Engineering the surface structure of a catalyst to improve its selectivity to target products is an important focus in heterogeneous catalysis [1-5]. Conventionally, this engineering mainly focuses on the regulation of composition, morphology, particle size and support, which can potentially alter the steric and electronic properties of the catalyst surface [6-10]. Recently, surface ligand modification has emerged as an effective and elegant strategy to tune the catalytic selectivity of metal nanoparticles (NPs). By modifying Pd/Al₂O₃ with alkanethiol coatings, Marshall et al. [11] improved the selectivity of 1-epoxybutane from 11% to 94% at equivalent reaction conditions and 1-epoxy-3-butene conversions. On the other hand, ligand modification offers a fine control over the steric or electronic structure of metal nanoparticles without introducing external active sites, which provides ideal models to investigate the reaction mechanism. For example, by utilizing ethylenediamine as an efficient ligand to tune the surface electronic structure of Pt nanowires, Chen et al. [3] revealed that the interfacial electronic effect played a key role in the selective catalysis. The electron-rich Pt surface favored the adsorption of electron-deficient reactants over electron-rich substrates, which consequently enhanced the selectivity of phenylhydroxylamine in nitrobenzene reduction. Nevertheless, to the best of our knowledge, there is no report about utilizing ligand modification to regulate the catalytic selectivity of high-temperature (> 300 ℃) gas-phase reactions due to the volatility and limited thermal stability of traditional organic ligands [2]. The successful discovery of thermal-stable ligands will eventually overcome this obstacle and extend this effective modification strategy to some important high-temperature industrial reaction systems.

Here, propane oxidative dehydrogenation (PODH) is chosen as a model reaction as it is a promising route for the industrial production of propene [12, 13]. Conventionally, the sequential oxidation of propene to CO₂ is thermodynamic favorable, which limits the catalytic selectivity towards propene [14, 15]. In this work, we demonstrate that the modification of polymeric phosphate on NiO nanoparticles (NPs) improves its selectivity to propene by 2–3 times at equivalent conversion of propane. Kinetic studies and DFT calculations reveal that the presence of polymeric phosphate can weaken the affinity of propene, which consequently reduce the possibility of over-oxidation and improve the selectivity towards propene.

Polymeric phosphate modified NiO was prepared by a post-treatment of NiO NPs in phosphate solution, followed by a high-temperature calcination to in situ generate polymeric phosphate (Scheme S1, Supporting Information). The content of P on NiO surface could be regulated by the concentration of phosphate solution. The corresponding catalysts were denoted as NiO-xSPP, where x is the molar ratio of P to Ni in the treatment solution.

Transmission electron microscopy images (Fig. S1) reveal that the modification of polymeric phosphate does not significantly change the morphology of NiO NPs. The size of NiO-1.8SPP (~11.4 nm) is slightly smaller than fresh NiO NPs (~14.0 nm), likely due to etching of NiO during the post-treatment. High-resolution transmission electron microscopy image (Fig. 1(a)) of NiO-1.8SPP shows lattice fringes with spacing of 0.21 nm, which could be readily attributed to the NiO(200) plane [16]. Besides, all NiO-SPP samples display similar X-ray diffraction patterns with NiO (Fig. 1(b)). No additional peaks belonging to other species are observed, excluding the structure change of NiO.

The structure of phosphate on NiO-SPP is determined by Fourier transform infrared spectra (FT-IR). Comparing to NiO NPs, all NiO-SPP samples exhibit obvious phosphorous peaks (Fig. 1(c)), indicating that phosphate is successfully modified onto the NiO surface. The peaks at 1069 and 1174 cm^–1 can be assigned to phosphate (PO₄^3–) while two other peaks at 974 and 890 cm^–1 are characteristic peaks of polymeric phosphate (P–O–P) [17, 18]. The latter is detected only in the samples after calcination (Fig. S2), suggesting that polymeric phosphate is in situ generated by the high-temperature calcination. This result is further confirmed by the obvious upshift of P 2p X-ray photoelectron spectroscopy (XPS) peak after calcination (Fig. 1(f) and Fig. S3). Taking NiO-1.8SPP as an example, the P 2p XPS peak before calcination locates at 132.9 eV, corresponding to PO₄^3– species. After calcination, the peak upshifts to 133.5 eV, corresponding to P₂O₇^4–. The 0.6 eV upshift clearly indicates the transformation of phosphate to polymeric phosphate. Besides, the presence of P ligand after a high-temperature (500 ℃) calcination also implies a good thermal stability of polymeric phosphate, consistent with the themogravimetric analysis (Fig. S4).

The presence of polymeric phosphate changes the surface electronic structure of NiO NPs. As for NiO NPs, the Ni 2p peak at 855.8 eV is attributed to Ni³⁺ species, and the peak at 854.0 eV is assigned to Ni²⁺ species along with its satellite peak at 861.3 eV (Fig. 1(d)) [19]. However, after modified by polymeric phosphate, all these three peaks upshift. Similar upshift is observed for O 1s as well, indicating that polymeric phosphate can withdraw electrons from NiO NPs. Specifically, NiO-1.8SPP shows the highest upshift. A 0.4 eV upshift in Ni 2p and a 0.3 eV upshift in O 1s are observed, indicating a strong interaction between NiO and polymeric phosphate. Besides, the modification of polymeric phosphate also shifts the reduction temperatures of NiO and desorption temperature of oxygen to higher values (Figs. S5 and S6), suggesting that the strong interaction between NiO and polymeric phosphate can suppress the reactivity of oxygen species.

With their good thermal stability and unique surface electronic structure, NiO-SPP readily serves as great models to investigate the effect of ligand modification in high-temperature catalytic reaction (Fig. S7). Propane oxidative dehydrogenation was chosen because it is an important reaction that remains challenging to control product selectivity [15, 20-25]. Fig. 2(a) shows the propene selectivity as a function of propane conversion over NiO NPs and NiO-SPP. In general, due to the higher reactivity of propene over propane, the propene selectivity decreases along with the increase of propane conversion [14]. It is important to highlight that at equivalent conversion of propane, the selectivity of propene on NiO-SPP is generally 2–3 times higher than that on NiO (Fig. 2(a) and (b)). In addition, as the surface content of P increases, the selectivity towards propene increases accordingly (Table S1). These results clearly demonstrate that the modification of polymeric phosphate can improve the catalytic selectivity towards propene. Propane conversion and propene selectivity remain almost unchanged over 70 h on-stream for NiO-1.8SPP (Fig. 2(d)), and the surface polymeric phosphate still existed on catalyst (Fig. S8), suggesting that NiO-1.8SPP is a promising candidate for long-term operation. Additionally, we can further improve the propene selectivity of NiO-1.8SPP by increasing the gas hourly space velocity (GHSV), though the propane conversion declines simultaneously (Fig. 2(b)). This result, on the other hand, implies that the facilitated propene desorption is beneficial to the propene selectivity. Interestingly, no cracking products such as ethane and methane are detected on both NiO-1.8SPP and NiO. Control experiment on NiO modified by H₃PO₄, however, exhibits considerable cracking products and coke (Figs. S4 and S9). These results demonstrate that a small amount of Na⁺ could prevent the coke formation on the catalyst. Similar results have been previously reported in Pt catalyzed propane dehydrogenation [26, 27].

To reveal how the ligand modification enhanced the catalytic selectivity, we carried out the propene oxidation reaction on NiO and NiO-1.8SPP. In principle, a catalyst with suppressed propene oxidation benefits the catalytic selectivity towards propene. Besides, the investigation of reaction kinetics also provides valuable information for the reaction mechanism. As can be seen from Fig. 3(a), NiO-1.8SPP shows 2–3 fold lower reaction rate than NiO at full range of propene partial pressure, indicating an intrinsic lower activity of NiO-1.8SPP in propene oxidation. It is important to highlight that the apparent reaction order of propene on NiO -1.8SPP (0.49) is much larger than that on NiO NPs (0.05), suggesting a weaker adsorption of propene on NiO-1.8SPP [28]. This result is further confirmed by C₃H₆ temperature-programmed desorption (C₃H₆-TPD). The propene desorption temperature of NiO-1.8SPP is roughly 50 ℃ lower than that of NiO (Fig. 3(b)), clearly demonstrating that the modification of polymeric phosphate weakens the adsorption of propene. These results could be explained by the acid-base property of polymeric phosphate and propene. According to the classic acid base theory, both polymeric phosphate and propene are Lewis base. The surface modification of polymeric phosphate therefore decreases the acidity of NiO and consequently weakens the interaction between NiO and propene [29]. To confirm the acidity of NiO NPs and NiO-1.8SPP, we further performed NH₃ temperature-programmed desorption (NH₃-TPD). As can be seen from Fig. S10, NiO NPs exhibit a significantly higher amount of NH₃ adsorption than NiO-1.8SPP, confirming that NiO NPs have much more acid sites than NiO-1.8SPP. In addition to the weakened acidity, the introduced polymeric phosphate might also isolate the lattice oxygen species of NiO, which is also beneficial to the inhibition of propene oxidation. Meanwhile, kinetic studies on propane oxidative dehydrogenation exhibit similar reaction rates for oxygen and different reaction rates for propane, indicating that the oxygen adsorption and activation is not significantly altered by the ligand modification (Fig. S11).

To rationalize the experimental observations, we further performed density functional calculations (DFT) to determine the structure and adsorption properties of NiO and NiO-SPP [30-33]. (NiO)₁₀(OH)₉ and (NiO)₁₀(OH)₃-(P₂O₇H₂)₃ were employed as models to interpret the local structure of NiO and NiO-SPP (Fig. S12). Hirshfeld charge analysis suggests that phosphate group withdraws electrons from NiO, consistent with the electron deficiency of NiO revealed by XPS (Fig. 1(d) and (e)). The adsorption energy of propene on NiO-1.8SPP (1.85 eV) is smaller than that on NiO (2.30 eV), indicating that the phosphate modification weakens the affinity of propene to NiO, which might therefore facilitate the desorption of C₃H₆ on NiO-SPP (Table 1). Taking all together, it is convincing that the improved propene selectivity on NiO-1.8SPP is due to the weakened adsorption and oxidation of propene on NiO-1.8SPP.

In summary, we successfully implemented the ligand modification strategy to tune the selectivity of high-temperature gas-phase reaction. NiO modified with polymeric phosphate exhibited a high selectivity towards propene, which is 2–3 times higher than NiO under the same reaction conditions. Experimental evidence and DFT calculations confirmed that surface polymeric phosphate can facilitate the desorption of propene to improve its selectivity. The success of the present study not only extends the application scope of ligand modification strategy to high-temperature reactions but also provides a useful approach to engineering the surface structure of metal oxide nanoparticles.

The authors thank Dr. Xiaodong Wen and Yunlei Chen for constructive discussions.