Catalytically active Au-O(OH)x- species stabilized by alkali ions on zeolites and mesoporous oxides

See allHide authors and affiliations

Science  19 Dec 2014:
Vol. 346, Issue 6216, pp. 1498-1501
DOI: 10.1126/science.1260526


We report that the addition of alkali ions (sodium or potassium) to gold on KLTL-zeolite and mesoporous MCM-41 silica stabilizes mononuclear gold in Au-O(OH)x-(Na or K) ensembles. This single-site gold species is active for the low-temperature (<200°C) water-gas shift (WGS) reaction. Unexpectedly, gold is thus similar to platinum in creating –O linkages with more than eight alkali ions and establishing an active site on various supports. The intrinsic activity of the single-site gold species is the same on irreducible supports as on reducible ceria, iron oxide, and titania supports, apparently all sharing a common, similarly structured gold active site. This finding paves the way for using earth-abundant supports to disperse and stabilize precious metal atoms with alkali additives for the WGS and potentially other fuel-processing reactions.

Dispersing catalytic gold as widely as possible

In order to maximize the activity of precious metals in catalysis, it is important to place the metal on some support with a high surface area (such as a zeolite) and to maintain the metal as small clusters or even atoms to expose as much metal as possible. The latter goal is more readily achieved with oxides of reducible metals such as cerium or titanium than with the aluminum and silicon oxides that make up most zeolites and mesoporous oxides. Yang et al. show that sodium and potassium can stabilize gold along with hydroxyl and oxo groups to create highly active catalysts for the water-gas shift reaction at low temperatures, a reaction that can be useful in applications such as fuel cells.

Science, this issue p. 1498

The water-gas shift (WGS) reaction (CO + H2O → CO2 + H2) is an important reaction for hydrogen upgrading during fuel gas processing. Emerging applications in fuel cells require active, nonpyrophoric, and cost-effective catalysts. Along with a new group of platinum catalysts with atomically dispersed Pt sites to maximize activity and catalytic efficiency (13), the lower apparent activation energy Ea for the WGS reaction (~45 kJ/mol) for gold (Au) versus ~75 kJ/mol for platinum (35) can be exploited for low-temperature WGS and other reactions (6, 7). Low-temperature activity is important to avoid multiple-treatment units in practical low-temperature proton-exchange membrane (PEM) fuel cell systems, whereby the deleterious CO should be totally removed for stable, long-term operation. The active Au species in the WGS catalysts are atomic species anchored through –O ligands to different supports such as ceria (3, 8, 9), iron oxide (1012), lanthana (13), and titania (4), and the number of the active Au sites can be increased through a variety of catalyst preparation protocols. Gold nanoparticles (Au NPs) that can form during catalyst preparation are spectator species in these chemistries (3, 4, 10), in that most of the Au atoms are not activated by the support. Thus, the approach of “cage encapsulation” of Au NPs in mesoporous supports is not advantageous for the stability of the active (atomically dispersed) Au sites.

Other approaches—for example, AuCl3 vapor produced by sublimation and introduced into various zeolites (14, 15)—may be used to produce active Au(I)-Cl species for ambient-temperature NO reduction to N2O by CO. Mohamed and Ichikawa (16) have shown that the Au(I) species are the main active sites for the WGS reaction at temperatures as low as 50°C. Because these sites are not chloride-free (Au-Cl bonds exist) and have weak chemical binding to the zeolites, the Au(I) sites are easily reduced to inactive Au(0) and form Au NPs upon increasing the temperature to only 100°C (16). Similarly, low stability of gold on zeolites was found by Gates and co-workers (17, 18). Careful anchoring of mononuclear Au(III) complexes from organometallic precursors produced chloride-free single-atom Au(III)-O-NaY catalytic centers that were active for CO oxidation but unstable at 25°C and 760 torr, losing ~75% of their initial activity after 15 min on stream (17). Finally, attempts to ion exchange gold in zeolites have been unsuccessful. Thus, gold ions in zeolites tend to be unstable toward aggregation in realistic reaction gas environments at temperatures above the ambient, an issue already understood for other inert supports such as silica or alumina, minimally interacting with gold (19). Hence, it is difficult to determine if the gold catalysts operate through similarly structured Au-O(OH)x- species on inert supports as in the Au-CeOx, Au-FeOx, and Au-TiOx systems (20).

To study the nature of the active gold sites on inert supports, it is important to maximize the number of the atomically dispersed gold sites and fully eliminate the formation of Au NPs. Titania is inferior to ceria and iron oxide in that Au NP growth occurs rapidly on its surfaces (21), but with special ultraviolet (UV)–assisted preparation methods, mononuclear Au-O(OH)x- species can be stabilized on titania up to 1.2 weight % (wt %), and the cations remain stably anchored and active for the WGS reaction from ~ 80° to 250°C (4). Alkali ion addition was investigated in this work as a means to boost further the number of stable mononuclear Au-O(OH)x- species. This was reported as a successful approach to prepare single-site active Pt-O(OH)x-(Na or K)y species on silica and alumina, which were activated at low temperatures (~100°C) for the WGS reaction and were stable to temperatures exceeding 300°C and for many hours (1). Washing of the surfaces could not remove the alkali ions stably associated with the Pt ion through -O linkages. Zugic et al. showed that the Pt-O(OH)x-(Na or K)y species could be prepared, stabilized, and similarly activated for the WGS reaction on inert (800°C-annealed) carbon nanotube surfaces (2). However, the extension of the platinum findings to gold is neither obvious nor anticipated. A few reports exist on alkali (and alkaline earth) addition to gold to structurally stabilize small Au NPs (1 to 3 nm) on alumina for ambient-temperature CO oxidation (22). Miller et al. (19) adopted a NaOH wash to remove the adsorbed Cl ions from the Au/Al2O3 catalysts and estimated by extended x-ray absorption fine structure (EXAFS) that the gold was 100% dispersed on these washed samples under mild reduction (up to 200°C) or oxidation (up to 225°C), but their stability under reaction conditions was not reported.

We show how to use alkali addition to activate and stabilize atomic Au for the WGS reaction even on inert zeolite (KLTL) and mesoporous [Si]MCM-41 silica materials. The WGS activity was measured to be comparable to that of Au on reducible oxide supports, and good stability was found up to 200°C (Table 1). The Ea values measured for the reaction over alkali-stabilized gold on the inert supports are all 45 ± 5 kJ/mol, similar to those on the Au-CeOx, Au-FeOx, and Au-TiOx systems (20), indicating that the gold active sites are of similar structure on all support types. It is not straightforward to produce the active alkali-stabilized gold centers on the inert supports, as the alkali ions must interact with the gold, not the support. Notably, the KLTL-zeolites used here have an abundance of potassium ions (16.8 wt %), but gold addition from a typical precursor (e.g., HAuCl4) by incipient wetness impregnation (IWI) or deposition precipitation fails to prepare an active catalyst. However, using IWI of KAu(CN)2 on the KLTL-zeolites followed by solid-state impregnation of KOH (Au:K = 1:10) did form an active Au-K/KLTL catalyst after careful heating [details of the preparation methods are described in the supplementary materials (23)]. In another preparation, a gold sol formed from HAuCl4 with NaOH (Au:Na = 1:10) at pH ~14 was used to prepare an active Au-Na/[Si]MCM41 catalyst (table S1). For other supports, including a hydrothermally treated alumina in NaOH, see preparation details in the supplementary materials and in fig. S1 and tables S1 and S2.

Table 1 Composition and reaction activity of supported gold catalysts.

View this table:

The preparation protocols show that the requirement for an active catalyst is that the alkali ions are linked to the atomic gold through –O ligands and not merely be present on the support. It has been widely reported that the support plays a crucial role in the WGS reaction, providing facile dissociation of water molecules to supply –OH species to the vicinal gold sites where CO is adsorbed (24), but the active sites were not resolved. These may still involve just the Au-(OH)x species attached to the support through –O ligands. We present evidence here for the latter by surrounding the gold atom with a large number (8 to 10) of Na or K ions through –O ligands. The choice of the supporting surface (zeolite, silica, alumina, etc.) is then unimportant for the chemistry.

The lack of sensitivity of Ea for the WGS reaction on the type of support is a strong first indication of a structurally similar gold active site present on all supports. Corroborating this finding, kinetics measurements found that the reaction order for H2O is 0.7 to 0.8 for all the catalysts (with or without alkali) used in this work (fig. S2), demonstrating that the role of water is common on all the gold catalysts. The alkali addition modifies the support properties at higher temperatures; for example, after a thermal treatment to 600°C, Amenomiya and co-workers found that alkalized alumina was activated for the WGS reaction above 400°C with Ea of ~ 80 kJ/mol (25, 26).

CO temperature-programmed reaction (TPR) tests were conducted to titrate the WGS-active hydroxyls on the alkali-stabilized gold sites (up to 350°C) (Table 1, table S2, and fig. S3). The presence of alkali markedly increased the amount of these hydroxyls, and correspondingly the overall WGS activity of the catalyst within the same temperature window. These hydroxyls are regenerable, as shown by consecutive CO-TPR cycles with intermittent rehydration of the catalyst at 25°C (fig. S3). In the absence of gold, the addition of K+ only provides trace amounts of dry [O] through the surface transformation of –OK to K2O (27). With the gold present, as in Au-K/KLTL, gold associated with the potassium shows a large amount of CO2 formation from active –OH species corresponding to a Au:K atomic ratio of 1:8. This was accompanied by the production of a large amount of H2 (1/2 of CO2) from the –OH species (fig. S3). Notably, from the samples with fully dispersed gold (0.25Au-K/KLTL and 0.25Au-Na/[Si]MCM41) and earlier reports (4), we found that the total activity is proportional to the number of the surface –OH species, whereas the activity per gold atom (turnover frequency) is the same for all gold catalysts irrespective of the support (fig. S4). The Au-K/KLTL sample also shows good stability in 100 hour-long operation in a reformate-type gas mixture (fig. S5).

Electron microscopy studies revealed that alkali ion addition markedly increased the dispersion of gold on all the inert supports that we investigated in this work (Table 1, Fig. 1, and figs. S6 to S9). For example, 72% of the gold counted in the images was present as isolated atoms away from each other on the surface of the 0.25Au-Na/[Si]MCM41 sample (Fig. 1 and fig. S6). A minority of subnanometer gold clusters present on the same sample did not have the packed gold atom structure of Au NPs. These species appear to comprise a few atoms of gold anchored close to each other but nonaggregated. The presence of the surrounding alkali atoms could not be determined by imaging because of the low contrast of these light elements. EXAFS analysis under in situ conditions for the working catalysts further confirmed that the gold species were atomically dispersed and associated with alkali ions before and after reaction (table S3 and fig. S10). For both the KLTL-zeolite and [Si]MCM41-supported samples, the Nax-Oy linkages to gold effectively reduced the Au-Au coordination number from 11 or 12 to the 3 or 4 range, which means that 100% dispersion of Au had been achieved (19). Although microscopic analysis (transmission electron microscopy or scanning transmission electron microscopy) provides a number-weighted particle distribution, the volume-weighted distribution provided by EXAFS can be heavily skewed by a few large particles (23). Along with the EXAFS results showing an enrichment of the Au-O shell for the alkali-containing samples, in situ x-ray absorption near-edge structure spectra (fig. S11) and x-ray photoelectron spectroscopy data (fig. S12) show the cationic nature of the gold species, whereas the alkali-free, catalytically inactive samples contain metallic Au NPs exclusively. The extra K+ ions stabilizing the Au-Ox site in the Au-K/KLTL sample are different from the ion-exchanged K-O-Al sites in the as-received KLTL zeolite, but are associated with –O (28), –OH, and H2O in their vicinity (29), as indicated by the K2p x-ray photoelectron spectra (fig. S12). These results demonstrate the substantial interaction between gold and potassium through oxygen bonding.

Fig. 1 Aberration-corrected high-angle annular dark-field–STEM images of the 0.25Au-Na/[Si]MCM41 catalyst.

The circles are drawn around isolated gold atoms. The size distribution (inset) is based on >150 observed gold species counted from the high-magnification images (recorded at 8 to 10M× original magnification).

We thus propose a cationic gold-centric active site [Au-O(OH)x- species] for the WGS reaction that is stabilized by several alkali ions via -O- linkages. To gain further insights, we used density functional theory (DFT) calculations, performed in the framework of ab initio molecular dynamics (30), by using the VASP code (31) (details of the computational methods are described in the supplementary materials). To create candidate structures for the Au-O(OH)x-Nay system, we began by inspecting a series of AuNax clusters (x = 1 to 10, without oxygen). Out of these, the AuNa9 precursor was selected for further studies because of its high stability (9 is the maximum number of Na atoms that can fit in a single shell around Au; beyond 9, additional Na atoms must be accommodated in a second shell around the gold atom; see supplementary materials for details) and because its Na:Au ratio closely matches the experimental facts/environment. To create the cationic gold species detected experimentally and to account for the involvement of oxygen in the active site, electron-withdrawing groups (O/OH) were then added to the AuNa9 cluster, and the Bader charge was calculated (32). For reference, the calculated Bader charges of Au in bulk gold oxides Au2O and Au2O3 (used as references for Au(I) and Au(III), respectively) are +0.41 and +1.11, respectively. In a systematic study of cluster sites of the general composition AuOx(OH)yNa9, several clusters in which the charge of the Au atom is similar to that of Au in bulk Au2O and Au2O3 were identified (the circled models in Fig. 2). To evaluate if any of these structures that feature an oxidized Au atom would be active for WGS, the thermochemical properties relevant to the key steps of WGS (including CO and H2O adsorption, OH binding, and H2O activation) were calculated on these candidate AuOx(OH)yNa9 structures and compared with those on extended Cu(111) and Cu(211) surfaces. As Cu is one of the best low-temperature WGS catalysts, these are the properties that are most relevant to mimic.

Fig. 2 Bader charge of Au in AuOx(OH)yNa9 cluster sites.

For reference, the Bader charges of Au in bulk Au2O (+0.41) and Au2O3 (+1.11) are given by the dashed horizontal lines. Au, Na, O, and H atoms are shown with yellow, purple, red, and white spheres, respectively. The clusters shown are examples of promising structural candidates for the WGS active site. The circled data correspond to clusters that have Au oxidation states between that of Au2O and Au2O3.

Based on both the Bader charge and the thermochemical analysis (see table S4), AuO7(OH)2Na9, AuO6(OH)2Na9, AuO3(OH)7Na9, and AuO2(OH)9Na9 (structures of the first two clusters on this list are shown in Fig. 2, and the last two are shown in fig. S13) were identified as promising candidates for the active WGS reaction site. Compared with Cu(111), these clusters bind both OH and CO more weakly. These structures (Fig. 2 and fig. S13) share some common geometric features: The –O ligands bind directly to the central gold atom; OH groups are primarily bound on the many three-fold “Na3” sites that surround the Au atom; and Na atoms are linked to the Au atom through –O ligands.

The O/OH groups in the clusters play vital roles in H2O dissociation. First, O/OH stabilize H2O binding through hydrogen bonding (33). As a result, H2O capture should be easier on the clusters than on Au(111). Second, the hydrogen atom released during H2O dissociation prefers to bind to a surface oxygen atom to form OH, leading to a much more exothermic water dissociation event than on clean Au(111) or Cu(111) surfaces. The more exothermic activation should increase activity for H2O dissociation. Although CO binding on the gold cation is weak, it is possible. The colocation of adsorbed CO and the –OH groups bound on the threefold “Na3” sites that surround the gold atom offers the possibility for facilitating COOH formation, a critical intermediate in the WGS reaction (34). An analog of this structure was recently proposed for single gold atoms stabilized over uncapped oxygen and surrounded by Fe(III) cations on Fe3O4(111) surfaces (12). Notably, the catalytic activity of Au on iron oxide (35) is similar to that of Au in the alkali-stabilized structure reported here, as shown in Fig. 3.

Fig. 3 TOF plot for the WGS reaction over samples with atomically dispersed gold in a reformate-type gas mixture of 11% CO, 26% H2O, 7% CO2, and 26% H2-He.

The data for the 1.16Au/TiO2 (G5, UV) sample are from (4), for the 0.50Au/CeO2 (La doped CeO2, DP) sample from (8), and for the 0.12Au/Fe2O3 (DP) sample from (35). The experimental error for the reaction rate measurements is less than 10%. The R2 for the linear fit is 0.9927.

In Fig. 3, we scaled the steady-state reaction rates by the amount of gold loading for the atomically dispersed gold catalysts studied here, and for other gold catalysts reported to comprise only atomic gold on other supports—e.g., after leaching of gold particles—and cast them in terms of a turnover frequency (TOF) in an Arrhenius plot. In addition to the similar Ea values, the closeness of the TOFs over gold catalysts prepared on different supports and from different precursors, and subjected to different treatment methods, is noteworthy. With these findings, we conclude that, being structurally similar as on CeO2, Fe2O3 and TiO2, single-site cationic Au-O(OH)x- species, stabilized by a number of alkali ions in the form of AuOy(OH)z(Na or K)x clusters, may be formed in appreciable amounts on inert supports. These species are highly active for the WGS reaction, the single gold atom maximizing the catalyst efficiency.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S13

Tables S1 to S4

References (3646)

References and Notes

  1. Supporting materials are available on Science Online.
  2. Acknowledgments: The financial support by the U.S. Department of Energy, Office of Basic Energy Sciences (DOE-BES) under grant DE-FG02-05ER15730 is gratefully acknowledged. M.Y. thanks H. Luo (Massachusetts Institute of Technology) for some of the surface area and pore structure analysis, C. Wang (Tufts University) for acquiring the XRD data, and B. Reinhart [Argonne National Laboratory (ANL)] for assisting with the in situ XAS experiments. Microscopy research was sponsored in part by a user project supported by the Center for Nanophase Materials Sciences (CNMS), which is sponsored at Oak Ridge National Laboratory (ORNL) by the Scientific User Facilities Division, DOE-BES, and by the U.S. DOE Office of Energy Efficiency and Renewable Energy, Vehicle Technologies Office, Propulsion Materials Program. The XAS research is sponsored by the Advanced Photon Source at ANL under contract DE-AC02-06CH11357. Work at the University of Wisconsin–Madison was supported by DOE-BES, Office of Chemical Sciences. Computational work was performed in part with supercomputing resources from the following institutions: Environmental Molecular Sciences Laboratory (EMSL), a national scientific user facility at Pacific Northwest National Laboratory (PNNL); the Center for Nanoscale Materials (CNM) at ANL; the CNMS at ORNL; and the National Energy Research Scientific Computing Center (NERSC). EMSL is sponsored by the Department of Energy’s Office of Biological and Environmental Research located at PNNL. CNM and NERSC are supported by the U.S. Department of Energy, Office of Science, under contracts DE-AC02-06CH11357 and DE-AC02-05CH11231, respectively. Work at Sydney was supported by USyd Early Career Researcher Scheme. The authors declare no conflict of interest.
View Abstract

Stay Connected to Science

Navigate This Article