Investigation of the Structure of Platinum Clusters Supported in Zeolite

Results from the Pair Distribution Function Method. The experimental PDF curves for zeolite Beta, Pt-Beta sample, and their difference are shown in Fi...
0 downloads 0 Views 398KB Size
J. Phys. Chem. C 2007, 111, 8573-8579

8573

Investigation of the Structure of Platinum Clusters Supported in Zeolite Beta Using the Pair Distribution Function Marı´a M. Martı´nez-In˜ esta† and Rau´ l F. Lobo* Chemical Engineering Department, UniVersity of Delaware, 150 Academy Street, Newark, Delaware 19716 ReceiVed: March 20, 2006; In Final Form: January 29, 2007

Structural characterization of supported metal catalysts is a challenge because the metal clusters are small, usually polydisperse, and present in low loadings. Here, we describe the use of the pair distribution function (PDF) method to study the structure of platinum clusters supported in zeolite beta at a loading of 1.1% w/w. In addition to the characterization by the PDF method, the sample was characterized using X-ray diffraction, X-ray photoelectron spectroscopy, X-ray absorption, and dark-field scanning transmission electron microscopy. With these techniques, we were able to determine that the platinum clusters were in a reduced state, monodisperse, and located mainly in the pores. With the PDF technique, we were able to determine that the local structure of the clusters was similar to that of the bulk platinum. Contrary to what might be expected, the PDF results showed that the lattice parameters did not change significantly. However, attempts to model the PDF using a finite cluster with the bulk platinum structure did not yield quantitative agreement to experiment, and it was determined that there were other interactions besides that of Pt-Pt affecting the PDF. Models that take into consideration different possible cluster shapes and sizes are necessary to completely describe the PDF at these low loadings. However, the PDF technique was shown to be valuable for the experimental determination of the structure of supported nanoclusters.

1. Introduction Metals are among the most important catalysts being used on a large scale for the refining of petroleum, pollution abatement of automobile exhaust, hydrogenation of carbon monoxide, hydrogenation of fats, and many other processes. The metal, however, is often expensive, and it is generally dispersed on a high-surface area porous support where it constitutes about 1 wt % of the catalytic material.1-3 Desired dispersed metal particles are less than 10 Å in size and are considered to have a dispersion (fraction of exposed metal atoms) of 1, ensuring a maximum metal activity. A general review by Gates on supported metal clusters is available.4 The fact that the metals are dispersed on supports, however, poses a challenge to their characterization because the metal clusters are small, usually polydisperse, and present in low loadings. The structures are often modeled as crystallites having the symmetry of bulk metals.5 For example, a 20 atom platinum particle assuming bulk structure structure has an edge length of 10 Å and a dispersion of 1. Zeolites as supports can be key for the study of the structure of the dispersed metals as metals dispersed in zeolites are usually close to monodisperse with an average size around 10 Å, which is approximately the size of the zeolite pore. In practice, metal clusters supported on zeolites have found industrial application for selective reforming of naphta to give aromatics Pt/Zeolite L (LTL).4 There are only a few experimental techniques capable of studying the local structural properties of zeolite-supported metal catalysts.6-8 The most commonly used techniques are edge X-ray absorption fine structure (EXAFS), high-resolution * To whom correspondence should be addressed. E-mail: (R.F.L.) [email protected]. † Permanent Address: Chemical Engineering Department, University of Puerto Rico, P.O. Box 9046, Mayagu¨ez PR 00681-9046.

transmission electron microscopy (HRTEM), and chemisorption. With EXAFS, element-specific first coordination shells can be studied to determine the interatomic distances and metal coordination numbers. This last number can be used to determine the size of the clusters. HRTEM, but more specifically Z-contrast images of the metal clusters, can be obtained with a scanning transmission electron microscopy (STEM) high-angle annular dark-field detector (HAAD) that can be used to determine their distribution throughout the matrix and their polydispersity. These images, however, cannot be used to directly determine the cluster sizes due to electron-scattering effects. Chemisorption, on the other hand, relates the amount of reductant chemisorbed to the size of the cluster assuming metal bulk properties. It is evident that the information available from these techniques is still limited and imprecise. Another important aspect that affects the structure of the clusters is the oxidation state of the metal as the effective radius decreases with increased oxidation state. Platinum, for example, has a metallic radius of 1.39 Å and thus a Pt-Pt distance of 2.78 Å, while the ionic radius of the 2+ cation is 0.94 Å. The oxidation state is also critical for the catalytic applications of the metal. X-ray photoelectron spectroscopy (XPS) and X-ray absorption near-edge absorption spectroscopy (XANES) are two techniques that can probe the oxidation state of the metal atoms. The first technique yields the binding energy of the electrons in the sample using a monochromatic X-ray source and is used mainly for surface metals as it has limited depth penetration. Because the binding energy is element specific, XPS can be used to identify the elements present in the sample, while any shift in the binding energy from the known values of the neutral species is used to determine its oxidation state. In XANES, an X-ray source with tunable energy is used to determine the X-ray energy absorption edge, which is equal to

10.1021/jp061728q CCC: $37.00 © 2007 American Chemical Society Published on Web 05/27/2007

8574 J. Phys. Chem. C, Vol. 111, No. 24, 2007

Martı´nez-In˜esta and Lobo

the binding energy of the electrons. As with XPS, a shift in the absorption edge is indicative of a charge. However, contrary to XPS, this technique probes all the atoms in the sample including those located in the bulk. XANES spectra also contain information of the oxidation state in its functionality that can be easily identified by comparison with reference samples. Despite the limitations of the abovementioned techniques, useful insights have been obtained for the system of platinum supported in zeolites. Previous EXAFS studies have determined that the Pt-Pt first-neighbor distance can vary from the bulk value of 2.78 Å, when the EXAFS experiment is done under a hydrogen atmosphere, to 2.70 Å, when the sample is under helium.9 Another EXAFS study by Vaarkamp et al.10 was able to resolve Pt-O interactions of the platinum atoms in the clusters and the oxygen from the zeolite framework and has determined that their interatomic distance can vary from 2.77 to 2.24 Å with increasing reduction temperature. It was concluded that the larger Pt-O distances were probably caused by spillover hydrogen (from the reduction) located at the interface of the Pt and the oxygen (Pt-H-O). Here, we investigate platinum clusters supported in zeolite Beta using the pair distribution function (PDF) method. The pair distribution function G(r) has been thoroughly described by Egami and Billinge11 and has been used before by the authors to describe structural disorder in zeolites.12,13 It describes the distribution of interatomic distances in a material and can be obtained either from high-energy X-ray or neutron diffraction data. The relation is described as

G(r) ) 4πr(F(r) - Fo) )

2 π

∫QQ

max

min

Q[S(Q) - 1]sin(Qr) dQ (1)

where r is the interatomic distance, F(r) is the microscopic pair density, Fo is the average atomic density of the material, Q ) 4πsin(θ)/λ is the scattering vector, and S(Q) is the corrected and normalized scattering data. The integration must be done to high Q to minimize termination errors that come from cutting the integration short, thus the need for a high-energy radiation source. The experimental PDF can be calculated from a structural model by using the equation

G(r) )

1

∑ r i,j

bibj 〈b〉2

δ(r - rij) - 4πrFo

diffraction data is model-independent, while G(r) obtained from EXAFS is calculated on the basis of a molecular model. EXAFS also gives information that is limited to a few coordination spheres, while the PDF method gives information up to much higher interatomic distances, depending on the quality of the measurement. Another difference is the robustness of the determination of the experimental data. In the PDF method, if you know the approximate composition of the material all the corrections are tabulated and straightforward, while with EXAFS there is usually a phase shift in the experimental data that needs to be corrected with a structural model. The results from EXAFS are, therefore, model dependent while the PDF results are not. On the other hand, the PDF method, in general, is not element sensitive as it gives the interatomic distance distribution of all the atoms. It is possible to do anomalous scattering experiments to obtain an element specific PDF, yet this method requires very precise data and, thus, is more difficult to obtain.14 We found a previous report that used the mathematically related radial distribution function to study the structure of platinum clusters in zeolite Y.15 There are, however, some differences in our approaches. The main difference is that the amount of platinum in their samples is 12 wt % Pt and as a consequence the radial distribution function shows mainly PtPt peaks and does not show the zeolite-zeolite nor the zeolitePt interatomic distances. Also, the high platinum content increases the probability that the platinum is located outside the zeolite pores. A technical difference is that their experiments were conducted in an in-house diffractometer using a filtered Mo KR radiation, and ours were done at a synchrotron facility. The result is that the resolution for our experiment is inherently better. Finally, here we are modeling the platinum clusters based on the pair distribution function observations, and Gazellot et al. just make qualitative observations from experimental radial distribution functions. In this article, we build on a previous study where we studied the framework of zeolite Beta using the PDF method,12 and we concentrate on modeling the structure of the platinum clusters. In practice, zeolite Beta has potential commercial value as metal support as, in its cation-exchanged version (nonacidic), it has been found to lead to better dispersion of the platinum and is more stable to sulfur poisoning than the commercially used zeolite LTL.16

(2)

where the sum is over all atoms in the sample, rij is the distance between atoms i and j, bi is the scattering length of atom i and 〈b〉 is the average scattering length of the sample. The scattering lengths using X-rays are the form factors and are usually assumed constant at a specific value of Q. Systematic comparison between the G(r) obtained from the experimental data and the G(r) obtained from structural models provides a mechanism to analyze disorder present in the sample. This last equation also serves to illustrate the relative advantages of X-ray and neutron radiation for PDF studies. The key is to use the radiation with the most contrast (or difference in scattering lengths) between the nuclei directly related to the disorder. Here, we are interested in understanding the Pt-X interactions where X can be atoms in the zeolite system or other platinum atoms, and we chose an X-ray source to obtain the best contrast between the Pt atoms and the zeolite atoms. This method can be thought of as being similar to EXAFS but there are some essential differences. To start, the experimental PDF from

2. Experimental Methods 2.1. Synthesis of Pt-Beta. For our experiments, we used an Al-Beta sample obtained from Zeolyst (CP814E) with a SiO2/ Al2O3 ) 25, [NH4][Al0.074Si0.926O2]. This sample was ionexchanged in an aqueous 0.2 M NaBr solution overnight at 89 °C. The sample was then washed and filtered with enough deionized water to ascertain that there was no Br- left in solution. This was checked using a silver nitrate solution. This zeolite was our precursor Beta sample. Our platinum-loaded experimental sample was obtained by incipient wetness impregnation of the zeolite precursor with a PtII(NH3)4(NO3)2 aqueous solution to get a platinum loading of 1.1 wt %. The amount of water to wet the sample was first determined empirically (∼0.3 g/g zeolite), and the concentration of the PtII(NH3)4(NO3)2 aqueous solution was determined accordingly. After impregnation, the samples were dried at room temperature in air. The loading was confirmed by atomic absorption spectroscopy. The impregnated sample was then calcined in air by ramping the temperature in a furnace at a rate of 5 °C/min to 150 °C, holding at that temperature for 1 h,

Platinum Clusters Supported in Zeolite Beta and then ramping at a rate of 1 °C/min to 350 °C and holding for 8 h. The sample was then reduced with hydrogen in a flow reactor using a ramp rate of 5 °C/min to 450 °C and holding for 2 h. These conditions were chosen to optimize platinum dispersion based on results obtained with zeolite KL.17 This sample is from here on referred to as the Pt-Beta sample. Three more samples containing 0.7, 1.5, and 3.0 wt % Pt were also synthesized using the method outlined above and are referred to as the reference samples. 2.2. Experimental PDF. The data for the X-ray PDF was obtained at the SRI-CAT 6-ID-D beamline at the Advanced Photon Source, Argonne National Laboratory. Data was taken for the sample holder, the precursor zeolite Beta and the PtBeta sample. The measurements were conducted in symmetric transmission geometry at room temperature. A Si (311) crystal monochromator was used to obtain incoming X-rays with 76.0 KeV energy (λ ) 0.163 Å) below the Pt absorption edge. The diffracted photons were collected with an intrinsic germanium detector connected to a multichannel analyzer. Several diffraction runs were conducted with the sample at room temperature, and the intensities were averaged to get better statistics. Each run was divided in two sets: from 2θ ) 0.88° the scan was measured with a step size of 0.01° with 5 s per step and from 8-60° the scattering was measured with a 0.02° step size for 5 s per step. The averaged data was normalized for flux, corrected for detector dead time, background, Compton scattering, and for absorption, and then was converted to total scattering structure factor, S(Q). The structure factor was used up to Q ) 11 Å-1 to minimize the Fourier transformation of statistical noise in the data at high Q into real space. The data treatment was done with the program PDFgetX.18 We attempted to measure anomalous scattering experiments for which two datasets were taken, one well below the absorption edge of platinum (76.0 KeV) and another almost at the absorption edge (78.35 KeV)), to get the scattering data just from the platinum, but the crystal monochromators in the beamline were mosaic and not perfect crystals that resulted in a small distribution of incoming energies that was inappropriate for this type of experiment. An X-ray absorption experiment was also carried out at Argonne National Laboratory on the Pt-Beta sample. Two scintillator detectors were located before and after the sample, and a sweep of incoming X-ray energies was done around the Pt-K edge to determine the sample absorption energy spectrum. In-house X-ray powder diffraction (XRD) patterns were obtained for the original and reference samples on a Philips X’Pert diffractometer using Cu KR radiation. A Si standard was added as a reference for peak positions. The patterns were collected from 1-50° in 0.02° steps with 1.75 s per step. An X-ray photoelectron spectroscopy (XPS) spectrum was obtained from the experimental sample with an instrument equipped with an Al KR X-ray source. Scanning electron microscopy (SEM) images were obtained in a JSM 3400F instrument of the Pt-Beta sample and Z-contrast STEM images of the platinum clusters from the same sample were obtained with a HAAD in a JEM-2010F microscope at 200 KeV using a 0.2 nm probe. 2.3. Real-space Rietveld Refinement. The structure of the zeolite Beta precursor was refined using the program PDFfit19 following the procedure described before for the refinement of Si-beta12 and zeolite chabazite.13 The structure of the platinum clusters was also refined employing this procedure using the difference PDF obtained by substracting the zeolite PDF from the Pt-zeolite PDF.

J. Phys. Chem. C, Vol. 111, No. 24, 2007 8575

Figure 1. Powder XRD patterns of the experimental and reference samples. The reference samples were mixed with a silicon standard to monitor changes in the lattice parameters with increasing platinum content. The inset shows the Pt 〈111〉 peak in the samples with a platinum loading higher than 0.7 wt %.

2.4. Reverse Monte Carlo (RMC). We used RMC simulations to model the platinum particles inside the zeolite pores using the program DISCUS.20 In these simulations, atoms were randomly selected and they were shifted by a random amount in a random direction. To avoid structures that are either physically or chemically improbable, we set a minimum distance for the Pt-Pt atomic distances. The actual shift applied to the atom is the generated shift multiplied by 0.05 of each unit cell parameter. We also tried displacement sizes of a fraction of 0.005 and 0.01 of the unit cell parameter; however, those simulations did not improve significantly the agreement of the structures. 3. Results and Discussion 3.1. General Characterization of the Platinum Clusters. The XRD pattern of the experimental Pt-Beta sample (Figure 1) shows an extra peak (compared to Na-exchanged zeolite Beta) at around 39.9° 2θ. The position of this peak is close to the position of the Pt 〈111〉 diffraction peak (39.8° 2θ). This peak also appears at approximately the same position on the XRD patterns of the reference samples with 1.5 and 3.0 wt % Pt. The Pt 〈111〉 peak does not appear in the 0.7 wt % Pt sample XRD. The shift of the peak to a higher angle might indicate a reduction in the cubic lattice parameter of the sample from a bulk value. However, definite conclusions cannot be drawn as the Pt peaks are very diffuse, and thus the determination of the peak positions is not precise. The Pt 〈111〉 peaks increase in intensity and decrease in width with increasing platinum concentration. An analysis of the peaks using the Scherrer equation indicates that the platinum cluster size in our experimental sample increases with platinum concentration. However the actual particle size values, especially of our experimental sample, are not reliable as the peaks were very diffuse and outside the reliability range of this method. In the 0.7 wt % Pt, the clusters are probably too small to be detected with this method. These XRD patterns also show that the zeolite framework does not appear to be under strain as is apparent with the constant zeolite peak positions with increasing platinum loading. The peaks at 28.56° 2θ and at 47.32° 2θ correspond to the 〈111〉 and the 〈220〉 diffraction peaks from a silicon standard that was mixed with the samples to use as a peak position reference. The peaks in the XPS spectra (Figure 2) were identified by comparison to the reference spectra of the different elements.21 The most intense peaks correspond to oxygen as is expected from the concentration of elements in the sample. The silicon peaks can also be identified at 153 eV (2s) and 102 eV (2p). The presence of these peaks assures that the sample is being

8576 J. Phys. Chem. C, Vol. 111, No. 24, 2007

Martı´nez-In˜esta and Lobo

Figure 4. SEM image of experimental Pt-Beta sample.

Figure 2. XPS spectrum of the experimental sample (a) complete spectrum and (b) range highlighting the area where the Pt-binding energy peak should appear.

Figure 3. Platinum X-ray absorption data at the K edge. The derivative of the data used to obtain the absorption edge is shown in the bottom.

probed by the X-ray source. However, in the location where platinum’s most intense peaks should be (74 eV for 4f5/2 and 71 eV for 4f7/2) there is only one small peak at 73 eV (shown in the bottom of Figure 2) that also happens to be in the area where the aluminum 2p peak is located. The fact that a small peak exists in the area where the Al 2s peak should appear (118 eV) while no peak is apparent where the Pt second most intense peaks should be (332 eV for 4d3/2 and 315 eV for 4d5/2) indicates that the peak at 73 eV corresponds to the aluminum 2p binding energy and that no platinum binding energy peak is evident in the spectrum. This result strongly suggests that the platinum is located inside zeolite pores and not on the external surface. This is an important result as it shows that within experimental error, all Pt is occluded in the zeolite micropores. The X-ray absorption spectrum of platinum at the K edge is shown in Figure 3. The position of the Pt absorption edge of our sample is located at 78.400 KeV, which is only 5 eV higher than the textbook value of 78.395 KeV for bulk metal platinum. Although this experiment was not done in a conventional way, we consider that the results are a good indication that the platinum in our sample is in the reduced state, only as it agrees with the other experimental results.

Figure 5. Z-contrast STEM image of the Pt-Beta sample. The bright spots are the platinum clusters in the zeolite.

SEM images were obtained of the approximately 130 × 30 nm enlongated Pt-Beta particles with the purpose of determining if there were any platinum particles formed on the zeolite surface. The images provided no evidence of platinum clusters present on the zeolite surface (Figure 4). Following the results of Rice et al.,22 Z-contrast images with an STEM highangle annular detector were also obtained to gain further insight into the size and location of the platinum clusters in our PtBeta sample. The image (Figure 5) shows bright white spots corresponding to the platinum clusters, gray areas corresponding to the zeolite matrix, and black areas corresponding to blank spaces. The platinum clusters represented show a relatively monodisperse distribution of sizes with symmetrical shapes. It is difficult to determine the actual cluster size from these images, as it is known that the observed particle widths are deceptive with this technique.22-24 From these images, we can also observe that there is no significant number of clusters around the borders of the zeolite matrix, which is a good indication that the platinum

Platinum Clusters Supported in Zeolite Beta

J. Phys. Chem. C, Vol. 111, No. 24, 2007 8577

GPt-Z - GZ ) 2

∑ Z

Figure 6. Experimental PDF curves of zeolite Beta, zeolite Pt-Beta, and their difference. The later is denominated the difference PDF.

Figure 7. Difference PDF and calculated bulk Pt PDF.

clusters are formed inside the pores. This agrees with the conclusions from the XPS measurements. 3.2. Results from the Pair Distribution Function Method. The experimental PDF curves for zeolite Beta, Pt-Beta sample, and their difference are shown in Figure 6. In principle, the difference between the two experimental PDF curves might be caused by changes in the zeolite atomic positions as well as to the presence of the platinum. However, there are three important results that indicate the changes in the PDF involve only Pt-X distances. First, the zeolite lattice parameters are not changing with increasing platinum loading as is evident from the XRD patterns (Figure 1). Second, the loading of the platinum is very small and thus is unlikely to cause changes in the overall zeolite PDF (although it is possible that local distortions exist). Finally, the third and most important characteristic is that all the peaks in the difference PDF correspond approximately to the positions of the peaks of bulk platinum (Figure 7). Theoretically the total pair distribution function is defined in terms of difference PDFs (GR(r)) as

GR(r) )

∑β

cβbβ 〈b〉

GRb(r)

(3)

where the difference PDF is the pair correlation function between all atoms and a particular chemical species that can be thought of being located in the origin. The total PDF is then given by

G(r) )

∑β

cRbR 〈b〉

GR(r)

(4)

In these equations, c is the fractional concentration, b is the scattering length, and 〈b〉 is the average scattering length of the sample. So strictly speaking, substracting the zeolite Beta PDF from the Pt-Beta PDF would yield

cZbZcPtbPt 〈b〉2

GPt-Z +

cPt2bPt2 〈b〉2

GPt-Pt (5)

where Z represents all the zeolite framework atoms (i.e., Si, O, Al, and Na). If we compare the GPt-Pt and GPt-O coefficients, we can observe that the GPt-O coefficient is approximately 87 times bigger than the GPt-Pt coefficient if we use the scattering lengths of the atoms at Q ) 0 Å-1 due to the low platinum concentration. So, if there is a Pt-O interaction, we would expect it to affect our difference PDF. Initially, however, we assumed that the difference curve corresponded only to the PtPt distances based on the principle that the Pt-Z interactions, and thus GPt-Z will be small compared to the Pt-Pt interactions.11 On the basis of the above observations, we substracted the Beta experimental PDF curve from the Pt-Beta experimental PDF curve to obtain a curve that contained only the Pt-X interatomic distances where X is either a Pt atom or any zeolite atom. From here on, we call this curve the difference PDF, and it is plotted along with the calculated PDF of bulk platinum for comparison purposes in Figure 7. For the calculated bulk Pt PDF, the profile and instrumental parameters were refined for proper comparison to the experimental PDF. The lattice parameter was refined to be 3.932 Å, a difference of 0.008 Å from the original (3.924 Å). Another one of the relevant results of this refinement is that the final scale factor of the model was 0.1 instead of the value of approximately one that is normally obtained with regular experimental data. This low scale factor reflects the fact that the platinum atoms are only a fraction of the atoms in the Pt-Beta sample. Normally, the coordination number can be calculated from the area of the peaks. However, in this case because the value of the scale is not one as the platinum is only a fraction of the total atoms the coordination number has to be determined through structural modeling. Nevertheless, contrary to EXAFS there is more information in the PDF curve that can help determine the average cluster size. We can observe that the functionality of the difference PDF is similar to that of the bulk platinum such that we can identify the Pt-neighbors interatomic distances. This suggests that our assumptions that the Pt-O interactions are less than the Pt-Pt interactions are correct. Thus, from the difference PDF we can determine the Pt-Pt neighbor distances as tabulated in Table 1 along with those of the bulk platinum. All the tabulated peaks are well defined in the difference PDF except that at 3.93 Å, which is convoluted with the termination ripple present in that area and actually appears at 4.02 Å. However, the fact that this peak is well described by the calculated bulk Pt PDF (Figure 7) assures us that the second neighbor Pt-Pt distance is 3.93 Å. After ∼15 Å, the difference PDF becomes very noisy. One of the disadvantages of using difference PDFs for refinement is that the PDF is always noisier than the original counterparts. The found bulk value of the first neighbor Pt-Pt distance (2.78 Å) agrees with our previous absorption spectroscopy finding that the platinum was in metallic state. Despite the good agreement between the calculated and experimental PDFs, the peaks of the fourth, fifth, and seventh neighbors are shifted from the bulk positions. Because the majority of the peaks have a good correspondence, we can conclude that all the platinum clusters have the same symmetry and the same Pt-Pt interactions as their bulk counterpart, and the cause of these shifts are other Pt-X interactions. Moreover, the absence of any strong Pt-zeolite peaks indicate that the Pt

8578 J. Phys. Chem. C, Vol. 111, No. 24, 2007

Martı´nez-In˜esta and Lobo

TABLE 1: Bulk and Experimental Neighbor Pt-Pt Distances distances (Å) coordination shell

bulk Pt

experimental

1 2 3 4 5 6 7 8

2.78 3.93 4.82 5.56 6.22 7.36 8.34 10.02

2.78 4.02 (3.93) 4.82 5.64 6.26 7.38 8.18 10.02

clusters are located at random locations within the pore. On the other hand, the fit between the calculated bulk PDF and the experimental difference PDF is not quantitative (intensity wise) so it is evident we need a better structural model. However, from the experimental PDF we can conclude that the average cluster has a dimension with at least 15 Å in mean effective diameter, as, after this distance, the experimental PDF has no significant structural information. One of the most evident differences in peak intensities is in the first peak at 2.78 Å. Two different effects can cause this difference; the first is the size effect and the other is the Ptzeolite interactions. We explored the possibility of the size effect by calculating the PDF of clusters of different sizes. One of the properties of calculating the PDF of size-limited clusters is the fact that the relative intensities of subsequent peaks decrease with increasing cluster size such that the ratio of intensities of the first neighbor peak to the third neighbor peak for the 1 × 1 × 1 platinum cluster is infinite as there is no third neighbor Pt-Pt peak, and for the 2 × 2 × 2 the ratios of intensities are finite for the six-coordination shell neighbors but infinite for the seventh, etc. We found that while roughly all the clusters bigger than 4 × 4 × 4 (15.72 Å wide) unit cells had peaks at the same position as that of the experimental PDF, all the peaks tended to be above the G(r) ) 0 line (Figure 8). The reason for this is that when the individual cluster PDF curves are calculated it is assumed that the average atomic density Fo is zero (there are no atoms) after a distance equal to the size of the cluster because there are no atoms after this distance. These results suggest that there are long distance interactions (after 15.72 Å) with other atoms that are influencing the difference PDF by increasing the average atomic density (from eq 2 it is evident that a nonzero value of Fo lowers the G(r)). These interactions could either be with other Pt clusters at larger distances, with the zeolite nearby, or with both. Figure 9 shows how the dimensions of a cubic cluster of this size compare to the dimensions of the intersection of the pores

Figure 9. Comparison of the dimensions of three platinum clusters with the dimensions of a pore intersection in zeolite beta. The schematics at the top are the structures viewed down the z-axis and the schematic at the bottom are the structures viewed down the x or y-axis. On beta schematic, the gray atoms represent the O and the black are the Si atoms.

in zeolite Beta. In this figure, we observe that a 4 × 4 × 4 cluster is bigger than the dimensions of the pore intersections, and that the 3 × 3 × 3 and the 2 × 2 × 3 clusters would fit better in the topology of the pore intersection. However, it is known that zeolite frameworks are flexible and can expand to fit large adsorbents, and it is also known that some adsorbed metals break the zeolite framework when they grow.7 Using a different approach and taking advantage of the relative good agreement of the bulk Pt PDF and the difference PDF, we refined the atomic positions of the bulk platinum using PDFfit (least-squares refinement) but found that the equilibrium structure was not changed even when the symmetry conditions were relaxed. We pursued this venue further with a RMC simulation of a crystal cluster of increasing size, and what we found was that the fit was not significantly improved. Only when a large crystal repeated in three dimensional was used were we able to describe quantitatively all the details in the experimental PDF. However, this model was completely unphysical as from the STEM image we know that the clusters occupy a discrete space in the pores and the success of the simulation is indeed an effect of having a large number of atoms being allowed to move. One difference that can be observed between the Pt PDFs and the experimental PDFs is in the intensity of the peak at 2.2 Å (Figures 7 and 8). This difference appears to go along with the results of the EXAFS study by Vaarkamp et al. mentioned above,10 and this peak might correspond to an interaction of the platinum with the framework zeolite oxygen. However, this peak is in an area of the PDF where it is probably convoluted with the termination ripples found at small interatomic distances in experimental PDFs as an artifact of the Fourier transform (eq 1). 4. Conclusions

Figure 8. Comparison between the difference PDF and the modeled PDF of the finite 4 × 4 × 4 Pt cluster. The PDF of bulk platinum is included for comparison.

We have shown that we can obtain the Pt-Pt interatomic distances from Pt clusters inside a zeolite Beta matrix at low loadings by substracting an experimental PDF curve of Al-Beta to that of the experimental PDF curve of the Pt-Beta. This is important because it shows that we can probe the supported Pt nanocluster structures in commercially relevant samples (i.e., samples with ∼1 wt % metal). By attempting to model the difference PDF using the bulk platinum structure, we learned that all the peaks had a significant correspondence suggesting

Platinum Clusters Supported in Zeolite Beta that the Pt clusters have the same symmetry as the bulk Pt. This is contrary to what might be expected for nanoclusters, where the surface tension of structures of this size generally changes their atomic conformation. We have shown that supported nanoclusters are an exception to this rule probably because of their energetic interaction with the zeolite matrix. Also, because there are no strong Pt-zeolite peaks indicates that the Pt clusters are located at a random orientation within the pore. From observations of the experimental PDF we conclude that there is a distribution of Pt cluster sizes with a minimum size of 15 Å and that the Pt is located in random locations within the zeolite. On the basis of the experimental results, we surmise that the sample Pt clusters differ from our ideal 4 × 4 × 4 Pt model due to the following: (a) That there are neighboring Pt clusters and zeolite atoms that interact with one Pt cluster and need to be taken into consideration. b) That the Pt clusters are not symmetrical and probably conform to the shape of the pores of the zeolites (elongated). c) That the Pt clusters are not of uniform size. Comparing the results of the pair distribution technique with the more commonly employed technique EXAFS, we can conclude that the two techniques are complementary in the structural information they provide. The element specificity of EXAFS makes it simpler to recognize the interactions of the target atom to other atoms. Also, with this method the coordination number is easily calculated. On the other hand, the PDF method provides larger range structural information of the platinum clusters in terms of interatomic distances, which gives more insight into the possible interactions of platinum atoms inside the cluster, interactions of platinum atoms with the zeolite matrix, and the interactions of platinum atoms in one cluster to those in another cluster without the need of any complex data manipulation, as is the case with EXAFS (in EXAFS an a priori model is needed to calculate the experimental data). The complexity of the information obtained in the difference PDF, however, demands more complex models to describe it completely. Parameters that the model would need to take into consideration are distribution of cluster sizes and shapes and distances between clusters and between Pt atoms and zeolite atoms.

J. Phys. Chem. C, Vol. 111, No. 24, 2007 8579 Acknowledgment. We would like to acknowledge Didier G. Wermeille and Douglas Robinson from Argonne National Lab for help taking the PDF data, Erich Weigert for the XPS data, and Dr. Chaoying Ni for the TEM and SEM images. Use of the Advanced Photon Source was supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No. W-31-109-Eng-38. We thank NSF award CTS-0085036 for the funding of this research. References and Notes (1) Boudart, M. J. Mol. Catal. 1985, 30, 27. (2) Boudart, M.; Dje´ga-Mariadassou, G. Kinetics of Heterogeneous Catalytic Reactions; Princeton University Press: Princeton, N.J., 1984. (3) Anderson, J. R. Structure of Metallic Catalysts; Academic Press: London, 1975. (4) Gates, B. C. Chem. ReV. 1995, 95, 511. (5) Poltorak, O. M.; Boronin, V. S. Russ. J. Phys. Chem. 1966, 40, 1436. (6) Sachtler, W. M. H. Catal. Today 1992, 15, 419. (7) Guzman, J.; Gates, B. C. Dalton Trans. 2003, 3303. (8) Gallezot, P. Catal. ReV. 1979, 20, 121. (9) Samant, M. G.; Boudart, M. J. Phys. Chem. 1991, 95, 4070. (10) Vaarkamp, M.; Modica, F. S.; Miller, J. T.; Koningsberger, D. C. J. Catal. 1993, 144, 611. (11) Egami, T.; Billinge, S. J. L. Underneath the Bragg Peaks : Structural Analysis of Complex Materials, 1st ed.; Pergamon: Amsterdam, 2003. (12) Martinez-Inesta, M. M.; Peral, I.; Proffen, T.; Lobo, R. F. Microporous Mesoporous Mater. 2005, 77, 55. (13) Martinez-Inesta, M. M.; Lobo, R. F. J. Phys. Chem. B 2005, 109, 9389. (14) Petkov, V.; Jeong, I. K.; Mohiuddin-Jacobs, F.; Proffen, T.; Billinge, S. J. L. J. Appl. Phys. 2000, 88, 665. (15) Gallezot, P.; Bienenstock, A. I.; Boudart, M. NouV. J. Chim. 1978, 2, 263. (16) Zheng, J.; Dong, J. L.; Xu, Q. H.; Liu, Y.; Yan, A. Z. Appl. Catal., A 1995, 126, 141. (17) M’Kombe, C. M.; Dry, M. E.; O’Connor, C. T. Zeolites 1997, 19, 175. (18) Jeong, I.-K.; Thompson, J.; Proffen, Th.; Turner, A. M. P.; Billinge, S. J. L. J. Appl. Cryst. 2001, 34, 536. (19) Proffen, T.; Billinge, S. J. L. J. Appl. Cryst. 1999, 32, 572. (20) Proffen, T.; Neder, R. B. J. Appl. Cryst. 1997, 30, 171. (21) Moulder, J. F.; Chastain, J. Handbook of X-ray Photoelectron Spectroscopy : A Reference Book of Standard Spectra for Identification and Interpretation of XPS Data; Physical Electronics Division, Perkin-Elmer Corp.: Eden Prairie, MN, 1992. (22) Rice, S. B.; Koo, J. Y.; Disko, M. M.; Treacy, M. M. J. Ultramicroscopy 1990, 34, 108. (23) Pennycook, S. J.; Howie, A.; Shannon, M. D.; Whyman, R. J. Mol. Catal. 1983, 20, 345. (24) Treacy, M. M. J.; Rice, S. B. J. Microsc. 1989, 156, 211.