, O₂*, and OH* and the weak adsorption energy of GLYA* lead to the enhanced catalytic activity of Pd@BO₃ for the GOR.

Fig. 5. DFT calculations of the GOR.

(A) Adsorption energies of glycerol*, O₂*, H₂O*, OH* and GLYA* on Pd and Pd@BO₃ models. (B) Schematic explanation for the effect of the d-band center on the adsorption of glycerol*, O₂*, H₂O*, and OH* adsorbates. HOMO, highest occupied molecular orbital. (C and D) Integrated COHP of Pd─B and Pd─O bonds for the Pd@BO₃ model. (E) Energy profiles of Pd and Pd@BO₃ toward the GOR.

Density of state analysis was performed to investigate the origin of the increased adsorption energy for multiple reactants on Pd@BO₃ (fig. S43). Compared with the d-band center of Pd in a pure metallic catalyst (−1.74 eV), the presence of the BO₃ cluster in Pd@BO₃ can greatly induce the upward shift of the d-band center to −1.55 eV, resulting in more antibonding state filling and approaching the LUMO (lowest unoccupied molecular orbital) of the reactants (table S9). Furthermore, the upward shift of the Pd d-band center enhances the adsorption of glycerol*, H₂O*, O₂* reactants, and OH* intermediates on the surface of Pd@BO₃, which explains the high adsorption energy of reactants and the excellent catalytic activity of the GOR (Fig. 5B) (41–43).

Furthermore, we calculated the crystal orbital Hamilton population (COHP) of Pd─B and Pd─O bonds in Pd@BO₃ to further investigate the effect of interatomic orbital hybridization on the d-band properties of Pd@BO₃ (44–46). The Pd 4d orbital has strong bond states with O 2s-2p and B 2s-2p orbitals at the Fermi level (Fig. 5, C and D). This demonstrates that there is strong orbital hybridization between Pd(4d), O(2s, 2p), and B(2s, 2p) atoms, which induces the upward shift of the d-band center, thus promoting the adsorption of substrates and the enhancement of the catalytic activity of the GOR.

To gain further insight into the catalytic mechanism of the GOR on Pd@BO₃, we calculated the energy profiles of each catalytic elementary step (Fig. 3A), including the deprotonation of glycerol (step 2), the dehydrogenation of CH₂OHCHOHCH₂O* (step 3), hydroxylation of GLYHD* (step 4), and the production of the GLYA (step 5). Compared with Pd, the result identifies that the above four steps are promoted with the assistance of the BO₃ cluster in the Pd@BO₃ catalyst (Fig. 5E). It is worth noting that the dehydrogenation of CH₂OHCHOHCH₂O* is the RDS for glycerol oxidation to GLYA. The energy barrier of the RDS on Pd@BO₃ (1.07 eV) is lower than that of Pd (2.37 eV), which could explain the reason for the increased catalytic activity and enhanced selectivity of GLYA on Pd@BO₃. The calculated result also reveals that OH* can directly bond with the H of the primary C─H bond of the CH₂OHCHOHCH₂O* intermediate in the RDS using the Pd catalyst. In contrast, the BO₃ cluster in Pd@BO₃ first bonds the H of the primary C─H bond in the CH₂OHCHOHCH₂O* intermediate to form BO₃-H, followed by further transfer of the H in resultant BO₃-H to the adsorbed OH* on the surface of Pd@BO₃ (fig. S44). Therefore, Pd@BO₃ as a catalyst can reduce the energy barrier of the RDS toward the GOR because it can promote the site-selective activation of the primary C─H bond for the key CH₂OHCHOHCH₂O* intermediate in the RDS.

In addition, we compare the charge density difference of the adsorbed GLYHD* intermediate for Pd and Pd@BO₃ surfaces (fig. S45). The surface of Pd@BO₃ with GLYHD* gains fewer electrons than Pd. This result proves that Pd@BO₃ as a catalyst is conducive to the occurrence of the subsequent catalytic pathway, thus further improving the catalytic performance of the GOR (47, 48).

DISCUSSION

In summary, we have synthesized a PdBO_x@Pd/CNTs catalyst with strong electronic interaction in which Pd nanoparticles were covered by in situ–exsoluted PdBO_x clusters. Our studies have provided insight into the enhanced activity of the GOR over the PdBO_x@Pd/CNTs catalyst. The strong orbital hybridization between Pd(d), O(s, 2p), and B(s, p) atoms was induced by the strong electronic interaction between PdBO_x and Pd, which facilitates the strong adsorption of multiple reactants and the crucial GLYHD intermediate. PdBO_x in the PdBO_x@Pd/CNTs catalyst promotes the site-selective activation of the C─H bond for CH₂OHCHOHCH₂O* intermediates, thus reducing the energy barrier of RDS and further improving GLYA selectivity toward the GOR. These findings emphasized the importance of adjusting the electronic structure for the performance optimization of supported catalysts toward the GOR and established a general design rule for producing advanced catalysts.

MATERIALS AND METHODS

Materials

Na₂PdCl₄, H₃BO₃, and BH₃-THF (1 M solution in THF) were purchased from Aladdin Chemistry Co., Ltd. The carbon nanotube with a diameter of 60 to 80 nm was obtained from Shenzhen Nanotech Port Co., Ltd. (China). Glycerol (99.9%), ethanol, ethylene glycol (EG), and KOH were purchased from Sinopharm Chemical Reagent Co. GLYA (20% in water), EG, PrOH, D₂O, and GLYHD were purchased from TCI. LA (98%), tartronic acid (98%, Sigma-Aldrich), OA (99.99%), and FA (88%) were purchased from Aladdin.

Preparation of the catalysts

Preparation of Pd/CNTs

Na₂PdCl₄ (13.8 mg) and CNTs (95.0 mg) were added into the mixed solution of 80 ml of EG and 40 ml of deionized H₂O. The dispersion was heated and stirred in an oil bath at 140°C for 4 hours. The obtained black powder was washed several times with deionized water and ethanol. The black sample was dried at 60°C to obtain Pd/CNTs.

Preparation of PdBO _x @Pd/CNTs

The Pd/CNTs sample (50 mg) was added into 13.0 ml of Ar-saturated BH₃-THF solution. The homogeneous solution was transferred into a 15-ml Teflon-lined autoclave, which was then heated at 150°C for 72 hours. The obtained black powder was washed several times with deionized water and ethanol. The black sample was dried at 60°C to obtain PdB/CNTs, where the molar ratio of Pd to B is ~1/1. When the reaction temperature (40°C) and time (1 hour) were set, the resulting catalyst was denoted as PdB/CNTs-1 (1/0.75). When the reaction temperature (80°C) and time (24 hours) were set, the resulting catalyst was denoted as PdB/CNTs-2 (1/1.25). The above as-prepared PdB/CNTs, PdB/CNTs-1, and PdB/CNTs-2 catalysts with different Pd/B ratios were calcined at 150°C for 2 hours under the air atmosphere to obtain the PdBO_x@Pd/CNTs, PdBO_x@Pd/CNTs-1, and PdBO_x@Pd/CNTs-2 catalysts with different PdBO_x contents, respectively.

Preparation of PdBO _x -Pd/CNTs

Na₂PdCl₄ (13.8 mg), H₃BO₃ (13.8 mg), and CNTs (95.0 mg) were added into the mixed solution of 80 ml of EG and 40 ml of deionized H₂O. The dispersion was heated and stirred in an oil bath at 140°C for 4 hours. The obtained black powder was washed several times with deionized water and ethanol and then dried at 60°C. The obtained black powder was heated from room temperature to 300°C for 4 hours with a rate of 3°C min⁻¹ in a muffle furnace to obtain PdBO_x-Pd/CNTs.

Catalyst characterizations

The powder XRD patterns of the samples were recorded with an x-ray diffractometer (Rigaku D/Max 2550) using Cu Kα radiation (λ = 1.5418 Å). The transmission electron microscopy images were obtained by a Tecnai F20 equipped with a field emission gun operating at 200 kV. The XPS spectrum of samples was measured on a Thermo Fisher Scientific ESCALAB 250Xi with a photoelectron spectroscopy system using a monochromatic Al Kα (1486.6 eV) x-ray source. H₂ temperature-programmed reduction was performed using a Micromeritics AutoChem 2920 II system. O₂ temperature-programmed desorption data were recorded on a Micromeritics AutoChem 2920 II system by ramping the catalyst temperature from 30° to 750°C at a rate of 10°C/min in an Ar atmosphere. The solid-state nuclear magnetic resonance spectra were recorded on an AVANCE NEO 600 MHz with 2 s of D1, Ns of 100, single impulse sequence, and a frequency of 10,000. The determination of Pd d-band center position was performed by the high-resolution ultraviolet photoelectron spectroscopy measurements by using He II (hυ = 21.22 eV) ultraviolet excitation lines excited with a Vacuum Generators He discharge source, and a negative bias voltage of 15 V was applied to all samples to accelerate electrons of low kinetic energy and therefore to allow for an accurate determination. The electron paramagnetic resonance spectra of the OH-DMPO radical were recorded with a magnetic field of 3500.00 G and an amplitude of 1.000 G. The CO stripping curves were tested in a 1.0 M KOH solution at the potential range from −0.6 to 0.6 V (V versus reversible hydrogen electrode). The contact angles of glycerol, water, GLYHD, and GLYA on Pd/CNTs and PdBO_x@Pd/CNTs catalysts were measured by using a drop shape analyzer (DSA100 system, Krüss GmbH). X-ray PDF tests were conducted on the Brockhouse High Energy Wiggler Beamline of the Canadian Light Source (CLS), a 2.9-GeV synchrotron, using high-energy x-rays of 65.75 keV. These samples were loaded into 0.9-mm Kapton capillaries. The Q_max value is 22 Å⁻¹. The x-ray absorption spectroscopy (XAFS) study was performed at the BL14B2 of SPring-8 (8 GeV, 100 mA), Japan, in which the x-ray beam was monochromatized with a water-cooled Si (111) double-crystal monochromator and focused with two Rh-coated focusing mirrors with beam sizes of 2.0 mm in the horizontal direction and 0.5 mm in the vertical direction around the sample position to obtain XAFS spectra in both near and extended edges. Pd foil and PdO samples were used as references. The CO pulse adsorption measurements of the catalyst were performed on an AutoChem II instrument (Micromeritics, US). Before CO pulse adsorption measurements, 50 mg of the catalyst was reduced in flowing hydrogen (30 ml/min) at 573 K for 2 hours and flushed with flowing pure He (30 ml/min) for 2 hours at 573 K. When the temperature was then reduced to room temperature, CO pulses over the reduced catalyst were measured in the CO and He mixture containing 10% CO. The stoichiometry of CO/Pd = 1 was taken to determine Pd dispersion.

Catalytic performance measurement of the GOR

The reaction was carried out in a 100-ml glass reaction flask equipped with a magnetic stirrer. In a typical reaction, 200 mM glycerol, 60 mg of PdBO_x@Pd/CNTs catalyst, and 240 mg of NaOH were introduced to 50 ml of deionized water in a reaction flask with a magnetic stirrer. When the temperature of 60°C was stable, the reaction flask was purged by O₂ at a flow rate of 150 ml/min⁻¹. The selectivity of all products was investigated by using an Agilent 1260 Infinity HPLC instrument. The HPLC instrument was equipped with an ultraviolet detector (210 nm) and refractive index detector using an Alltech OA-1000 organic acid column (4.6 mm by 250 mm, Agilent) operating at 353 K. In a typical experiment procedure, the calibration curves and the retention time of all products can be acquired by adding the solution for the standard sample. The stability of samples was measured by using initial catalysts with five repeated experiments.

The TOF values were obtained quantitatively using Eq. 1

$$\text{TOF}=n/m/\text{reaction time}$$ (1)

where n represents the conversion moles of glycerol, and m represents the total moles of Pd on the catalysts.

The yield was obtained quantitatively using Eq. 2

$$\begin{array}{c}\begin{array}{c}\text{Yield}=\text{Conversion of glycerol}\times \text{Selectivity of GLYA}\times \text{Carbon balance}\hfill \end{array}\hfill \end{array}$$ (2)

Twenty milligrams of catalyst for PdBO_x@Pd/CNTs or Pd/CNTs was placed into 20 ml of 400 mM glycerol solution to conduct the kinetic experiments operating at different reaction temperatures (40°, 50°, 60°, and 70°C). Twenty milligrams of catalyst for PdBO_x@Pd/CNTs or Pd/CNTs was placed into 20 ml of 400 mM glycerol solution with the solvent of H₂O or D₂O to conduct the kinetic isotope effect experiments operating at 60°C.

In situ FTIR measurements

Glycerol (20 wt %) or PrOH aqueous solution (2 ml), NaOH (0.024 g), and catalyst (8 mg) were fully blended and loaded into an in situ autoclave that allows infrared light to pass through. The autoclave was filled with O₂ (1 MPa), and then the mixture was stirred at a given temperature (20°, 40°, 60°, and 80°C) for 2 hours. FTIR spectra were collected in a transmission mode on a Bruker Vertex 70 infrared spectrometer at a resolution of 4 cm⁻¹.

Theoretical section

The calculations were performed using the periodic DFT method implemented in the Vienna Ab initio Simulation Package (49). The generalized gradient approximation–based Perdew-Burke-Ernzerhof exchange-correlation functional was used to describe the exchange-correlation interaction of valence electrons (50). In this work, we ignored the van der Waals interaction, such as the DFT-D3 method, because the commonly used approach for van der Waals interaction may lead to the overestimation of chemisorption energies (51). The cutoff energy for the plane wave basis set was taken as 450 eV. A Monkhorst-Pack grid with a size of 2 by 2 by 1 was used to sample the Brillouin zone for structural optimization, and a 3 by 3 by 1 k-point grid was adopted to calculate the density of states (52). The ground-state wave functions and atomic geometries were converged within 10⁻⁵ eV for energy and 0.02 eV/Å for the maximal force, respectively. The transition states were obtained by relaxing the force <0.05 eV Å⁻¹ by using the dimer method (53, 54). The COHP was calculated by LOBSTER (55). Atomic charges were calculated using the atom-in-molecule theory proposed by Bader (56). The optimized structures were drawn to permit visualization for electronic and structural analysis using the VESTA software package (57). The data calculated by the Vienna Ab initio Simulation Package were processed by the vaspkit1.3.3 package (58).

The р(4 × 4) Pd (111) surface was represented by a periodic slab model using the cubic Pd with cell parameters of a = b = c = 3.89070 Å, which contains four layers of 64 atoms. All of the slab models were built with vacuum layers of 15 Å, and the first layer slabs, as well as the BO_x clusters (BO₂ and BO₃) and glycerol molecules of the surface, were allowed to relax, while the other layers were frozen during the geometry optimization.

The adsorption energy (E_ads) for glycerol, other reactants (O₂*, H₂O*, O*, and OH*), and products (GLYA) was calculated using Eq. 3 (59)

$$E_{\text{ads}}=E_{\text{system}}-E_{\text{model}}-E_{\text{substrate}}$$ (3)

where E_system, E_model, and E_substrate are the total energy of the system with the absorbed substance, the isolated energies of models (glycerol*, O₂*, H₂O*, O*, OH*, and GLYA) and the substrate (Pd, PdBO₂, PdB₂O₄, and PdBO₃), respectively.

Supplementary Materials

This PDF file includes:

Figs. S1 to S45

Tables S1 to S9

References