Report

Photochemical route for synthesizing atomically dispersed palladium catalysts

See allHide authors and affiliations

Science  13 May 2016:
Vol. 352, Issue 6287, pp. 797-800
DOI: 10.1126/science.aaf5251

Lightly dispersed palladium

Catalysts made from atomically dispersed metal atoms on oxide supports can exhibit very high per atom activity. However, the low loadings needed to prevent metal particle formation can limit overall performance. Liu et al. stably decorated titanium oxide nanosheets with relatively high loadings of single palladium atoms by reducing the ions with ultraviolet light and ethylene glycol. These catalysts cleaved H2 into atoms and were highly effective for hydrogenating alkenes and aldehydes.

Science, this issue p. 797

Abstract

Atomically dispersed noble metal catalysts often exhibit high catalytic performances, but the metal loading density must be kept low (usually below 0.5%) to avoid the formation of metal nanoparticles through sintering. We report a photochemical strategy to fabricate a stable atomically dispersed palladium–titanium oxide catalyst (Pd1/TiO2) on ethylene glycolate (EG)–stabilized ultrathin TiO2 nanosheets containing Pd up to 1.5%. The Pd1/TiO2 catalyst exhibited high catalytic activity in hydrogenation of C=C bonds, exceeding that of surface Pd atoms on commercial Pd catalysts by a factor of 9. No decay in the activity was observed for 20 cycles. More important, the Pd1/TiO2-EG system could activate H2 in a heterolytic pathway, leading to a catalytic enhancement in hydrogenation of aldehydes by a factor of more than 55.

Atomically dispersed catalysts with mononuclear metal complexes or single metal atoms anchored on supports have recently attracted increasing research attention (115). With 100% metal dispersity, atomically dispersed catalysts offer the maximum atom efficiency, providing the most ideal strategy to create cost-effective catalysts, particularly those based on Earth-scarce metals such as Pt (15), Au (58), Pd (912), and Ir (13, 14). Moreover, the uniform active sites of atomically dispersed catalysts make them a model system to understand heterogeneous catalysis at the molecular level (4, 6, 10, 1214, 1621), bridging the gap between heterogeneous and homogeneous catalysis.

During the past decade, several strategies for atomically dispersing metal sites on catalyst supports have emerged; these include lowering the loading amount of metal components (1, 810, 12, 20), enhancing the metal-support interactions (4, 6, 9, 19), and using voids in supports or vacancy defects on supports (3, 11, 14, 22). In most cases, the supports for atomically dispersed catalysts are deliberately chosen. Zeolites provide effective voids to anchor individual metal atoms therein and prevent them from sintering during catalysis (23, 24). Defects on reducible oxides (e.g., TiO2 and CeO2) (25, 26) and on graphene or C3N4 (9, 11, 22) help to stabilize atomically dispersed metal atoms on supports. Coordinatively unsaturated Al3+ ions on γ-Al2O3 act as binding centers to maintain the high dispersion of Pt atoms, but Pt rafts form as the loading amount of Pt increases (3). Currently, two major challenges remain in the field of atomically dispersed catalysts: (i) to ensure a loading content high enough for practical applications while maintaining the metal centers as individual sites under catalytic conditions (27, 28), and (ii) to address whether atomically dispersed catalysts offer distinct active sites and/or undergo catalytic pathways different from those of conventional metal catalysts (1, 46, 810, 12, 1621).

We report a room-temperature photochemical strategy to fabricate a highly stable, atomically dispersed Pd catalyst (Pd1/TiO2) on ultrathin TiO2 nanosheets with Pd loading up to 1.5%. Ultraviolet (UV) light–induced formation of ethylene glycolate (EG) radicals on TiO2 nanosheets was shown to be critical for preparing Pd1/TiO2. With abundant Pd-O interfaces, Pd1/TiO2 activates H2 in a heterolytic pathway distinct from the homolytic pathway on conventional Pd heterogeneous catalysts. The Pd1/TiO2 catalyst exhibits extremely high catalytic activities and stabilities in hydrogenation of C=C and C=O. A turnover frequency (TOF) greater than that of surface Pd atoms on commercial Pd catalysts by a factor of >55 was demonstrated on Pd1/TiO2 in the hydrogenation of aldehyde at room temperature, and no decay in the catalytic activity was observed during catalysis.

Two-atom-thick TiO2(B) nanosheets [figs. S1 to S4 (29)] were prepared by reacting TiCl4 with EG and were used as the support (30). H2PdCl4 was introduced into a water dispersion of TiO2(B) to allow the adsorption of Pd species (figs. S5 and S6). The mixture was then irradiated by low-density UV provided by a Xe lamp (fig. S7). After 10 min of irradiation, the Pd1/TiO2 catalyst was collected and washed thoroughly with water. No formation of Pd nanoparticles (NPs) was observed in transmission electron microscopy (TEM) images (Fig. 1A and fig. S8) or in the x-ray diffraction pattern (fig. S9) of the obtained Pd1/TiO2 catalyst, even with the loading content of Pd as high as 1.5 weight percent (wt %) as analyzed by inductively coupled plasma mass spectrometry (ICP-MS). Energy-dispersive x-ray spectroscopy (EDS) analysis in a scanning transmission electron microscope (STEM) revealed that atomic Pd was evenly dispersed in Pd1/TiO2 (Fig. 1B), unlike in supported Pd NPs prepared by a conventional impregnation method followed by calcination in air at 350°C (fig. S10). To verify that Pd atoms were dispersed in Pd1/TiO2, we performed x-ray absorption near-edge structure (XANES) and extended x-ray absorption fine structure (EXAFS) spectrometry (Fig. 1C and figs. S11 and S12). There was only one notable peak in the region 1 to 2 Å from the Pd-O contribution, and no peak in the region 2 to 3 Å from the Pd-Pd contribution, confirming the sole presence of dispersed Pd atoms in Pd1/TiO2 (table S1). The calcination of the as-prepared Pd1/TiO2 in air at 350°C removed the organic residues on the surface of Pd1/TiO2 and thus allowed direct observation of the atomic dispersion of Pd by aberration-corrected STEM (Fig. 1D and fig. S13).

Fig. 1 Structural characterizations of Pd1/TiO2.

(A) Representative TEM image of Pd1/TiO2. The inset is an aberration-corrected STEM image for cross sections of ultrathin TiO2(B), showing that it is composed of only two layers of Ti atoms. (B) STEM-EDS elemental mapping of a single Pd1/TiO2 nanosheet. (C) FT-EXAFS spectra of Pd1/TiO2 and bulk palladium foil at the Pd K-edge, showing the surrounding atoms adjacent to Pd atoms. (D) High-resolution high-angle annular dark-field (HAADF) STEM image of Pd1/TiO2. The sample was calcined in air at 350°C for better contrast.

After calcination, there were still a large number of dispersed Pd atoms on the TiO2 support, even with the Pd loading up to 1.5 wt%. We also investigated the CO adsorption behavior of Pd1/TiO2 (fig. S14) to confirm the atomic dispersion of Pd on the catalyst. There was only a weak band at 2100 cm−1 ascribed to CO adsorbed on Pdδ+ in a top configuration (31). No signals attributed to CO adsorbed on bridge or hollow sites were observed, quite unlike supported Pd nanoparticulate catalysts (fig. S15).

Styrene hydrogenation was chosen as a model reaction to evaluate the catalytic activity of Pd1/TiO2. The Pd1/TiO2 catalyst displayed an extremely high activity and stability (Fig. 2, A and B) relative to commercial Pd/C (fig. S16), TiO2(B)-supported Pd NPs (fig. S17), and unsupported homogeneous H2PdCl4 catalysts. We achieved 100% styrene conversion in 1 hour at a molar ratio of 1:104 (Pd:styrene). The calculated TOF value of Pd1/TiO2, 8973 hours−1, was greater than that of surface Pd atoms on the Pd/C catalyst (972 hours−1) by a factor of 9. The reaction rate of Pd1/TiO2 was maintained even after 20 cycles with the same catalyst (fig. S18), which suggests that the atomically dispersed structure of Pd1/TiO2 was robust under the catalytic conditions. There was no detectable change in the EXAFS fitting profiles after 20 catalytic cycles (fig. S19 and table S2). In contrast, an obvious decreased activity was observed on unsupported H2PdCl4 in the second cycle and even at the end of the first run (fig. S20). Because styrene hydrogenation is a zero-order reaction whose reaction rate is independent of the styrene’s concentration (32, 33), the decline in the reaction rate of H2PdCl4 (Fig. 2B) was caused by the changing status of the catalyst. After reaction, small Pd NPs were detected in the reaction mixture (figs. S21 and S22).

Fig. 2 Catalytic performances of Pd1/TiO2 and reference materials in styrene hydrogenation.

(A and B) Catalytic performances for the first run (A) and TOF (B) of several recycles of repeated reactions for Pd1/TiO2, H2PdCl4, and commercial Pd/C. The same portion of Pd1/TiO2 catalyst was recycled and used for 20 runs without loss of activity. (C) FT-EXAFS spectra at the Pd K-edge of PdCl2/TiO2 before and after catalysis reaction. (D) First- and second-run catalytic performances of PdCl2/TiO2. Reaction conditions: ethanol, 10 ml; Pd, 0.005 μmol; styrene, 50 μmol; T = 303 K; pressure = 0.1 MPa.

To better understand why the Pd1/TiO2 catalyst possessed such a high catalytic activity and stability, we prepared a catalyst (denoted PdCl2/TiO2) by the same method as for Pd1/TiO2 but without the UV treatment (fig. S23). No Pd-Pd bonds in PdCl2/TiO2 were detected by EXAFS (Fig. 2C and fig. S24), similar to Pd1/TiO2. The coordination numbers of Pd-O and Pd-Cl in the obtained PdCl2/TiO2 were 2.2 and 1.7, respectively (table S3). The Pd:Cl molar ratio of ~1:2 in the catalyst was confirmed by the elemental analysis (fig. S25). All of these data indicated that Pd atoms in PdCl2/TiO2 were in the form of individual PdCl2 species bound on TiO2(B). The presence of two Cl ligands on each Pd atom made the catalytic performance of PdCl2/TiO2 much poorer than that of Pd1/TiO2 (Fig. 2D). The reaction rate already declined during the first run and kept decreasing after every recycle, suggesting a deleterious effect of Pd-Cl bonds on the catalysis. Similar to H2PdCl4, the decreased activity of PdCl2/TiO2 was attributed to the sintering of Pd atoms into NPs during catalysis. EXAFS studies revealed that a peak in the region 2 to 3 Å from the Pd-Pd contribution emerged for the PdCl2/TiO2 catalyst after catalysis (Fig. 2C and table S4). Pd NPs were observed in TEM images for the used PdCl2/TiO2 catalyst (fig. S26), indicating that the presence of Pd-Cl bonds would destabilize atomically dispersed Pd on TiO2 and induce their sintering into NP during catalysis.

The removal of Cl ligands on Pd under mild UV conditions appears vital for preparing highly stable and active Pd catalysts. To confirm this, we thoroughly washed PdCl2/TiO2 with water until no Cl was detected in the supernatant. The water dispersion of PdCl2/TiO2 was then exposed to UV for 10 min. As expected, all Cl ligands on PdCl2/TiO2 were released into the supernatant (figs. S25 and S27). The molar ratio of the released Cl to the anchored Pd was measured to be ~2, confirming the formation of the Cl-free Pd1/TiO2 catalyst after the UV treatment.

The UV-induced elimination of Cl from PdCl2/TiO2 was attributed to the photoreactivity of TiO2(B) nanosheets. As shown in Fig. 3A, TiO2(B) nanosheets treated by UV alone (denoted as TiO2-UV) already displayed an electron spin resonance (ESR) spectrum with an intense peak corresponding to a Ti3+ species and a set of six peaks that matched perfectly with EG radicals (HOCH2•CHOH) (34). Similar signals were observed for the as-prepared Pd1/TiO2 catalyst (fig. S28). However, no obvious ESR peaks were found on the original TiO2(B) nanosheets without UV treatment (Fig. 3A).

Fig. 3 Mechanism of promotion of Cl removal by EG radicals.

(A) ESR of TiO2(B) and TiO2-UV. (B) Energies and models of intermediates and transition states in the stepwise preparation mechanism of Pd1/TiO2. (C) Molar ratios of Cl to Pd in PdCl2/TiO2, –OCH2•CHOPdCl, and Pd1/TiO2. (D) Change in pH after H2 was introduced into the water dispersion of PdCl1/TiO2.

Together with selected-area electron diffraction (SAED) (fig. S1) and the aberration-corrected STEM (fig. S2), thermogravimetric analysis (fig. S29) and infrared (IR) spectroscopy (fig. S30) suggested that two-atom-thick TiO2 nanosheets used in this work had TiO2(B)(010) as their major exposed facets, and these exposed facets were highly covered by deprotonated EG (~19 wt %) (fig. S31). Once exposed to UV, electron-hole pairs were generated on TiO2(B) nanosheets. Electrons were trapped in Ti-3d orbitals to form Ti3+ sites (35), and holes broke Ti-O bonds between glycolate and TiO2, leading to the formation of –OCH2CH2O• radicals (from I to II in Fig. 3B) (figs. S32 and S33). Because of the presence of α-H, –OCH2CH2O• was not stable and would thus undergo hydrogen transfer to give –OCH2•CHOH. According to density functional theory (DFT) calculations, such a process (from II to III in Fig. 3B) was predicted to be exothermic by 0.38 eV. Such UV-generated surface organic radicals are not unusual, as the oxidation potentials of most organic compounds lie below that of the holes in the valence band of TiO2 (3638). The ESR signals from the samples after washing and drying processes suggest that, once formed upon UV irradiation, the EG radicals on the surface of TiO2 nanosheets were quite stable.

To understand how EG radicals promoted the release of Cl from Pd sites, we also designed a stepwise route (fig. S34) to prepare the Pd1/TiO2 catalyst. UV treatment was first used to obtain TiO2-UV nanosheets containing EG radicals on their surfaces. H2PdCl4 was then introduced into the water dispersion of TiO2-UV. Our calculations showed that once adsorbed onto TiO2, each PdCl42– liberated two Cl ligands, yielding an intermediate with individual PdCl2 units adsorbed on TiO2 (IV in Fig. 3B) (fig. S35). Such a process was predicted to be slightly exothermic by 0.03 eV. Subsequently, the OH group in –OCH2•CHOH attacked its nearby Pd site by replacing one Cl, leading to the formation of PdCl1/TiO2 intermediate (V in Fig. 3B) with an exothermicity of 0.81 eV. As shown in fig. S35, PdCl1/TiO2 has three Ti-O bonds and one Pd-Cl bond. Experimentally, both EXAFS data and elemental analysis showed a Cl:Pd molar ratio of ~1:1 for PdCl1/TiO2 (Fig. 3C, figs. S36 and S37, and table S4), lower than the 2:1 molar ratio in PdCl2/TiO2 made from untreated TiO2. Moreover, mixing TiO2(B)-UV with H2PdCl4 solution decreased the amount of EG radicals, as evidenced by the reduced intensity of each ESR peak (fig. S38).

All of these results strongly confirmed that the UV-generated EG radicals facilitated the removal of Cl on Pd and stabilized individual Pd atoms by forming more Pd-O bonds. The remaining Cl on PdCl1/TiO2 could be easily removed by using H2 treatment, giving rise to H+ and Cl (from V to VII in Fig. 3B) (table S5). This result explained why treating the water dispersion of PdCl1/TiO2 resulted in a pH drop from 6.8 to 5.3 (Fig. 3D). Alternatively, further UV treatment completely removed Cl from PdCl1/TiO2 (fig. S39), also leading to the formation of Pd1/TiO2. The catalyst prepared in the stepwise procedure showed the same catalytic properties as that prepared by the one-pot method in which the aqueous mixture of TiO2 and H2PdCl4 was directly treated with UV (fig. S40). More important, the insight into the formation mechanism of Pd1/TiO2 allowed us to prepare the catalyst in large scale by using a continuous UV-flow reactor (fig. S41).

To evaluate the importance of EG radicals in the preparation of the atomically dispersed Pd1/TiO2 catalyst, we also synthesized EG-free TiO2 by calcination of TiO2(B) nanosheets at 350°C in air and used it as the support for the catalyst preparation. A photochemical strategy similar to that used in the one-step preparation of Pd1/TiO2 was applied to load Pd onto EG-free TiO2, but Pd NPs were formed (fig. S42); this result shows that surface EG helps to stabilize atomically dispersed Pd catalysts during their preparation. When surface EG was removed by calcination, the obtained Pd1/TiO2-cal catalyst displayed a substantially decreased activity with a TOF of only 1930 hours−1 (fig. S43).

It is generally accepted that H2 would undergo homolytic dissociation on conventional Pd particulate catalysts into H atoms with partially negative charge (Hδ–) (39). In this case, the presence of more than two Pd atoms in the vicinity is required. However, all Pd atoms in Pd1/TiO2 are individually dispersed, with no Pd-Pd pairs available for homolytic dissociation of H2, so the dissociation of H2 must go via an alternative pathway on Pd1/TiO2. According to our DFT calculations (Fig. 4A and figs. S44 to S46), H2 adsorbed on Pd was readily split into two H atoms. One of the H atoms moved to a nearby oxygen on EG to yield O-Hδ+, leaving the other H atom on Pd as Hδ– (Fig. 4A). This step was calculated to be exothermic by 0.69 eV and exhibited a barrier of 0.40 eV. We expected that both Pd-Hδ– and O-Hδ+ should then be involved in the hydrogenation catalysis. DFT calculations revealed that the hydrogenation of styrene using Pd1/TiO2 followed a stepwise process. Computationally, we considered two possible pathways (figs. S44 to S46), one beginning with Hδ– transfer from Pd to C=C and the other beginning with Hδ+ transfer. The first of these is energetically favorable, with a barrier of only 0.47 eV required for the Hδ– transfer from Pd to the terminal CH2 to make the half-hydrogenated intermediate, which in turn adds Hδ+ from a nearby O-H group by overcoming a barrier of 0.73 eV. This pathway leads to the formation of ethylbenzene and simultaneously recovers the Pd-EG interfaces.

Fig. 4 Catalytic mechanism of Pd1/TiO2 in hydrogenation reactions.

(A) Energies and model of intermediates and transition states in the heterolytic H2 activation process for Pd1/TiO2. (B) Primary isotope effect observed for Pd1/TiO2 in styrene hydrogenation. (C) First-run reaction performances for Pd1/TiO2, Pd/C, and H2PdCl4 in benzaldehyde hydrogenation.

To test the proposed mechanism, we explored the kinetic isotope effect (KIE) with the use of D2 in styrene hydrogenation. For Pd/C, the reaction was slowed down by a factor of 1.43 (fig. S47) as a result of the zero-point energy difference between isotopic isomers. However, on Pd1/TiO2, a larger KIE was observed (ratio of reaction rates using H2 and D2, kH/kD = 5.75) (Fig. 4B) because the bond cleavage was O-D rather than Pd-D in the rate-determining step. Both our IR spectroscopy and nuclear magnetic resonance measurements, which were performed with deuterium-labeled reagents, confirmed the proposed mechanism (figs. S48 and S49). Such a large KIE in hydrogenation caused by the participation of both Hδ– and Hδ+ has usually been observed on homogeneous catalysts (e.g., Au, Pd, and Ru complexes) (7, 40, 41) but has not been reported on heterogeneous Pd catalysts. In this regard, atomically dispersed metal catalysts can share the same hydrogenation mechanism as homogeneous catalysts.

Because the heterolytic activation of H2 yielded both Hδ– and Hδ+ at the Pd-O interface, Pd1/TiO2 should allow better hydrogenation of polar unsaturated bonds. As expected, in the hydrogenation of benzaldehyde, we observed a much superior catalytic performance by Pd1/TiO2 (Fig. 4C and fig. S50). Pd1/TiO2 readily converted all of the benzaldehyde into benzyl alcohol in 3.5 hours at room temperature with a TOF of 1002 hours−1. No decay in the catalysis was observed after the catalyst was used for five cycles. In comparison, both Pd/C and H2PdCl4 showed negligible activity under the same catalytic condition, with TOF less than 18 hours−1. This work demonstrates that upgrading catalytically active components from nanoparticles to single atoms not only boosts the catalytic reaction because of the high atom efficiency, but also endows atomically dispersed catalysts with catalytic capability that conventional nanocatalysts do not possess.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/352/6287/797/suppl/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S50

Tables S1 to S5

References (4255)

REFERENCES AND NOTES

  1. See supplementary materials on Science Online.
  2. Acknowledgments: Supported by Ministry of Science and Technology of China grant 2015CB932303; National Natural Science Foundation of China grants 21420102001, 21131005, 21390390, 21133004, 21373167, 21573178, and 21333008; a NSERC CGS Alexander Graham Bell scholarship (D.M.C.); and a NSERC Discovery grant (P.Z.). We thank the XAFS station (BL14W1) of the Shanghai Synchrotron Radiation Facility.
View Abstract

Stay Connected to Science

Navigate This Article