Report

Coordinatively Unsaturated Al3+ Centers as Binding Sites for Active Catalyst Phases of Platinum on γ-Al2O3

See allHide authors and affiliations

Science  25 Sep 2009:
Vol. 325, Issue 5948, pp. 1670-1673
DOI: 10.1126/science.1176745

Bonding Oxides and Metals

The binding of noble metals that can act as catalysts to metal oxides that are reducible is assumed to occur at the exposed cation of the oxide. For nonreducable oxides such as aluminum oxide, it is not so obvious how the metal can bind strongly. Kwak et al. (p. 1670) used a combination of high-resolution transmission electron microscopy and solid-state magic-angle spinning nuclear magnetic resonance to study the anchoring of platinum at high and low loadings on alumina. At the surface, the Al3+ ions were penta-coordinated. Density functional calculations support a model in which the cation binds three oxygen atoms in the alumina and two from platinum oxide.

Abstract

In many heterogeneous catalysts, the interaction of metal particles with their oxide support can alter the electronic properties of the metal and can play a critical role in determining particle morphology and maintaining dispersion. We used a combination of ultrahigh magnetic field, solid-state magic-angle spinning nuclear magnetic resonance spectroscopy, and high-angle annular dark-field scanning transmission electron microscopy coupled with density functional theory calculations to reveal the nature of anchoring sites of a catalytically active phase of platinum on the surface of a γ-Al2O3 catalyst support material. The results obtained show that coordinatively unsaturated pentacoordinate Al3+ (Al3+penta) centers present on the (100) facets of the γ-Al2O3 surface are anchoring Pt. At low loadings, the active catalytic phase is atomically dispersed on the support surface (Pt/Al3+penta = 1), whereas two-dimensional Pt rafts form at higher coverages.

The ability to control the dispersion and morphology (typical characteristics that determine the performance of catalysts) of oxide-supported metal catalysts is a primary goal of catalyst design and can be enabled by understanding the nature of metal–support surface interactions. Precious metals (e.g., Pt, Pd, and Rh) supported on oxide surfaces are the most widely used industrial catalyst materials. For these classes of catalysts, dispersion of the precious metal on the oxide support is an especially critical factor because of the expense of the metal.

The so-called strong metal-support interaction (SMSI) is often seen as critical to sustaining high catalytic activity under demanding catalyst operation conditions (i.e., high temperature, high water vapor pressure, etc.). Indeed, SMSI has been directly linked to the presence of electronic defects that can be prepared with ease on the surfaces of reducible oxides (e.g., CeO2 and TiO2) (1). Anchoring the active metal components to these electronic defects has been documented (2, 3), with the strong interaction between the active metal and defects of reducible oxides fundamentally determining the dispersion, morphology, and, therefore, the catalytic activity of metal clusters. For example, the correlation between the number of oxygen defects on the TiO2 support and the dispersion and morphology of nanosized Au particles has been clearly demonstrated by the results of density functional theory (DFT) calculations of Lopez et al. (2). Experimentally, Chen and Goodman (3) have described the formation of stable, two-dimensional (2D) Au clusters on a defect-rich TiO2 thin film.

Electronic defects, however, are not present on γ-Al2O3, a nonreducible oxide that is one of the most commonly used catalyst support materials in practical applications. Still, high dispersion and thermal stability of the active metal phase can be achieved and maintained on γ-Al2O3 supports, with Koningsberger and co-workers (4) suggesting that specific interactions between “defect sites” of γ-Al2O3 and Pt clusters play an essential role. Their extended x-ray absorption fine structure (EXAFS) analysis revealed changes in the morphology of γ-Al2O3–supported Pt particles upon annealing at 723 K. In particular, 3D Pt particles were converted into 2D rafts during the annealing process. Additional evidence for the presence of special (so-called “defect”) sites for anchoring precious metals comes from the presence of monoatomically dispersed Pt on γ-Al2O3 in high-angle annular dark-field scanning transmission electron microscopy (HA-ADF STEM) images reported by Pennycock and co-workers (5, 6). However, in spite of considerable efforts to understand the interaction between catalytically active metal particles and γ-Al2O3 support surfaces, the exact nature of the defect sites where these metals can anchor and be stabilized has not been determined (4, 79). We present data obtained from ultrahigh magnetic field 27Al magic-angle spinning (MAS) nuclear magnetic resonance (NMR) spectroscopy and HA-ADF STEM techniques that suggest that coordinatively unsaturated Al3+ centers [i.e., pentacoordinate Al3+ (Al3+penta)] on the γ-Al2O3 surface are the anchoring sites for PtO, the precursor of metallic Pt. We used DFT to confirm the experimental observations for the feasibility of the formation of 2D PtO rafts interacting with Al3+penta surface sites on the γ-Al2O3 support surface.

High-resolution 27Al solid state MAS-NMR spectra collected at ultrahigh magnetic fields for the γ-Al2O3 support and 10 weight percent (wt %) Pt/γ-Al2O3 samples after annealing at 573 K are displayed in Fig. 1A. The spectrum of the metal-free oxide support exhibits three peaks centered at 13, 35, and 70 parts per million (ppm) chemical shifts [referenced to 1 M aqueous Al(NO3)3]. The two characteristic 27Al NMR features of γ-Al2O3 at 13 and 70 ppm represent Al3+ ions in octahedral (Al3+octa) and tetrahedral (Al3+tetra) coordination, respectively. The NMR peak at 35 ppm chemical shift has been assigned to Al3+ ions in pentahedral coordination (Al3+penta) (10, 11). These pentacoordinate sites are created on the γ-Al2O3 surface by dehydration and dehydroxylation at elevated temperatures (here at 573 K) (10, 11). Loading Pt onto the γ-Al2O3 support and calcining the catalyst material at 573 K result in a substantial decrease in the number of Al3+penta sites, as evidenced by the large drop in the intensity of the 35-ppm NMR peak. Concomitantly, the intensity of the 13-ppm NMR feature increases with the loading of Pt onto γ-Al2O3, which suggests the conversion of the Al3+penta sites into Al3+octa (coordinative saturation; the total number of Al3+ ions remains constant). These results strongly suggest that Pt atoms bind to the Al3+penta sites on the γ-Al2O3 surface through oxygen bridges, thereby coordinatively saturating these sites (penta- to octahedral conversion).

Fig. 1

(A) The 27Al MAS-NMR spectra of γ-Al2O3 (black) and 10 wt % Pt/γ-Al2O3 (red) (both samples were calcined at 573 K before NMR measurements). (B) Number of Al3+penta sites as a function of calcination temperature in γ-Al2O3 (black) and 10 wt % Pt/γ-Al2O3 (blue) samples. The number of Pt per Al3+penta site as a function of calcination temperature is displayed in red. (C) The number of Pt atoms per Al3+penta sites as a function of Pt loading on γ-Al2O3 support. (Samples were calcined at 573 K before NMR measurements.)

Quantitative evaluation of the number of Al3+penta sites (exclusively present on the surface of the oxide support) converted to Al3+octa sites by the addition of Pt is made possible by using the NMR peak areas in Fig. 1A and the results of our previous work on BaO/γ-Al2O3 systems (11, 12). The number of Al3+penta sites on the metal-free, 573 K–calcined γ-Al2O3 sample has been estimated to be 1.8 × 10−4 mole/g (11), whereas the number of Pt-free Al3+penta surface sites still present after 10 wt % Pt loading is about 0.7 × 10−4 mole/g. With the known total amount of Pt in the catalyst, the difference between the number of Al3+penta sites before and after Pt loading can be used to estimate the average number of Pt atoms per reacted (i.e., converted to Al3+octa) pentacoordinate surface Al3+ ions. Assuming an even distribution of Pt on the γ-Al2O3 surface, on average, ~one of every four Pt atoms is anchored to each of the occupied Al3+penta sites. HA-ADF STEM images obtained from this sample (discussed later) show the formation of 2D raftlike Pt structures. Thus, the strong interactions between some Pt atoms in these clusters and the Al3+penta sites on the γ-Al2O3 surface are likely responsible for the formation of 2D Pt rafts.

All of these results suggest that Pt is selectively anchored to the pentacoordinate aluminum ions present on the surface of mildly calcined γ-Al2O3. Thus, we propose that the coordinative saturation of these pentacoordinate aluminum ion sites is the driving force for the strong catalytic active phase–support interaction. This structure represents a fundamentally different anchoring of the active catalyst component onto the oxide support from the SMSI observed in metal–reducible support oxide catalyst systems. For γ-Al2O3, the driving force is the coordinative saturation of specific sites on the support oxide surface (i.e., Al3+penta), whereas in reducible oxides it is an electronic interaction between the metal and the electronic defects on/in the oxide (2). The interaction between Pt and γ-Al2O3 surface described here is similar to our recent discussion of specific interactions between oxides (e.g., BaO and La2O3) and Al3+penta sites on γ-Al2O3 (1113). In this work, such interactions were shown to be responsible for the thermal stabilization of the γ-Al2O3 structure. Pentacoordinate Al3+ ions formed on the (100) facets of γ-Al2O3 upon high-temperature calcination interacted selectively with these oxides, forming highly (even atomically) dispersed nanostructures (12).

The number of Al3+penta sites increases with increasing annealing temperature. For example, Fig. 1B shows the number of Al3+penta sites (determined from NMR spectra) as a function of calcination temperature for γ-Al2O3 and 10 wt % Pt/γ-Al2O3. As the temperature was raised from 573 K to 873 K, the number of Al3+penta sites increased from 1.8 × 10−4 to 2.8 × 10−4 mole/g for the metal-free γ-Al2O3 support, whereas it changed from 0.7 × 10−4 to 2.1 × 10−4 mole/g for the 10 wt % Pt/γ-Al2O3 catalyst. The results presented in Fig. 1B reveal that, although the number of Al3+penta sites increased on both metal-free and metal-loaded γ-Al2O3 support, the difference in the number of Al3+penta centers between the clean and metal-loaded support decreased substantially as the calcination temperature increased. As described above, we can estimate the average number of Pt atoms per occupied Al3+penta [denoted in the following as Pt/Al3+penta(occupied by Pt)] as a function of calcination temperature (red symbols in Fig. 1B) by using the experimental values of occupied Al3+penta sites and the known quantity of Pt in the catalyst. At 573 K, the number of Pt/Al3+penta(occupied by Pt) is ~four (as we have discussed previously), and this number is unchanged as the calcination temperature is raised to 673 K. As the sample is calcined to even higher temperatures (773 and 873 K), the number of Pt/Al3+penta(occupied by Pt) approximately doubled. The observed stepwise increase above 673 K calcination temperature can be explained by the sintering of metallic Pt particles that begin to form at around 723 K even under oxidizing conditions.

Given evidence for a specific and strong interaction between Pt and the Al3+penta sites, can we control the number of Pt/Al3+penta(occupied by Pt) during catalyst preparation? This parameter ultimately determines the dispersion and likely also the morphology of the active metal phase supported on γ-Al2O3. Theoretical studies of the structure of γ-Al2O3 have predicted that the (100) crystal facets (representing about 17% of the total surface area) readily dehydrate and dehydroxylate under conditions of calcination applied throughout this study, producing the stable Al3+penta sites (14, 15). By keeping the number of Al3+penta sites constant (calcining the samples to the same temperature of 573 K) and varying the loading of Pt, we can prepare catalysts with a range of Pt/Al3+penta(occupied by Pt) values. To this end, we prepared a series of Pt/γ-Al2O3 catalysts with a range of Pt loading of 1 to 10 wt %. After calcining the samples under identical conditions (573 K), the number of Al3+penta ions was estimated for each Pt-loaded sample by ultrahigh magnetic field 27Al MAS-NMR. These measurements again allowed us to quantitatively determine the average number of Pt/Al3+penta(occupied by Pt), and the results are displayed in Fig. 1C. Interestingly, a practically linear relation was found between the Pt/Al3+penta(occupied by Pt) and the Pt loading in the coverage range of this study.

The results presented in Fig. 1C suggest that, at loadings of 1 wt % Pt or lower, each occupied Al3+penta site interacts with exactly one Pt atom, meaning that all Pt must be atomically dispersed. This result [Pt/Al3+penta(occupied by Pt) = 1] is similar to our observations for a 2 wt % BaO/γ-Al2O3 sample calcined at 773 K (12). In the latter case, HA-ADF STEM images revealed the presence of mostly single BaO units dispersed on certain, most likely (100), facets of the γ-Al2O3 support. In order to substantiate our claim of single-atom Pt dispersion at loadings ≤ 1 wt %, we collected HA-ADF STEM images for the 1 wt % Pt/γ-Al2O3 sample calcined at 573 K. The HA-ADF STEM image shown in Fig. 2A demonstrates the almost-perfect atomic dispersion of Pt on the γ-Al2O3 surface at 1 wt % loading. Atomic dispersion of Pt on γ-Al2O3 at low loadings has been observed previously in a STEM study (5, 6); however, the nature of the sites where Pt was anchored to the support surface had not been elucidated.

Fig. 2

High-resolution STEM images of 1 wt % Pt/γ-Al2O3 (A) and 10 wt % Pt/γ-Al2O3 (B to D) samples. The image in (A) shows the presence of mostly atomically dispersed Pt. (C) The presence of atomically dispersed Pt. (D) Pt rafts that dominate the STEM image at high Pt loadings. (E and F) Normalized image intensities in directions 1 and 2 from the 2D raft shown in (D).

At higher Pt loadings, the formation of 2D Pt rafts has been proposed on the basis of the results of Pt EXAFS measurements (4). The NMR results presented here also suggest the formation of PtO clusters on the Al3+penta sites, consisting on average of four Pt atoms per anchoring site at 10 wt % loading after calcination at 573 K. HA-ADF STEM images, obtained after calcination of a 10 wt % Pt-loaded sample at 573 K (Fig. 2, B to D), substantiate the formation of 2D Pt rafts on the γ-Al2O3 surface with 1-nm average size. The high-resolution image in Fig. 2C also reveals that a substantial fraction of atomically dispersed Pt is still present on the γ-Al2O3 surface, presumably bound to Al3+penta surface sites as evidenced by the NMR and HA-ADF STEM results obtained at lower Pt loadings. The 2D rafts (a good representative close-up image is shown in Fig. 2D) with 1-nm average size consist of about 20 Pt atoms, suggesting that these rafts may interact with as many as four to five Al3+penta sites in order to maintain the average of ~4 Pt/Al3+penta(occupied by Pt) ratio. From cross sections of one of the images (Fig. 2, E and F), the distance between two Pt atoms is estimated to be 2.78 Å, shorter than the expected Pt-Pt distances in PtO (3.1 Å) and almost identical to the bond distance in metallic Pt (2.76 Å). However, the results of x-ray absorption near-edge structure analysis substantiate that the rafts observed after 573 K calcination are PtO (fig. S1) (16).

Analysis of a number of cross sections on different particles also supports the assignment of these PtO particles to 2D rafts instead of 3D clusters. Furthermore, DFT calculations (discussed below) suggest a Pt-Pt distance of 2.8 Å in 2D PtO overlayers on γ-Al2O3(100) facets. On the basis of DFT calculations, Lopez et al. (2) recently proposed the formation of 2D metallic gold rafts over defect sites of a TiO2(110) surface. The formation of 2D Au rafts on the defect sites of a TiO2 film was experimentally confirmed by Chen and Goodman (3). The strong interaction between Au and electronic defects (oxygen vacancies) on TiO2 results in the formation of 2D metallic particles. In our system, the strong interaction between PtO and the coordinatively unsaturated Al3+penta sites on the γ-Al2O3 surface is responsible for the formation of PtO rafts (4, 17).

In order to further substantiate the role of Al3+penta sites in determining Pt dispersion and, thus, morphology on alumina supports, we compared the above results obtained on γ-Al2O3 with high-resolution STEM (HR-STEM) images for a 0.25 wt % Pt/α-Al2O3 catalyst calcined at 573 K. The low Pt loading on the low-surface-area α-Al2O3 support was chosen to match the Pt/support surface area ratio of the 10 wt % Pt/γ-Al2O3 sample. A representative TEM image (Fig. 3) shows the formation of very large 3D Pt particles (10 to 15 nm) on the α-Al2O3 support, even after the relatively low–temperature (573 K) calcination. The formation of these large Pt clusters on the α-Al2O3 support indicates a very weak interaction between PtO and the support. The particles can readily migrate even at this relatively low calcination temperature because the α-Al2O3 surface does not possess any Al3+penta sites (18) that would provide anchoring points for PtO to the surface. Another consequence of the weak interaction between PtO and α-Al2O3 is the formation of metallic Pt upon calcination at temperatures as low as 573 K. Reduction temperatures increase with increasing catalyst precursor (here PtO)–support oxide interaction in supported Pt catalysts (17).

Fig. 3

TEM image of a 0.25 wt % Pt/α-Al2O3 sample obtained after calcination at 573 K.

We performed DFT calculations (16) in order to obtain specific information about (i) the energetics of the interaction of PtO particles with the (100) facets of γ-Al2O3 and (ii) the likely stable structures of the experimentally (HA-ADF STEM) observed 2D PtO rafts. The surface model for γ-Al2O3 was taken from the work of Digne et al. (14, 15) that showed the formation of pentacoordinated Al3+ ions on the (100) facets upon termination of the bulk structure. The γ-Al2O3(100)-2×1 surface model used in this work consists of eight Al2O3 units with all surface Al3+ ions being pentacoordinated and the O2- ions tricoordinated. Our calculations, in agreement with those reported previously (14, 15), indicate that the γ-Al2O3(100) surface was fully dehydrated under the experimental conditions of this study (i.e., 573-K calcination). Various structures extracted from one layer of an optimized tetragonal PtO bulk on the γ-Al2O3(100) surface were examined. The O-terminated PtO(101) structure on the γ-Al2O3(100) surface was the thermodynamically most stable structure. Interestingly, this PtO(101) overlayer structure was obtained from the relaxation of an O-terminated PtO(100) structure on the γ-Al2O3(100) surface. The relaxation can be attributed to the thermodynamic instability of the PtO(100) surface, which has a negative surface energy (19). The 2D PtO(101) overlayer (Fig. 4) strongly interacts with the γ-Al2O3(100) surface via four O-Al3+penta bonds, with a binding energy of 7.1 eV. The ratio of PtO/Al3+penta(occupied by Pt) is three, similar to our experimental estimation of about four (although it should be noted that the experimental number is affected by the presence of atomically dispersed Pt evident in the HA-ADF STEM images). The Pt-Pt bond lengths in the 2D PtO overlayer structure after relaxation are calculated to be between 2.7 and 2.9 Å, in excellent agreement with the 2.78 Å estimated from the HR-STEM images.

Fig. 4

Optimized PtO overlayer structures on the γ-Al2O3(100) surface. (A) Top view. (B and C) Side views.

The work presented here provides a new understanding about the nature of strong interactions between catalytic phases and oxide support materials in the industrially important catalyst system, Pt/γ-Al2O3. Our explanation for the role of pentacoordinate Al3+ ions in the anchoring of the catalytically active phase (Pt) to the γ-Al2O3 support is arrived at on the basis of the results of ultrahigh magnetic field 27Al MAS-NMR measurements. At low (≤1 wt %) Pt loadings, Pt is atomically dispersed on the support surface (Pt/Al3+penta = 1). When the loading of Pt exceeds the number of Al3+penta sites, 2D PtO rafts form, as evidenced by HA-ADF STEM measurements. DFT calculations provide further confirmation for the energetic feasibility of the formation of these 2D rafts as well as for their energetically most stable overlayer structure. (Although data presented here are consistent with the mechanism of Pt anchoring to the Al3+penta sites of γ-Al2O3, other mechanisms, including the reaction between the Al3+penta sites and the ligands of the Pt precursor, cannot be completely ruled out.) A substantial challenge remains to develop synthesis methods that allow for the systematic and controllable variation of the number of Al3+penta sites, ultimately enabling us to tailor the dispersion and morphology and, therefore, catalytic activity of active metals on the commonly used γ-Al2O3 support material.

Supporting Online Material

www.sciencemag.org/cgi/content/full/325/5948/1670/DC1

Materials and Methods

SOM Text

Fig. S1

References

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We gratefully acknowledge the U.S. Department of Energy (DOE), Office of Science, Basic Energy Sciences, Division of Chemical Sciences, and Office of Energy Efficiency and Renewable Energy, Vehicle Technologies Program, for the support of this work. The research described in this paper was primarily carried out in the Environmental Molecular Sciences Laboratory (EMSL), a national scientific user facility sponsored by the DOE’s Office of Biological and Environmental Research and located at Pacific Northwest National Laboratory (PNNL). PNNL is operated for the DOE by Battelle Memorial Institute under contract number DE-AC05-76RL01830. Computing time was granted by National Energy Research Scientific Center (NERSC). The HR-STEM images were recorded at Oak Ridge National Laboratory's High Temperature Materials Laboratory, sponsored by the DOE, Office of Energy Efficiency and Renewable Energy, Vehicle Technologies Program.
View Abstract

Navigate This Article