Alkene streams from industrial naphtha cracking often contain about 1% diene or alkyne side products. The concentrations of these side products must be reduced to less than 10 ppm because they can quickly poison the downstream catalysts used in the polymerization process [1-6]. The selective hydrogenation of 1, 3-butadiene to butenes is the most promising way to solve this issue [5], and Pd-based catalysts have been widely used for this purpose owing to their high hydrogenation activity [7-13]. However, Pd monometallic catalysts generally exhibit low alkene selectivity at high conversions, and are quickly deactivated by heavy coking during reactions.

Extensive efforts have been devoted to the modification of Pd catalysts to improve the butene selectivity and catalyst stability [1, 3, 7, 13-17]. The decoration of Pd with a secondary metal to form bimetallic nanoparticles is a common method, and allows the Pd electronic properties and surface ensembles to be tailored to enhance the selectivity of the catalyst [1, 11, 13, 17-19]. However, improvements in the selectivity have generally been achieved at the expense of catalytic activity. For example, Kolli et al. [10] synthesized a series of AuPd/δ-Al₂O₃ catalysts with different atomic Au/Pd ratios, and found that although increasing the Au/Pd ratio above 15 markedly increased the selectivity, the catalytic activity was considerably decreased. Similarly, Zhang [5] reported that Ag-alloyed Pd single-atom catalysts, which are processed with only ppm levels of Pd, showed excellent ethylene selectivity at high acetylene conversions over a wide temperature range. However, the activity was considerably decreased and 100% conversion was achieved only above 160 ℃ owing to the extremely low Pd concentration. Recently, Chen and co-workers [19] showed that NiPd/γ-Al₂O₃ catalysts exhibited both higher 1-butene selectivity and higher hydrogenation activity than the monometallic catalysts. Unfortunately, the improved butene selectivity was only achieved at 1, 3-butadiene conversions below 60%.

Applying oxide coatings to Pd catalysts has proved to be another effective way of dividing Pd surface ensembles for selectivity improvement [7, 15, 17, 20, 21]. For example, Kang et al. [21] reported that Pd/SiO₂ catalysts coated with TiO_x, NbO_x, or CeO_x exhibited considerably higher ethylene selectivity than uncoated Pd/SiO₂ in acetylene hydrogenation after reduction at 500 ℃. Later, they also found that TiO₂-modified Pd/SiO₂ catalysts showed improved selectivity for conversion of 1, 3-butadiene to 1-butene, essentially without changing the activity [15]. Furthermore, Crabb et al. [17] reported that addition of FeO_x onto Pd or Pt catalysts could suppress the total hydrogenation to butane to some extent. Nevertheless, the reduction in hydrogenation activity caused by carbon deposition or blocking of the active sites by FeO_xwas still severe. In our previous studies, we have shown that deposition of microporous alumina layers with a thickness of approximately 3.8 nm onto Pd/Al₂O₃ catalysts by atomic layer deposition (ALD) not only considerably improved the butene selectivity to close to 100% at 95% 1, 3-butadiene conversion, but also exhibited very high stability against coking [7]. The enhanced catalytic performance was attributed to the confinement effect induced by micropores within the ALD alumina layer. We also found that the alumina coating largely blocked the surface Pd active sites, thereby considerably reducing the hydrogenation activity. Recently, we further showed that deposition of Ga₂O₃ on Pd/Al₂O₃ catalysts through ALD could generally enhance ethylene selectivity; however, the activity was very sensitive to the coverage of Ga₂O₃ [20]. Taken together, most studies reported in the literature show that improvement in the alkene selectivity is achieved at the expense of catalytic activity, and there has been very limited success in improving both the activity and selectivity in a Pd-based catalyst system.

ALD is an effective method for catalyst synthesis and post-modification, and has near atomic precision [7, 8, 22-26]. Herein, we precisely deposited FeO_x onto a Pd/Al₂O₃ catalyst using several cycles of ALD to tune the coverage of FeO_x. We show that decoration of Pd nanoparticles (NPs) with FeO_x remarkably enhanced both hydrogenation activity and butene selectivity in the selective hydrogenation of 1, 3-butadiene. The enhanced catalytic performance was attributed to electronic modulation, formation of a Pd-FeO_x interface, and geometric effects.

The Pd/Al₂O₃ catalyst was synthesized using a wet-impregnation method [22]. Therein, 0.1 g Pd(NO₃)₂·2H₂O (Aladdin, ≥97.7%) and 0.45 g citric acid (Sinopharm Chemical Reagent Co., Ltd., ≥99.5%) were dissolved in 2.6 mL water to form a Pd-citric acid solution. Then, 4.4 g spherical Al₂O₃ powder (Nanodur, Alfa Aesar, 99.5%) was added and the solution was mixed uniformly. The resulting mixture was dried in an oven at 70 ℃ for 14 h. Finally, the dried material was calcined in 10% O₂ in He at 400 ℃ for 2 h, then reduced at 250 ℃ for 30 min in 10% H₂ in He to obtain the Pd/Al₂O₃ catalyst.

FeO_x ALD was carried out using a viscous flow reactor (GEMSTAR-6^TM Benchtop ALD, Arradiance) [7]. Ultrahigh-purity N₂ (99.999%) was used as the carrier gas at a flow rate of 200 mL/min. The Pd/Al₂O₃ catalyst was loaded into the ALD reactor, and FeO_x ALD was performed at 150 ℃ by exposing the sample to alternating cycles of ferrocene and oxygen. The ferrocene precursor was contained in a stainless steel reservoir and heated to 90℃ to achieve a sufficient vapor pressure. The inlet lines were heated to 110℃ to avoid condensation. The time sequence for FeO_x ALD was 300, 200, 500, and 200 s for ferrocene exposure, N₂ purge, O₂exposure, and N₂ purge, respectively. The resulting Pd/Al₂O₃ samples with different numbers of FeO_x ALD cycles were denoted xFe/Pd/Al₂O₃, where x denotes the number of FeO_x ALD cycles.

The contents of Pd and Fe were determined by inductively coupled plasma atomic emission spectroscopy (ICP-AES) (University of Science and Technology of China). The morphology of the catalysts was characterized on an aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) instrument at 200 kV (JEOL-2010F, University of Science and Technology of China).

Diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) CO chemisorption measurements were performed on a Nicolet iS10 spectrometer equipped with an MCT detector and a low-temperature reaction cell (Praying Mantis Harrick). After the sample was loaded into the cell, it was first calcined in 10% O₂ in He at 200 ℃ for 1 h, and this was followed by reduction in 10% H₂ in He at 200 ℃ for 1 h. The sample was then cooled to room temperature under He and a background spectrum was collected. Subsequently, the sample was exposed to 10% CO in He at a flow rate of 20 mL/min for about 30 min until saturation. Next, the sample was purged with He at a flow rate of 20 mL/min for another 30 min to remove the gas phase CO, and then the DRIFT spectrum was collected with 128 scans at a resolution of 4 cm^-1.

X-ray photoelectron spectroscopy (XPS) measurements were obtained on a Thermo-VG scientific Escalab 250 spectrometer equipped with an Al anode (Al K_α = 1486.6 eV). Before the XPS experiments, all samples were pretreated ex situ at 200 ℃ in 10% O₂ in Ar for 1 h, then reduced in 10% H₂ in Ar for another 1 h, and finally cooled to room temperature in 10% H₂ in Ar. The samples were then carefully sealed and quickly transferred for the XPS measurements.

Selective hydrogenation of 1, 3-butadiene was conducted in a fixed-bed flow reactor. The feed gas consisted of 1.9% 1, 3-butadiene, 4.7% H₂, and Ar as the balance gas. The total gas flow rate was maintained at 25 mL/min and 50 mg of each catalyst was used. The catalysts were diluted with 1 g quartz chips (60-80 mesh). Prior to each test, the catalyst was pretreated at 200 ℃ in 10% O₂ in Ar for 1 h, and then reduced in 10% H₂ in Ar at the same temperature for another 1 h. After the reaction stream was introduced, the catalyst was first stabilized at room temperature for several hours, and then the reaction products were collected and analyzed using an online gas chromatograph equipped with an FID detector and a capillary column (ValcoPLOT VP-Alumina-KCl, 50 m × 0.53 mm).

The Pd and Fe loadings in the xFe/Pd/Al₂O₃ catalysts were determined by ICP-AES. As shown in Table 1, the Pd loading was 1.0 wt% in all samples, and the Fe loadings were 0.26 wt%, 0.41 wt%, 0.67 wt%, and 0.73 wt% in 6Fe/Pd/Al₂O₃, 10Fe/Pd/Al₂O₃, 20Fe/Pd/Al₂O₃, and 30Fe/Pd/Al₂O₃, respectively. The gradual increase in the Fe loading as a function of the number of FeO_x ALD cycles clearly confirms the successful deposition of FeO_x on the Pd/Al₂O₃ catalyst. Aberration-corrected HAADF-STEM measurements were employed to investigate the morphology of the 30Fe/Pd/Al₂O₃ catalyst. The size of the Pd NPs was on average about 7 nm in this sample, as shown in Fig. 1(a)-(c). Unfortunately, we could not observe the FeO_x layer on the Pd NPs with high resolution STEM (Fig. 1(b)), probably because the FeO_x layer was too thin and therefore showed negligible contrast with the Pd NPs. Nonetheless, elemental mapping using energy-dispersive X-ray spectroscopy (EDS) demonstrated overlapping Pd L_ɑ and Fe K_ɑ signals, which suggests that the Pd NPs were decorated with FeO_x (Fig. 1(d)-(f)).

Infrared (IR) spectroscopy of CO chemisorbed on Pd NPs has been extensively studied, and the relationship between the IR features and the Pd structure is well established [10, 21, 22, 27]. Here, DRIFTS CO chemisorption measurements were carried out to evaluate the accessibility of Pd NPs after FeO_x deposition. As shown in Fig. 2, the uncoated Pd/Al₂O₃ shows three characteristic peaks at 2083, 1976, and 1942 cm^-1, which are assigned to linear CO on corner sites, bridge-bonded CO on edge and step sites, and bridge-bonded CO on (111) facets of Pd NPs, respectively [24]. The intensities of these three CO peaks decreased after application of FeO_x ALD on Pd/Al₂O₃, and, as expected, they further decreased as the number of FeO_x ALD cycles was increased. This result also suggests that the coverage of FeO_x on Pd NPs can be increased gradually with precise control, similar to the trend observed for Al₂O₃ ALD on Pd NPs [7]. In addition, the decrease in the CO peak intensity at 1942 cm^-1, assigned to bridge-bonded CO on (111) facets of Pd NPs, was considerably more dramatic than that of the other two CO peaks. This is in contrast to the observed formation of an Al₂O₃ ALD coating on Pd NPs, in which the intensity of bridge-bonded CO on low-coordinate sites decreased more quickly than that of bridge-bonded CO on (111) facet sites [24]. Our result thus implies that FeO_x preferentially decorates the Pd (111) facets and leaves the low-coordinate Pd sites accessible for catalytic function.

The influence of FeO_x on the electronic properties of Pd NPs was investigated by XPS, as shown in Fig. 3. In the Fe 2p region (Fig. 3(a)), 6Fe/Pd/Al₂O₃, 10Fe/Pd/Al₂O₃, and 20Fe/Pd/Al₂O₃ showed the same Fe 2p_3/2 binding energy of 710.8 eV, which suggests that the FeO_x species in these samples were mainly in the 3+ oxidation state [28, 29]. The increased peak intensities are in line with the increased FeO_x loadings from the increased number of FeO_x ALD cycles (Table 1). In the Pd 3d region (Fig. 3(b)), the deconvoluted Pd 3d_5/2 peak at 336.3-336.9 eV in xFe/Pd/Al₂O₃ (x = 0, 6, 10, 20) was assigned to Pd²⁺ [23, 30], which implies that the Pd NPs in these samples were partially oxidized as a result of air exposure before the XPS measurements. The deconvoluted Pd 3d_5/2 peak at 335.0 eV in the spectrum of uncoated Pd/Al₂O₃ was assigned to metallic Pd [23, 30]. After deposition of FeO_x on the Pd/Al₂O₃ catalyst, the Pd 3d binding energy of metallic Pd gradually shifted to higher energies and reached 335.4 eV in the spectra of 10Fe/Pd/Al₂O₃ and 20Fe/Pd/Al₂O₃. Clearly, the deposition of FeO_x on Pd NPs resulted in an electron-deficient Pd surface owing to the strong interaction between Pd and FeO_x [31, 32]. Interestingly, we also noted that the intensities of the Pd 3d XPS peaks in all samples were nearly identical, which implies that the FeO_x film on the Pd NPs was very thin. This is consistent with the low Fe loadings (Table 1) and the STEM observations (Fig. 1(b)).

When the xFe/Pd/Al₂O₃ catalysts were examined for selective hydrogenation of 1, 3-butadiene, all samples were first stabilized in the reaction stream at room temperature. The reaction temperature was then slowly increased to obtain different 1, 3-butadiene conversions. As shown in Fig. 4, the 1, 3-butadiene conversion was 6.7% on the uncoated Pd/Al₂O₃ catalyst at 25 ℃ and reached 99% at 43 ℃. Surprisingly, all of the Pd/Al₂O₃ samples decorated with FeO_x exhibited considerably higher 1, 3-butadiene conversions at room temperature; 30Fe/Pd/Al₂O₃ showed the highest conversion (45%) at 26 ℃. However, we did not see an obvious trend in the activity with increasing numbers of FeO_x ALD cycles. Such a remarkable increase in the hydrogenation activity of FeO_x-coated Pd/Al₂O₃ catalysts is rather unusual because the active sites of the Pd NPs were partially blocked by FeO_x, as indicated by the CO chemisorption results in Fig. 2. Indeed, in previous work, we have shown that the microporous Al₂O₃ ALD coating on Pd/Al₂O₃ largely blocks the Pd active sites, thereby considerably reducing the hydrogenation activity [7]. Kang et al. [21] also reported that Pd/SiO₂ catalysts coated with TiO_x, NbO_x, or CeO_x generally exhibited lower activities than uncoated Pd/SiO₂ after reduction at 300 ℃ in acetylene hydrogenation reactions because the Pd surface of the modified catalysts was partially covered with metal oxides. In addition, these authors observed a negative shift of approximately 0.3 eV in the Pd 3d binding energy on metal-oxide-coated Pd/SiO₂ samples. Therefore, an electron-deficient Pd surface, resulting from the strong interaction between Pd and FeO_x, could potentially increase the hydrogenation activity. In contrast, the generated Pd-FeO_x interface may also play an important role in the activity increase [33-36].

The reaction products were plotted as a function of the 1, 3-butadiene conversion, as shown in Fig. 5. With the uncoated Pd/Al₂O₃ catalyst, the total butene selectivity was 100% below 75% conversion (Fig. 5(a)). The most desired product, 1-butene, was the major product, with a selectivity of 74% (Fig. 5(b)), and the trans-and cis-2-butene selectivities were 20% and 6%, respectively (Fig. 5(c) and (d)). When the conversion was higher than 75%, the 1-butene selectivity dropped rapidly, and this was accompanied by a sharp increase in butane selectivity. At the same time, the selectivity for trans-and cis-2-butene first increased slightly at conversions between 75% and 90%, then declined at higher conversions. It is clear that on increasing the conversion of 1, 3-butadiene, secondary hydrogenation of 1-butene to butane occurred first, and secondary hydrogenation of trans-and cis-2-butene only became pronounced at conversions above 90%. At near complete conversion of 1, 3-butadiene (99%), selectivity for butenes decreased to 52%, which is consistent with the previous literature [4, 7].

Deposition of FeO_x on Pd/Al₂O₃ using ALD caused a slight decrease in the 1-butene selectivity and an increase in the trans-2-butene selectivity at low conversions, at which the total butene selectivity was preserved at 100% (Fig. 5(a)-(c)). For example, the selectivity of 30Fe/Pd/Al₂O₃ for 1-butene and trans-2-butene was about 60% and 35%, respectively, at conversions below 80%. At high 1, 3-butadiene conversions (> 75%), the total butene selectivity increased dramatically with increasing numbers of FeO_x ALD cycles. The total butene selectivity was about 95% for 30Fe/Pd/Al₂O₃ at 99% conversion, which is much higher than that observed for the uncoated Pd/Al₂O₃ catalyst (52%). The secondary hydrogenation of all butenes (1-butene, and trans-and cis-2-butene) to butane was effectively suppressed by the FeO_x coating (Fig. 5(e)). The increase in butene selectivity at high conversions can be mainly attributed to a geometric effect, in which large ensembles of Pd surface atoms were divided into smaller ones by FeO_x; this is consistent with previous studies [10, 15, 21, 37]. In addition, at conversions above 75%, formation of trans-and cis-2-butene through isomerization became the dominant secondary reaction on the FeO_x-coated Pd/Al₂O₃ catalysts, similar to results seen for the Al₂O₃ ALD-coated Pd catalyst [7]. This change in the reaction pathway, from butene hydrogenation to isomerization, is likely due to the lower concentration of dissociated hydrogen atoms on small Pd ensembles compared with that on larger Pd ensembles as a result of the enhanced steric effect. The concentration of dissociated hydrogen atoms on small Pd ensembles is likely to be too low for butene hydrogenation, but still sufficient for isomerization. This effect is similar to that of CO additives on a Pd surface [14], and the improvement in alkene selectivity is in line with results observed for single-atom catalysts with high hydrogenation selectivity [5, 8, 38, 39].

In this work, we have demonstrated that ultrathin FeO_x coatings can be precisely grown on a Pd/Al₂O₃ catalyst using ALD. DRIFT CO chemisorption suggested that FeO_x preferentially decorated the Pd (111) facets, and XPS showed that the Pd NPs became positively charged on addition of FeO_x owing to the strong Pd-FeO_x interaction. In the selective hydrogenation of 1, 3-butadiene, we showed that deposition of FeO_x onto the Pd/Al₂O₃ catalyst remarkably improved both the hydrogenation activity and butene selectivity. The increase in hydrogenation activity is attributed to electronic effects and the newly generated Pd-FeO_x interface, whereas the enhancement in butene selectivity was mainly induced by a geometric effect, in which large ensembles of Pd surface atoms were divided into smaller ones by FeO_x. This geometric and electronic modulation of Pd catalysts through transition-metal-oxide coatings may open up new opportunities for the improvement of catalysts for other hydrogenation reactions.

Notes

The authors declare no competing financial interest.