Active Nonmetallic Au and Pt Species on Ceria-Based Water-Gas Shift Catalysts

See allHide authors and affiliations

Science  15 Aug 2003:
Vol. 301, Issue 5635, pp. 935-938
DOI: 10.1126/science.1085721


Traditional analysis of reactions catalyzed by supported metals involves the structure of the metallic particles. However, we report here that for the class of nanostructured gold– or platinum–cerium oxide catalysts, which are active for the water-gas shift reaction, metal nanoparticles do not participate in the reaction. Nonmetallic gold or platinum species strongly associated with surface cerium-oxygen groups are responsible for the activity.

The heterogeneously catalyzed water-gas-shift (WGS) reaction (CO + H2O ↔ CO2 + H2) is a key step in fuel processing to generate H2. Such heterogeneous catalysts should combine both high activity and structural stability in air and in cyclic operation; these are stringent requirements not met by the commercially available low-temperature WGS catalysts. A new class of WGS catalysts based on cerium oxide (ceria) has been investigated extensively in recent years (16). To provide low-temperature WGS activity, Pt-group metals (PM), Au, or Cu are suitably added in amounts that vary from ∼1 to 10 weight percent (wt %). A critical problem with Pt-ceria catalysts is their prohibitive economics (3), due to the cost of Pt, even if their issues of deactivation with time-on-stream are resolved. However, ceria containing only trace amounts of Pt would be economical. Similarly attractive would be ceria containing base metals or oxides. We have reported recently that an excellent shift catalyst results from supporting Au or Cu on nanocrystalline ceria (46). This type of catalyst, if properly developed, would of course be much more economical, i.e., practical for large-scale fuel cell application. We show here that low loadings of metal can be as effective as much higher loadings.

The PM-ceria catalysts have received considerable attention because of their use in the automobile catalytic converter (7, 8). It is widely accepted that the oxygen in ceria plays an important role in the reaction pathway (1, 4, 9). However, identification of the active sites for low-temperature CO oxidation, the WGS reaction, and other oxidation reactions on PM-ceria catalysts remains an issue of contention. In most reports, the active sites are placed at the metal-ceria interface (1, 9), whereas in others Pt ions dispersed on the surface of ceria are assumed to be active (10). Encapsulation of Pt by reduced ceria nanoparticles has also been proposed (11).

In the work reported here, all catalyst components were nanocrystalline. Nanocrystalline ceria can be prepared by various techniques (2, 4). Ceria particles with diameters of less than 10 nm have markedly higher electronic conductivity over that of centered ceria (12), and doping with a rare earth oxide, such as La2O3, can be used to create oxygen vacancies and to stabilize ceria particles against sintering (13). We prepared such nanosized La-doped ceria and deposited Au or Pt on it. After thermal annealing, these materials had excellent and as-yet-unexplained properties for the low-temperature WGS reaction. Au or Pt exists as nanosized particles and in ionic state in these catalysts. The catalytic activity was not affected by the removal of metallic Au or Pt particles by cyanide leaching. Thus, metallic nanoparticles are not necessary for the activity; they are mere spectators in the WGS reaction. Nonmetallic Au or Pt species embedded in ceria catalyze the reaction of CO with H2O.

We prepared most of the Au-ceria catalysts by the same technique; ceria (doped with 10 atom % La) was synthesized by urea gelation/coprecipitation (UGC) (4) and then calcined in air at 400°C for 10 hours. This treatment produced ceria with a mean particle size of ∼5 nm (5), with a surface area of ∼150 m2/g. Au was then applied onto ceria by deposition-precipitation (DP) (5) at room temperature, through dropwise addition of HAuCl4 into a suspension of the ceria particles in an aqueous solution of (NH4)2CO3 at a constant pH (∼8). After several washes and drying, the Au-ceria particles were calcined in air at 400°C for 10 hours. Most of the Au thus prepared was in the form of metal nanoparticles with an average size of ∼5nm(5, 6). The deposition step had a negligible effect on the total surface area of ceria. For comparison, we prepared a Au-ceria sample by a single coprecipitation (CP) step. This involved mixing an aqueous solution of HAuCl4, Ce(NO3)3, and La(NO3)3 with (NH4)2CO3 at 60° to 70°C, keeping it at a constant pH (∼8), and aging the precipitate at the same temperature for 1 hour. Leaching of gold took place in an aqueous solution of 2% NaCN at room temperature. Sodium hydroxide was added to keep the pH at ∼12. [This same process is used to extract gold during gold mining (14).] No Ce or La was found in the leachate. The leached samples were washed, dried (at 120°C for 10 hours), and heated in air (at 400°C for 2 hours). More than 90% of the Au loading was removed from the ceria by this leaching procedure. Scanning transmission electron microscopy and energy-dispersive x-ray spectroscopy showed no Au particles remaining, only what appeared to be very fine clusters or atomically dispersed Au. X-ray photoelectron spectroscopy (XPS) identified ionic Au as the major or only Au species present in the leached materials (Fig. 3A).

A similar procedure was used to remove excess Pt from the ceria surface. First, Pt-ceria was prepared by incipient wetness impregnation (IMP). La-doped ceria (CL) powders were prepared by UGC as described above. They were then impregnated with an aqueous solution of H2PtCl6 of appropriate concentration, the volume of which equaled the total pore volume of ceria. After impregnation, the samples were degassed and dried at room temperature under vacuum. After drying in a vacuum oven at 110°C for 10 hours, the samples were crushed and calcined in air at 400°C for 10 hours. Calcined Pt-ceria samples were leached by the same procedure as the Au catalysts; the leached sample is denoted as Pt-CL(IMP, NaCN1). To further reduce the amount of Pt, Pt-CL(IMP, NaCN1) was leached in a 2% NaCN solution at 80°C for 12 hours. The corresponding sample is denoted as Pt-CL(IMP, NaCN2). The properties of the thus-prepared Au- and Pt-ceria samples are presented in Table 1.

Table 1.

Physical properties of ceria-based catalysts. All samples were calcined at 400°C; CL is Ce(10 atom % La)Ox, calcined at 400°C, for 10 hours. Numbers in sample names represent atom %. NM, not measured; ND, not detectable; NA, not applicable.

Sample Surface area (m2/g) Surface metal contentView inline (atom %) Au or Pt Bulk composition (atom %)View inline Particle sizeView inline (nm)
Metal (Au or Pt) Ce La Metal (Au or Pt) CeO2
<111> <220>
4.7 Au-CL (DP) 156.1 1.60 4.71 87.88 7.41 5.0 5.2 4.9
0.4 Au-CL (DP) (NaCN) 157.9 0.61 0.44 91.24 8.32 ND 5.2 4.9
2.8 Au-CL (DP) 159.2 1.58 2.81 89.16 8.03 4.7 5.0 4.9
0.2 Au-CL (DP) (NaCN) 162.2 0.43 0.23 93.10 6.67 ND 5.0 4.9
3.4 Au-CeO2View inline (DP) 25.9 NM 3.36 96.64 0 4.0 21.1 20.3
0.001 Au-CeO2 (DP)View inline (NaCN) 28.0 NM ∼0.001 ∼99.999 0 ND 21.0 20.4
CL (UGC) 156.9 NA 0 92.62 7.38 NA 5.1 4.8
4.4 Au-CL (CP) 47.8 3.29 4.35 88.00 7.65 12.9 7.2 6.3
0.7 Au-CL (CP) (NaCN) 47.5 0.24 0.67 91.52 7.82 ND 7.0 6.0
3.7 Pt-CL (IMP) 129.8 1.63 3.67 88.83 7.50 2.5View inline 6.2 6.1
2.7 Pt-CL (IMP, NaCN1) 147.5 1.79 2.70 89.78 7.52 ND 6.2 6.1
1.5 Pt-CL (IMP, NaCN2) 103.2 0.82 1.50 90.86 7.64 ND 6.2 6.1
  • View inline* Surface metal content was determined by XPS.

  • View inline Bulk composition was determined by ICP.

  • View inline Particle size was determined by XRD with the Scherrer equation.

  • View inline§ CeO2 was calcined at 800°C.

  • View inline Particle size was determined by HRTEM.

  • Arrhenius-type plots of the WGS reaction rate measured over the as-prepared Au-ceria catalysts and the Au-free ceria (CL) are shown in Fig. 1. The reacting gas mixture simulates a reformate gas composition: 11% CO, 7% CO2, 26% H2, and 26% H2O, in an inert gas carrier. Activation of catalysts was not necessary (15). Similar rates of CO2 production (per square meter of catalyst surface area) were measured over the parent [4.4 (CP), 4.7 (DP), or 2.8 (DP) atom % Au] and the corresponding leached (0.7, 0.44, or 0.23 atom % Au) ceria catalysts. The apparent activation energy (Ea) for the reaction was the same for parent and leached catalysts: 47.8 ± 1.5 kJ/mol for the DP samples and 36.8 ± 0.9 kJ/mol for the CP samples. The rate over the nanosized CL sample was much lower than the Au-containing CL over the temperature range of interest, with an Ea of 83 kJ/mol. Fig. 1 also shows the rate measured over a commercial Cu-ZnO-Al2O3 (United Catalysts Inc., G-66A) low-temperature WGS catalyst, which contains 42 wt % Cu. Although the rate was greater over this catalyst, its use in fuel cell applications is highly unlikely because of its air sensitivity and its narrow operating temperature window. Moreover, a careful activation in H2 is required for Cu/ZnO catalysts. However, ceria-based WGS catalysts require no activation and are not air-sensitive.

    Fig. 1.

    WGS rates measured in a reformate-type gas composed of 11% CO, 7% CO2, 26% H2, 26% H2O, and balance He [see Table 1 for sample properties, (15) for details]. Solid squares, 4.4 atom % AuCe(La)Ox (CP); open squares, 0.7 atom % AuCe(La)Ox (CP, leached); solid triangles, 4.7 atom % AuCe(La)Ox (DP); open triangles, 0.44 atom % AuCe(La)Ox (DP, leached); solid circles, 2.8 atom % AuCe(La)Ox (DP); open circles, 0.23 atom % AuCe(La)Ox (DP, leached); asterisks, Ce(La)Ox; diamonds, G-66A (42 wt % CuO, 47 wt % ZnO, 10 wt % Al2O3, surface area 49 m2/g). T(1/K), temperature in Kelvin.

    The reaction pathway on the Au-ceria catalysts was different from that on Au-free ceria (Fig. 1). Also, only the Au species present on the leached catalyst must be associated with the active sites, because the extra Au present in the parent material did not increase the rate; nor did it change the Ea for the reaction. If we assume complete dispersion of Au in the leached catalysts, we can calculate the turnover frequency (TOF) from the data in Fig. 1. For example, at 300°C, the TOF is 0.65 molecules of CO2 per Au atom per second.

    In kinetic studies with the Pt-ceria catalysts (16) (fig. S1), the Ea over the parent (3.7 atom % Pt) and the leached Pt-ceria (2.7 or 1.5 atom % Pt) was the same, 74.8 ± 0.6 kJ/mol. The WGS rate over these samples was similar. The isokinetic temperature for the Pt- and Au-ceria (DP) samples is 250°C. Transient light-off curves for the WGS reaction over the Pt-ceria catalysts (Fig. 2) were collected in temperature-programmed reaction mode. These profiles were reproduced after samples were cooled down from the high end-point temperature and the transient test was repeated. The light-off temperature was lowest for the catalyst that contained the lowest amount of Pt (by leaching). Thus, the removed Pt was not important for the reaction, and the leaching process must have increased the number of active sites.

    Fig. 2.

    Temperature-programmed reaction of as-prepared and leached Pt-ceria catalysts in a 2% CO-3% H2O-He gas (see Table 1 for sample properties).

    The oxidation states of Au and Pt in both the parent and leached ceria samples were checked by XPS (Fig. 3). The common features in both systems were (i) the existence of ionic states (Au+1,+3 and Pt+2,+4) both before and after leaching, and (ii) the complete removal of metallic Au or Pt nanoparticles after the leaching step. No Ce or La loss took place during the leaching step, as verified by inductively coupled plasma (ICP) analysis of the leachate solutions. The absence of Au or Pt particles on the leached ceria samples was also confirmed by high-resolution transmission electron microscopy (HRTEM). The intensities shown in Fig. 3A cannot be used to compare the amounts of Au between parent and leached samples. In fact, the surface metal content of the parent DP and CP samples was grossly underestimated (Table 1), because average metal particle sizes greatly exceeded the electron escape depth. The agreement is better for the leached Au-ceria samples. Finally, all Pt-ceria samples show much less Pt on the surface than expected on the basis of the ICP analysis and the surface area of each sample. In both Au- and Pt-ceria, diffusion of Au or Pt ions into subsurface layers of ceria is plausible.

    Fig. 3.

    X-ray photoelectron spectra of as-prepared and leached samples. (A) Au-ceria. (B) Pt-ceria [see Table 1 for sample properties, (17) for comments]. X5 and X25 represent signal magnification.

    The 4.4 atom % Au-CL catalyst prepared by CP shows Au0 binding energies at 83.8 and 87.4 eV; this sample contains metallic Au particles with a mean size of 12.2 nm (Table 1). Leaching removed all metallic gold (Fig. 3A) for the 0.7 atom % Au-CL sample. Both Au+1 and Au+3 were present in the leached sample. The 4.7 atom % Au-CL catalyst prepared by DP shows Au0 lines as well as ionic Au. The corresponding leached material shows ionic Au binding energies, as well as a positively shifted (by ∼0.1 eV) binding energy of Au0. This shift is within the experimental error of the analysis. Deconvolution of the spectra shows that the zerovalent species amount to only 14% of the total Au present in the leached 0.44 atom % Au-CL sample (Fig. 3A) (1719).

    It may be argued that the oxidic Au observed in our samples is due to the preparation conditions (air calcination at 400°C), and that during reaction under net reducing conditions, zerovalent Au would dominate. This possibility would require further studies. An important observation, however, is that the used catalyst, after more than 20 hours at reaction conditions, cannot be further leached; i.e., even if Au changes oxidation state during reaction, it does not migrate to form metallic particles. XPS analysis of Au-ceria catalysts after 15 hours of use in the reaction gas mixture in Fig. 1 showed predominance of ionic Au (16) (fig. S2). The Au-O-Ce structures are stable under the conditions in this work. Similar arguments can be made for the Pt-ceria catalysts. For this type of material, surface Pt-O phases strongly associated with ceria have been reported (10).

    The use of dry CO in temperature-programmed reduction (TPR) (CO-TPR) identified oxygen species of importance to the low-temperature WGS reaction on the parent and leached catalysts. Various types of oxygen have been identified on ceria (20, 21), ranging from weakly bound adsorbed oxygen to surface capping oxygen to lattice oxygen, depending on the operating temperature. A synergistic redox model for metal-CeO2 has been proposed, in which the metal particle participates by providing adsorption sites for CO (1, 4, 22, 23), while ceria supplies the required oxygen. This simple model does not provide atomic-level understanding and mechanistic resolution of several key questions; most importantly, it assigns the CO adsorption sites on metal particles. However, as Figs. 1 and 2 show, the WGS activity of metal-free (leached) ceria is similar to that of the metal-containing samples.

    CO-TPR (24) results from fully oxidized parent and leached Au-ceria (DP) samples and from the CL material are shown in Fig. 4. The first CO2 peak produced on the parent Au-ceria sample is absent in the leached sample and in the Au-free, CL material. This peak is thus assigned to oxygen adsorbed on metallic Au nanoparticles, which are present only on the parent 4.7 atom % Au-CL sample. The high-temperature oxygen species, Ob, is of similar reducibility in all three samples. Thus, the presence of Au did not affect the bulk (lattice) oxygen of ceria. However, the reducibility of the surface oxygen species of ceria, Os1 and Os2, was greatly increased for both Au-containing samples (Fig. 4). This result correlates well with the markedly higher WGS activity of the Au-CL catalyst compared to that of the CL material shown in Fig. 1.

    Fig. 4.

    CO-TPR profiles of as-prepared and leached Au-CL (DP) and CL samples; 10 mole % CO in He, 50 cm3 per min [see Table 1 for sample properties, (24) for details].

    The appearance of H2 along with CO2 elution during CO-TPR is attributed to surface hydroxyls that remain in ceria even after the oxidation pretreatment step in a dry O2/He mixture at 350°C (2). Indeed, when we repeated the CO-TPR after reoxidation at 400°C, very little H2 was produced, and by the fourth cycle, only trace amounts of H2 evolved. The amount of CO2 eluted in all cycles was the same, and its production began and peaked at the same temperatures as those shown for the first cycle in Fig. 4. A higher amount of CO2 was eluted from the leached catalyst (Fig. 4, the area under the Os1 peak). This difference may be due to unmasking of sites after the metallic particles that cover them are leached away.

    How do Au ions or adatoms interact with ceria to weaken both its Os1 and Os2 surface oxygens? A distribution of electronic charges between atomic Au or a small cluster of Au atoms and ceria could weaken the Ce-O bond. Evidence from H2-TPR and separate pulse reactor experiments with CO in our lab (6) strongly suggests that Au increases the amount of surface oxygen in ceria. This increase can occur through partial lattice filling of vacant Ce sites with Auδ+, which would create additional oxygen vacancies on the surface of the Ce4+-O2 fluorite-type oxide.

    The identification of Au ions (Fig. 3A), along with the increased amount of surface oxygen in the leached sample (Fig. 4), argues in favor of lattice substitution. Diffusion of Au ions into ceria must take place during the heating step in the preparation process, because attempts to leach the Au immediately after deposition and before heating failed to produce an active catalyst. The minimum metal loading required for a desired WGS activity may be determined from the ceria surface properties. Assuming uniform, monolayer, dimensioned metal surface coverage on the CL material [Ce(10 atom % La)Ox, 160 m2/g], we calculated the coverage to be 13.5 atom % Au or 15.5 atom % Pt, with the Au or Pt radius equal to 0.174 nm and 0.139 nm, respectively. Only a small fraction of a monolayer of Au or Pt was present on the leached catalysts (Table 1 and Figs. 1 and 2), but this amount correlates well with the concentration of surface oxygen defect sites of ceria (2527).

    The importance of the surface defects of ceria as the anchoring sites of Au, and in turn as the active sites for the WGS reaction, can be seen in ceria samples annealed at high temperatures, which effectively reduces the number density of these sites. The reaction rate measured over 3.4 atom % Au-CeO2 (calcined at 800°C for 4 hours) (Table 1) was very low, but the apparent activation energy was the same as for the other Au-ceria (DP) materials shown in Fig. 1. Removal of Au from this sample by leaching was essentially complete (Table 1) and the leached sample was inactive for the WGS reaction up to 400°C.

    Au nanoparticles are essential for some oxidation reactions, in view of their totally different oxygen adsorption properties compared with those of bulk Au (28, 29). However, they are unimportant in the WGS reaction over Au-ceria.

    Supporting Online Material

    Figs. S1 and S2

    References and Notes

    View Abstract

    Stay Connected to Science

    Navigate This Article