pubs.acs.org/Langmuir © 2010 American Chemical Society
Preparation of Supported Pd Catalysts: From the Pd Precursor Solution to the Deposited Pd2þ Phase Giovanni Agostini,† Elena Groppo,† Andrea Piovano,† Riccardo Pellegrini,‡ Giuseppe Leofanti,‡,§ and Carlo Lamberti*,† †
Department of Inorganic, Physical and Materials Chemistry, NIS Centre of Excellence and INSTM Reference Center, University of Turin, Via P. Giuria 7, I-10125 Torino, Italy, ‡Chimet SpA - Catalyst Division, Via di Pescaiola 74, Viciomaggio Arezzo, I-52041 Italy, and §Consultant, Via Firenze 43, 20010 Canegrate, Milano, Italy Received February 3, 2010. Revised Manuscript Received March 31, 2010 The preparation by the deposition-precipitation method (using Na2PdCl4 as a palladium precursor and Na2CO3 as a basic agent) of Pd catalysts supported on γ-Al2O3 and on two different types of active carbons has been followed by several techniques (UV-vis, EXAFS, XRPD, and TPR). This work consists of four successive parts: the investigation of (i) the palladium precursor liquid solution (in the absence of substrate), (ii) the solid precipitated phase (in the absence of substrate), (iii) the precipitated Pd2þ-phase on the supports as a function of Pd loading from 0.5 to 5.0 wt % (i.e., the final catalyst for debenzylation reactions), and (iv) the Pd0-phase formed upon reduction in H2 atmosphere at 393 K. A time/pH-dependent UV-vis experiment indicates that Pd2þ is present in the mother solution mainly as PdCl2(H2O)2] and [PdCl(H2O)3]þ. Upon progressive addition of NaOH (3.0 < pH < ∼3.8), the concentration of the two complexes is almost constant and then they rapidly disappear because of the precipitation of an amorphous aggregation of Pd2þ-polynuclearhydroxo complexes. This phase represents a model material for the active supported phase. Thermal treatments at increasing temperature of this phase cause progressive water loss and resulted in a progressive increase in crystallinity typical of a defective PdO-like phase. The EXAFS spectrum of the final catalysts has been found to be intermediate between that of the unsupported amorphous Pd2þ-polynuclearhydroxo complexes and that of the PdOlike phase. Independent of the support, EXAFS was not able to evidence any fraction of reduced metallic Pd, meaning that all Pd is in the 2þ oxidation state within the sensitivity of the technique (a few percent). Analogously, independent of the support, XRPD was not able to detect the presence of any crystalline supported phase. The Pd local environment of the as-precipitated samples changes slightly as a function of Pd loading from 0.5 to 2.0 wt %: at higher loadings, no further modification has been observed. After reduction in an H2 atmosphere, two trends have been observed: (i) the dispersion of Pd nanoparticles tends to decrease with increasing Pd concentration, less significantly on Al2O3-supported samples and more significantly on carbon-supported ones and (ii) the dispersion depends on the carrier following the sequence Al2O3 . Cp > Cw according to the increasing palladium-support interaction strength.
1. Introduction Pd metal-supported catalysts are widely used in hydrogenation reactions for the synthesis of fine chemicals (e.g., active pharmaceutical ingredients)1 and bulk chemicals (e.g., terephthalic acid).2 Their activity and selectivity toward different molecules are strongly related to the morphology and dispersion of the metal active phase and to its electrostatic interaction with the support.3 The morphology and dispersion of Pd clusters are strongly dependent on the type of support and on the preparation method and can be modified on the basis of different factors: Pd precursor, temperature, and pH of the impregnating solution and kinetic factors.4 In particular, the support can have a direct influence on the catalytic reaction because its surface is often active toward reactants and reaction products but also an indirect one because the physical-chemical properties of the support can influence the metal dispersion, its resistance to sintering, and the *Corresponding author. Tel: þ39011-6707841. Fax: þ39011-6707855. E-mail
[email protected]. (1) Blaser, H.-U.; Indolese, A.; Schnyder, A.; Steiner, H.; Studer, M. J. Mol. Catal. A: Chem. 2001, 173, 3–18. (2) Pernicone, N.; Cerboni, M.; Prelazzi, G.; Pinna, F.; Fagherazzi, G. Catal. Today 1998, 44, 129–135. (3) van der Eerden, A. M. J.; Visser, T.; Nijhuis, A.; Ikeda, Y.; Lepage, M.; Koningsberger, D. C.; Weckhuysen, B. M. J. Am. Chem. Soc. 2005, 127, 3272– 3273. (4) Pellegrini, R.; Leofanti, G.; Agostini, G.; Rivallain, M.; Groppo, E.; Lamberti, C. Langmuir 2009, 25, 6476–6485.
11204 DOI: 10.1021/la1005117
accessibility of active sites to reactants. Some of these aspects have been investigated in the literature, such as the effect of the supporting chemical properties (e.g., acidity or basicity)5-8 or porosity.9,10 In this work, we have used several techniques to follow the preparation of Pd catalysts supported on γ-Al2O3 and on active carbons from wood (Cw) and peat (Cp) performed via a depositionprecipitation method using Na2PdCl4 as a palladium precursor and Na2CO3 as a basic agent. The study starts from a time/ pH-dependent investigation of the solution during precipitation via UV-vis spectroscopy. The amorphous unsupported precipitated phase is then fully characterized because it represents the model material for the active supported phase. For this reason, thermal activation at increasing temperature of the unsupported precipitated phase has been followed by extended X-ray absorption (5) Markus, H.; Plomp, A. J.; Maki-Arvela, P.; Bitter, J. H.; Murzin, D. Y. Catal. Lett. 2007, 113, 141–146. (6) Venezia, A. M.; La Parola, V.; Pawelec, B.; Fierro, J. L. G. Appl. Catal., A 2004, 264, 43–51. (7) Ruta, M.; Semagina, N.; Kiwi-Minsker, L. J. Phys. Chem. C 2008, 112, 13635–13641. (8) Ruta, M.; Laurenczy, G.; Dyson, P. J.; Kiwi-Minsker, L. J. Phys. Chem. C 2008, 112, 17814–17819. (9) Min, K.-I.; Choi, J.-S.; Chung, Y.-M.; Ahn, W.-S.; Ryoo, R.; Lim, P. K. Appl. Catal., A 2008, 337, 97–104. (10) Okhlopkova, L. B.; Lisitsyn, A. S.; Likholobov, V. A.; Gurrath, M.; Boehm, H. P. Appl. Catal., A 2000, 204, 229–240.
Published on Web 04/21/2010
Langmuir 2010, 26(13), 11204–11211
Agostini et al.
fine structure (EXAFS) spectroscopy, X-ray powder diffraction (XRPD), and temperature-programmed reduction (TPR). Successively, a characterization of the supported active phase has been carried out by EXAFS. Finally, Pd metal nanoparticles have been obtained by reducing the dried precipitated samples in an H2 atmosphere at 393 K. Metal particle dispersion has been investigated by combining CO chemisorption for the whole set of samples with EXAFS spectra collected on the most highly loaded samples. The large emphasis on the as-precipitated Pd2þ samples is justified by the fact that unreduced catalysts are directly used without prereduction treatments in some hydrogenation reactions such as debenzylation.11 Conversely, when reduced Pd catalysts are investigated, a systematic examination of the as-precipitated Pd2þ phase allows us to discriminate between precipitation and reduction effects in determining the final characteristics of the Pd0 nanoparticles.
2. Materials, Experiments, and Methods 2.1. Sample Preparation and Nomenclature. Supported
Pd samples have been prepared in the Chimet Laboratories on γAl2O3 (surface area = 121 m2 g-1; pore volume = 0.43 cm3 g-1) and on two different activated carbons, from wood (hereafter Cw: surface area = 980 m2 g-1; pore volume = 0.62 cm3 g-1) and peat (hereafter Cp: surface area = 980 m2 g-1; pore volume = 0.47 mm3 g-1) origin, following the deposition-precipitation method4,12-14 with Na2PdCl4 as a palladium precursor and Na2CO3 as a basic agent. For each support, the following Pd loadings have been prepared: 0.5, 1.0, 2.0, 3.5, and 5.0 wt % Pd. With the terminology Pd/Al2O3, Pd/Cw, and Pd/Cp, we indicate the generic catalysts prepared on alumina, Cw and Cp. When we are referring to a specific catalyst, then its Pd loading in wt % is reported in parentheses (i.e., Pd(1.0)/Al2O3 indicates 1.0 wt % Pd supported on alumina). As a reference, an unsupported sample has been prepared following a similar procedure adopted for the catalyst preparation, hereafter Pduns. All catalysts and Pduns have been carefully washed until complete Cl- removal has been achieved and have remained in their wet condition until measurement. The Pduns sample has been divided into four subsets, three of which have been dried at 393, 493, and 773 K, hereafter called samples Pduns(393), Pduns(493), and Pduns(773). Reduced Pd nanoparticles have been obtained from the as-precipitated samples by drying at 393 K and treatment in an H2 atmosphere at 393 K. Reduced samples will be named Pd/Cw_H2, Pd/Cp_H2, and Pd/Al2O3_H2, with obvious meanings. 2.2. UV-Vis Spectroscopy. The nature of the Pd-solvated species in the Na2PdCl4 mother solution during the Pd precipitation process has been monitored by means of UV-vis spectroscopy performed in transmission mode on a Varian Cary 300 scan instrument, whereas the precipitated phase has been investigated by UV-vis in reflectance mode on a Perkin-Elmer Lambda 19 instrument. 2.3. TPR and CO Chemisorption. A Micromeritics Autochem 2910 instrument has been used for both the TPR and CO chemisorption measurements. For the TPR measurements, the instrument is equipped with a CryoCooler device for reaching subambient temperature by (11) Pearlman, W. M. Tetrahedron Lett. 1967, 8, 1663–1664. (12) Geus, J. W.; van Dillen, A. J. Preparation of Supported Catalysts by Deposition-Precipitation. In Handbook of Heterogeneous Catalysis; Ertl, G., Kn€ozinger, H., Weitkamp, J., Eds.; VCH: Weinheim, Germany, 1997; Vol. 1, pp 240-257. (13) Simonov, P. A.; Likholobov, V. A. Physicochemical Aspects of Preparation of Carbon-Supported Noble Metal Catalysts. In Catalysis and Electrocatalysis at Nanoparticle Surfaces; Wieckowski, A., Savinova, E. R., Vayenas, C. G., Eds.; Marcel Dekker: New York, 2003; pp 409-454. (14) Bertarione, S.; Prestipino, C.; Groppo, E.; Scarano, D.; Spoto, G.; Zecchina, A.; Pellegrini, R.; Leofanti, G.; Lamberti, C. Phys. Chem. Chem. Phys. 2006, 8, 3676–3681.
Langmuir 2010, 26(13), 11204–11211
Article means of liquid nitrogen. A molecular sieve trap has been put between the sample holder and the detector in order to absorb water that is formed during the reduction process. A wet sample (100 mg) has been introduced in the sample holder and dried in situ by nitrogen flow at 393 K for 2 h.15 Then the sample has been cooled to 153 K under Ar flow (50 cm3 min-1). After this step, the flow was switched to 5% H2 in Ar (50 cm3 min-1) and was maintained throughout the analysis. Once a baseline was established, the temperature ramp was started at a rate of 5 K min-1 up to 623 K. TPR experiments on Pduns(393), Pduns(493), and Pduns(773) reference compound samples have been carried out on 10 mg of sample following the same procedure described above for the catalysts. CO chemisorption measurements have been performed by a dynamic pulse method at 323 K on samples prereduced in the H2 atmosphere at 393 K, with the same procedure as described for TPR measurements, by using He instead of N2. A CO/Pd average stoichiometry of 1 has been assumed for the calculation of dispersion. This assumption has been verified by a parallel series of measurements performed in a static volumetric apparatus on three samples reduced at two different temperatures (393 and 673 K) with H2 and O2 as probe molecules that give a O/Pd average stoichiometry of close to 1.16,17 The two methods/ techniques give a CO/O ratio in the 0.94-1.13 range, which is strong support for the assumption that the CO/Pd average stoichiometry is 1. 2.4. X-ray Absorption Spectroscopy. EXAFS spectra in transmission mode with a high signal-to-noise ratio require the precise calibration of sample thickness. However, when working on samples where the Pd concentration varies within 1 order of magnitude (from 0.5 to 5.0 wt %), significant changes in the sample thickness are required: 35 to 9 mm (18 to 4 mm) for Pd/C (Pd/Al2O3) catalysts. For this reason, an ad hoc homemade sample holder (vide infra, Figure 2a) whose construction has been inspired from the cells for liquids was adopted. The thickness of the cell, representing an excellent compromise between simplicity and efficiency, can be optimized within the 4-50 mm range. The samples were hosted as wet unpressed powders and were protected with parafilm during the measurement to avoid a partial drying effect. As far as reduced catalysts are concerned, the asprecipitated samples, in the form of self-supported pellets of optimized thickness, have been located inside an ad hoc conceived cell18a that allows in situ evacuation at 393 K (in this case), in situ removal of the H2O reduction product, cooling down at room temperature and XAFS collection under vacuum conditions. X-ray absorption experiments at the Pd K-edge (24350 eV) were performed at the BM26A beamline18b of the ESRF facility (Grenoble, France). The white beam was monochromatized using a Si(111) double crystal; a harmonic rejection has been performed using Pt-coated silicon mirrors. The following experimental geometry was adopted: (1) I0 (10% efficiency), (2) sample, (3) I1 (40% efficiency), (4) reference Pd foil, and (5) I2 (80% efficiency). This setup allows a direct energy/angle calibration for each spectrum by avoiding any problem related to small energy shifts due to the small thermal instability of the monochromator crystals.19 The EXAFS parts of the spectra were collected with a variable sampling step in energy, resulting in Δk = 0.05 A˚-1 up to (15) Pellegrini, R.; Leofanti, G.; Agostini, G.; Bertinetti, L.; Bertarione, S.; Groppo, E.; Zecchina, A.; Lamberti, C. J. Catal. 2009, 267, 40–49. (16) Prelazzi, G.; Cerboni, M.; Leofanti, G. J. Catal. 1999, 181, 73–79. (17) Agostini, G.; Pellegrini, R.; Leofanti, G.; Bertinetti, L.; Bertarione, S.; Groppo, E.; Zecchina, A.; Lamberti, C. J. Phys. Chem. C 2009, 113, 10485–10492. (18) (a) Lamberti, C.; Prestipino, C.; Bordiga, S.; Berlier, G.; Spoto, G.; Zecchina, A.; Laloni, A.; La Manna, F.; D’Anca, F.; Felici, R.; D’Acapito, F.; Roy, P. Nucl. Instrum. Methods B 2003, 200, 196–201. (b) Nikitenko, S.; Beale, A. M.; van der Eerden, A. M. J.; Jacques, S. D. M.; Leynaud, O.; O'Brien, M. G.; Detollenaere, D.; Kaptein, R.; Weckhuysen, B. M.; W., B. J. Synchrotron Radiat. 2008, 15, 632–640. (19) Lamberti, C.; Bordiga, S.; Bonino, F.; Prestipino, C.; Berlier, G.; Capello, L.; D’Acapito, F.; Llabres i Xamena, F. X.; Zecchina, A. Phys. Chem. Chem. Phys. 2003, 5, 4502–4509.
DOI: 10.1021/la1005117
11205
Article
Agostini et al.
20 A˚-1, with an integration time that linearly increases with k from 4 to 25 s/point to account for the low signal-to-noise ratio at high k values. The extraction of the χ(k) function has been performed using the Athena code.20 For each sample, two consecutive EXAFS spectra have been collected and corresponding χ(k) functions have been averaged before data analysis. EXAFS data analysis has been performed using Arthemis software.20 The phase and amplitudes have been calculated by the FEFF6 code21 and have been successfully checked with Pd metal foil. For each sample, the averaged k3χ(k) function was Fourier transformed in the Δk = 2.00-15.75 A˚-1 interval. For the Pduns model compound, the experimental spectrum was fitted in R space in the ΔR = 1.00-4.06 A˚ range (2ΔkΔR/π > 25). At higher R values, the EXAFS spectra of the Pd2þ phase are dominated by noise, as appreciable in the plots reported in the Supporting Information, where the same curves are reported up to 10 A˚. 2.5. X-ray Powder Diffraction. XRPD data on the reference samples were collected at BM26A18b in parallel with EXAFS measurements with a circular position-sensitive INEL CPS 590 detector (sample-to-detector distance R = 50 cm) at an energy of 24.00 keV corresponding to λ = 0.5165 A˚ after 5 min of integration time. The experimental conditions were optimal for X-ray absorption measurements (self-supported pellets, diluted in BN) but were not optimized for diffraction experiments for two reasons: (i) the sample is placed in the center of the detector circle with a precision of ∼5 mm (i.e., channel-angle calibration is necessary for any acquisition) and (ii) the sample cannot be considered to be a point source of the diffracted beam because its thickness, optimized for EXAFS, is nonnegligible (on the order of millimeters) so as to cause peak-broadening effects. The channel-angle calibration was performed a posteriori using the well-defined peaks of the BN phase that acts as an internal standard. This procedure allows us to compare the different patterns qualitatively, thus observing the progressive crystallization of the PdO phase, in addition to the EXAFS data. Unfortunately, for the reasons outlined above a quantitative Rietveld refinement was not possible.
3. Unsupported Precipitated Pd2þ Phase 3.1. In Situ UV-Vis Monitoring of the Precipitation Process. The deposition-precipitation method used to prepare Pd/Cw, Pd/Cp, and Pd/Al2O3 samples implies the instantaneous precipitation of the Pd2þ species of the acidic precursor whose pH value immediately reaches that of the alkaline slurry. In this way, we are unable to follow the rapid pH change of the Na2PdCl4 drop in time. Consequently, to follow the nature and abundance of the Pd2þ-solvated species spectroscopically during the precipitation process we decided to invert the process, adding increasing amounts of base to the starting water solution of Na2PdCl4 (3.2 10-4 M, pH 3.0, no support is present in this case). After each base addition, the pH of the solution was recorded and a negligible fraction of the liquid has been sampled for UV-vis measurements in transmission mode. The results are displayed in Figure 1a,b. The optical spectrum of the starting Na2PdCl4 solution (black curve) is characterized by two main absorption bands at 48 000 and 42 500 cm-1 that, according to the literature,22 are ascribed to the charge-transfer (CT) transition of [PdCl(H2O)3]þ and [PdCl2(H2O)2] complexes, respectively. The two weaker components at 24 500 and 33 500 cm-1 are associated with two CT bands of the same species characterized by a much lower extinction coefficient. From the equilibrium constants (20) Ravel, B.; Newville, M. J. Synchrot. Radiat. 2005, 12, 537–541. (21) Ankudinov, A. L.; Ravel, B.; Rehr, J. J.; Conradson, S. D. Phys. Rev. B 1998, 58, 7565–7576. (22) Elding, L. I.; Olsson, L. F. J. Phys. Chem. 1978, 82, 69–74.
11206 DOI: 10.1021/la1005117
Figure 1. (a) From black to brown curves, evolution of the transmission UV-vis spectra of the Na2PdCl4 precipitation solution (3.2 10-4 M) as a function of the pH increase (where only a selection of curves have been reported). The CT bands of [PdCl(H2O)3]þ and [PdCl2(H2O)2] complexes are identified by circle and diamond symbols, respectively. (b) Absorption intensity of the two main CT components as a function of pH. The color code adopted in part b allows us to recognize the selected curves in part a. (c) UV-vis DRS spectrum of the precipitated Pd(OH)2 phase (K.M. = Kubelka-Munk).
reported elsewhere,23 the expected concentrations in the adopted conditions are ∼50% [PdCl2(H2O)2], ∼40% [PdCl(H2O)3]þ, and ∼10% [PdCl3(H2O)]-. The main spectroscopic feature of this last, less-abundant species expected at 41 800 cm-1 is not visible in Figure 1a because its extinction coefficient is too low.22 Upon progressive addition of base (i.e., increasing the pH), the two main CT bands are almost unaffected until pH ∼3.8, when they rapidly disappear because of the precipitation of the Pd2þ phase (Figure 1c), hereafter called Pduns. 3.2. Structural Characterization of the Unsupported Pd2þ Precipitated Phase. The optical properties of the precipitated phase, mainly constituted of Pd2þ-polynuclearhydroxo complexes,24-26 have been investigated by DRS UV-vis spectroscopy (Figure 1c). No vestige of the well-defined CT bands of isolated [PdClx(H2O)4 - x]2 - x complexes was present in the UVvis DRS spectrum, testifying that all Pd2þ ions lie in an aggregated phase characterized by a broad CT band at much lower energy centered at around 20 000 cm-1. The XRD pattern (not reported) of Pduns reveals its amorphous nature. Therefore, to obtain structural information, we used EXAFS spectroscopy, which is able to determine the local structure around the absorbing atom (Pd in this case) also in the absence of long-range order. The EXAFS spectrum of Pduns is shown in Figure 2c,d in both k- and R-spaces. The |FT| is dominated by a first-shell contribution centered at around 1.6 A˚ and by a weaker and complex contribution at a longer distance (two peaks at 2.7 and 3.2 A˚, respectively). To model our EXAFS data, we constructed the cluster reported in Figure 2b on the basis of the structural model proposed for the Pd2þ-polynuclearhydroxo (23) Elding, L. I. Inorg. Chim. Acta 1972, 6, 647–651. (24) Troitskii, S. Y.; Chuvilin, A. L.; Kochubei, D. I.; Novgorodov, B. N.; Kolomiichuk, V. N.; Likholobov, V. A. Russ. Chem. Bull. 1995, 44, 1822–1826. (25) Troitskii, S. Y.; Chuvilin, A. L.; Bogdanov, S. V.; Moroz, E. M.; Likholobov, V. A. Russ. Chem. Bull. 1996, 45, 1296–1302. (26) Rotunno, F.; Prestipino, C.; Bertarione, S.; Groppo, E.; Scarano, D.; Zecchina, A.; Pellegrini, R.; Leofanti, G.; Lamberti, C. In Preparation of Pd/C Catalysts: From the Pd-Precursor Solution to the Final Systems; Gaigneaux, E. M., Devillers, M., De Vos, D. E.,Hermans, S., Jacobs, P. A., Martens, J. A., Ruiz, P., Eds.; Studies in Surface Science and Catalysis; Elsevier: Amsterdam: 2006; Vol. 162, pp 721-728.
Langmuir 2010, 26(13), 11204–11211
Agostini et al.
Figure 2. (a) Photograph of the ad hoc-designed high-energy EXAFS cell. The cell is optimized for powdered samples. (b) Pd(OH)2 cluster used to model the EXAFS signal up to 4.1 A˚ around the absorbing Pd atom (labeled as Pd0). Gray and red spheres represent Pd and O atoms, respectively. H atoms, which are unable to contribute significantly to the EXAFS signal, have been omitted. The model has been constructed from the studies of Troitskii et al.24,25 The structural parameters optimized in the EXAFS fit are the Pd-O first-shell distance (considered to be equivalent for O1a, O1b, and O1c atoms), the Pd0-Pd1, Pd0-Pd2, and Pd0-O2 distances, and the corresponding Debye-Waller factors. Angles R and δ differ from the ideal 90° of square-planar geometry in order to account for the optimized Pd0-Pd1, Pd0-Pd2, and Pd-O1 distances. Because the two PdO4 units centered in Pd0 and Pd2 are linked via a single oxygen bridge (O1b), the PdO4 unit centered on Pd2 is, in principle, free to rotate. The optimized Pd0-O2 distance allows us to evaluate the angle θ between the Pd0-Pd2 and Pd2-O2 vectors. (c) Comparison between experimental (scattered black dots) and best fit (red full line) in k space, adopting a k3 weight. (c) k3-weighted, phase-uncorrected k3 χ(k) functions reported in part c. Both the modulus and imaginary parts have been reported. Vertical arrows in parts (c) and (d) indicate the FT and the fit ranges, respectively.
complexes by Troitskii et al.24 In this model, Pd2þ-polynuclearhydroxo complexes form filaments of about 100 Pd atoms. Because our EXAFS data exhibits signal at up to 4 A˚ (Figure 2c,d), we have truncated the cluster to the labeled atoms in Figure 2b. The local environment of Pd2þ consists of planar coordination squares of PdO4 units linked via two (O1a) or one (O1b) oxygen bridges with different geometry. The remaining Pd-O bond connects an OH group (O1c). To limit the number of scattering paths and thus free parameters in the EXAFS fit, PdO4 units linked via two oxygen bridges are forced to be coplanar, resulting in two Pd (Pd0 and Pd1 in Figure 2b) and six O atoms in the same plane. Moreover, Pd2 has been forced to lie on the same plane. Even if such a geometrical approximation seems to be very crude, it does not affect the fit results; in fact, no multiple scattering (MS) path involving Pd0 and Pd1 or Pd0 and Pd2 has significant weight, so both Pd1 and Pd2 scatterers contribute with their single scattering (SS) signal only, from which no relative angle can be estimated. According to this model, the most important scattering paths are the SS Pd0-O1 (4-fold degenerate), Pd0-Pd1, Pd0-Pd2, Pd0-O2, the collinear MS involving diametrically opposite O1 atoms, and the MS triangular paths involving Pd0 and two adjacent O1 atoms. The three Pd0-O1a, Pd0-O1b, and Pd0-O1c distances can, in principle, be different. A fit of the first-shell contribution (0.9-2.0 A˚ range in Figure 2) with two independent Langmuir 2010, 26(13), 11204–11211
Article
oxygen shells resulted, within experimental error, in the same distance and Debye-Waller factors for the two shells. Consequently, we assumed all Pd-O1 distances to be equivalent (within the accuracy of the technique). Therefore, the EXAFS spectrum was optimized with the following 10 free parameters: a single amplitude factor (S02) and a single energy shift (ΔE) common to all paths, the Pd0-O1, Pd0-Pd1, Pd0-Pd2, and Pd0-O2 distances (RO1, RPd1, RPd2, and RO2), and their corresponding Debye-Waller factors (σO12, σPd12, σPd22, and σO22). The quantitative results of the fit are reported in Table 1, and the quality of the fit can be appreciated in Figure 2c,d for k and the R space, respectively. Optimized distances RO1 = 2.02 ( 0.004 A˚ and RPd1 = 3.03 ( 0.01 A˚ allow us to determine R = 83 ( 3°. Analogously, RPd2 = 3.42 ( 0.03 A˚ implies δ = 135 ( 5°, and finally RO2 = 3.66 ( 0.04 A˚ results in θ = 75 ( 8°. The quality of the EXAFS fit (Rfactor = 0.027) validates the structural model proposed by Troitskii et al.24 and implies that the nature of the precipitated phase is a [Pd(OH)2]n colloid characterized by high local order (strong EXAFS signal up to 4.1 A˚) and by poor long-range order (XRD silent). With respect to the model proposed in 1995 by Troitskii et al.,24 the higher quality of our EXAFS data allows us to improve the local geometry by determining the R, δ, and θ angles fixed at 90, 122, and 90° in previous work. The basic validation of the model proposed by Troitskii et al.24 is very important because it proves that Cl- ions that are present in the first coordination shell of Pd2þ when it is present in solution (CT bands in Figure 1a) leave the Pd2þ environment in the precipitated phase so that no Pd-Cl signal is present in the EXAFS spectrum. 3.3. Thermal Treatment of the Unsupported Pd2þ Precipitated Phase. It has just been shown that Pduns is an amorphous, nonstoichiometric Pd-hydroxide phase. As will be reported in section 4, Pd2þ-supported samples exhibit a degree of local order higher than that found on Pduns; hence, more ordered Pd2þ model compounds are needed. For this reason, the Pduns sample has been subjected to progressive thermal treatments at increasing temperature, resulting in a gradual transformation into a more ordered Pd-oxide phase: samples Pduns(393), Pduns(493), and Pduns(773). The properties of the Pduns phase at each activation temperature have been investigated by XRPD (Figure 3a), EXAFS (Figure 3b,c), and TPR (Figure 3d). Starting with XRPD, the Pduns phase (pattern not reported) did not present any diffraction peak distinguishable from the background besides those of the BN phase used as a diluent. The crystallization and sintering processes of the starting amorphous Pduns phase upon increasing the treating temperature are clearly visible, on qualitative grounds, from the XRPD patterns reported in Figure 3a. The most intense reflection of the PdO phase at 2θ = 9.5° due to the (110) plane (dspacing = 3.18 A˚) is already visible after heating to 393 K, resulting in a broad peak (fwhm = 1.14°). Thermal treatments at 493 and 773 K progressively reduce the fwhm to 0.95 and 0.28° respectively, with the concomitant appearance of new well-defined reflections of the PdO phase (see asterisks) reflecting the occurrence of a disorder-order transition. Note that BN used to optimize the EXAFS measurements results in well-defined, sharp diffraction peaks that scarcely interfere with the Pd-oxide phase. The intensity of the BN diffraction peaks is strongly affected by the pressure adopted to prepare the pellets (compare the gray and black patterns in Figure 3a); the effect is due to preferred orientations. Smaller changes in the relative intensity of the BN peaks measured in the different samples can be explained in terms of slightly different pressures adopted in the pellet preparation. DOI: 10.1021/la1005117
11207
Article
Agostini et al. Table 1. Summary of the EXAFS Fit of the Pd(OH)2 Precipitated Phasea Pd0-O1
R(A˚)
Pd0-Pd1 σ (A˚2) 2
R(A˚)
Pd0-Pd2
σ (A˚2) 2
R(A˚)
Pd0-O2
σ (A˚2) 2
R(A˚)
ΔE
σ (A˚2) 2
eV
S0 2
2.023 ( 0.005 0.0041 ( 0.0005 3.03 ( 0.01 0.0040 ( 0.0005 3.42 ( 0.03 0.007 ( 0.002 3.66 ( 0.04 0.008 ( 0.005 2.5 ( 0.9 0.96 ( 0.07 a The fit validity can be appreciated from the quality factor R = 0.027 and from the low values of main correlations among optimized variables: S02/σO12 = 0.79 and ΔE/RO1 = 0.75. (The remaining correlations are below 0.7 in absolute value.).
Figure 4. (a) Modulus of the k3-weighted, phase-uncorrected FT of the EXAFS signal of Pd/Cw samples with increasing Pd loadings of 0.5, 1.0, 2.0, 3.5, and 5.0 wt % for samples Pd(0.5)/Cw, Pd(1.0)/ Cw, Pd(2.0)/Cw, Pd(3.5)/Cw, and Pd(5.0)/Cw, respectively. (b) The same as in part a for the imaginary part. Pd(OH)2 and PdO model compounds heated to 773 K are also reported for comparison. (c) Corresponding TPR curves.
Figure 3. XRPD, EXAFS, and TPR of the precipitated Pd(OH)2 phase as such (Pduns, orange curves) and thermally treated at 393, 493, and 773 K (red, green, and blue curves, respectively). (a) XRPD patterns collected just before the EXAFS data collection. Samples are in the form of pellets of optimized thickness for absorption measurements that have been pressed with BN to make them selfsupporting. For comparison, the XRPD patterns collected on pressed and unpressed BN are also reported (black and gray curves, respectively). Beam energy = 24 keV, corresponding to λ = 0.5165 A˚. (b) k3-weighted, phase-uncorrected FT of the EXAFS signal. (c) The same as for part b for the imaginary part. (d) TPR patterns.
Figure 3b,c reports the k3-weighted, phase-uncorrected FT of the EXAFS signal of the precipitated Pduns phase and that obtained after subsequent heating of the sample to 393, 493, and 773 K. The crystallization process that occurred on sample Pduns(773) K is clearly observed by EXAFS because the corresponding FT exhibits a signal up to R = 4.5 A˚, typical of longrange-ordered materials, but for the as-precipitated phase and the sample heated to 493 K, the signal stops at around 3.5 A˚. Going into more detail, the EXAFS data reported in Figure 3b, c confirm on local grounds what was concluded in the XRPD study discussed above. The first shell signal (around 1.6 A˚) is basically unaffected by the heating process because all Pd atoms lie in square-planar geometry, independent of the degree of hydroxylation and sintering (Figure 3b); this invariance will also hold for catalysts (section 4). Conversely, the second and third peaks (around 2.7 and 3.2 A˚) are significantly modified and increase in intensity according to the formation of new, more ordered Pd-O-Pd bridges upon dehydroxylation (Figure 2b). Of particular interest is the peculiar inversion between the intensity of these two peaks that can be considered to be the EXAFS fingerprint of effective dehydroxylation. It is evident that both hydroxide and oxide phases have very similar EXAFS spectra, with imaginary parts that are almost totally in phase (Figure 3c). On this basis, it is evident that on the catalyst samples, where Pd metal, hydroxide, and oxide phases can possibly coexist, it will be virtually impossible to discriminate EXAFS data between hydroxide and oxide phases. 11208 DOI: 10.1021/la1005117
The TPR pattern of sample Pduns (orange curve in Figure 3d) exhibits the main reduction peak starting at 252 K with a Tmax of 264 K. This peak is followed by a shoulder that is at least partially due to hydride formation and by the successive negative peak due to hydride decomposition.15 Upon heating, the main Pd2þ reduction peak is shifted to higher temperatures with a starting temperature of ∼263 K for both Pduns(493) and Pduns(773) and Tmax values of 274 and 277 K for Pduns(493) and Pduns(773), respectively. We can conclude that the dehydration caused by calcination is responsible for the increase in the reduction temperature by ∼10 K. This is the behavior expected as a consequence of the increase in order, which decreases the density of surface defects that act as preferential starting points for the reduction process.
4. Supported Precipitated Pd Phase from 0.5 to 5.0 wt % The use of the deposition-precipitation method to prepare supported palladium catalysts directly provides palladium in an unreduced state (oxide/hydroxide) deposited into the pores of the carrier material. The investigation of this Pd phase is important not only because it represents an indispensable step in the preparation of supported Pd nanoparticles but also because these unreduced catalysts can be used in hydrogenation reactions without further treatment (e.g., debenzylation11). Furthermore, the final industrial catalyst, being either in a reduced or unreduced state, is almost always kept in the wet form for safety handling and storage and is loaded as such into the hydrogenation reactor. Therefore, in this section, we have characterized the catalysts in the original wet state because the drying pretreatment could alter the catalyst properties, specifically, the degree of reduction of Pd. The local environment of Pd in Pd/Cw, Pd/Cp, and Pd/Al2O3 systems has been investigated as a function of Pd loading from 0.5 up to 5.0 wt %. As an example, Figure 4a,b reports both moduli and imaginary parts of the k3-weighted, phase-uncorrected FT of the EXAFS signal for the as-precipitated Pd/Cw samples. The spectra of Pduns Langmuir 2010, 26(13), 11204–11211
Agostini et al.
Figure 5. (a) Modulus of the k3-weighted, phase-uncorrected modulus of the FT of the EXAFS signal of 0.5 wt % Pd supported on Cw, Cp, and Al2O3. (b) The same as for part a for the 5.0 wt % Pd-supported catalysts. (c) TPR curves of 1.0 wt % Pd supported on Cw and Al2O3. (d) The same as for part c for the 5.0 wt % Pdsupported catalysts. Gray dashed lines in parts c and d show the small differences in the reduction temperature.
and Pduns(773) are also reported for comparison. The data on Pd/ Cw reported in Figure 4a,b are almost equivalent, showing the oxidic nature of the Pd precursor at any investigated loading. In particular, the signal is intermediate between that for model compound Pd(OH)2 and that for model compound PdO heated to 773 K. The third-shell contribution (centered at around 3.2 A˚) shows a small intensity increase upon increased Pd loading, and saturates (at 2.0 wt % Pd) at a value much lower than that observed for PdO heated to 773 K, indicating the dispersed nature of the oxidic Pd precursors on the Cw support. Independent XRPD measurements (not shown) are characterized by the broad bumps typical of carbons and did not show the presence of any crystalline phase. In the EXAFS spectra, no evidence of reduced Pd metal is appreciable. According to the sensitivity of the technique, we can affirm that if reduced Pd particles are present at this stage then their concentration is below 1 to 2%. This evidence is important, particularly for samples prepared on carbon supports, where the presence of several chemically active species could, in principle, have a reducing effect on the precipitated phase. The effects of the nature of the support and of the Pd loadings are shown in Figure 5, where the data for both low and high Pdloaded samples on the three supports are reported. At the lowest investigated loadings (Figure 5a), the difference in particle dimensions of the amorphous PdO precursor is appreciable from the intensity of the higher shell signal at 2.3-3.5 A˚: Pd/Cp < Pd/Cw < Pd/Al2O3. For the 5.0 wt % Pd-loaded samples (Figure 5b), this trend holds but becomes borderline within the sensitivity of the technique. All TPR profiles (Figures 4c and 5c,d) have two main features: (i) complex H2 consumption formed by a peak plus a shoulder in the 260-330 K region due to the Pd2þ f Pd0 reduction and the surface PdHx and bulk Pd-hydride phase formation and (ii) a negative peak due to Pd-hydride decomposition. Note that 1.0 wt % Pd-supported samples have been chosen to be representative of the low loaded catalysts in TPR experiments because the results obtained on 0.5 wt % Pd samples resulted in a TDC signal that was too low. TPR results can be summarized as follows. (a) The TPR profile of the Pduns sample is markedly different from those of the Langmuir 2010, 26(13), 11204–11211
Article
supported catalysts, and the same holds true for the EXAFS spectra. (b) The starting temperature of reduction (Tstart) as well as the peak temperature (Tpeak) of Pd/Cw samples decreases with increasing Pd concentration, more markedly at low Pd concentration and only slightly at high concentration (Figure 4c). This trend mirrors that observed in the FT of the EXAFS spectra (Figure 4a,b). TPR is more sensitive than EXAFS to Pd concentration because Tstart and consequently Tpeak depend on the Pd crystal surface rather than on the bulk. (c) Conversely, samples supported on Al2O3 are characterized by TPR and EXAFS results that are much less dependent on the Pd loading: the main peak temperature ranges from 268 to 273 K in Pd/Al2O3 samples and from 267 to 278 K in Pd/Cw samples. (d) The difference among the supports is clearly observed at low Pd loading (Figure 5a,c), but it is much less visible at high loading (Figure 5bd); in particular, the Tpeak of Pd/Cw and Pd/Al2O3 samples differs by 5 and 1 K for Pd(1.0) and Pd(5.0) samples, respectively. Analogies and differences observed among the different samples by EXAFS spectroscopy are fully supported by parallel TPR, with the latter being more sensitive to the type of support and Pd concentration because it is affected by a Pd surface that is, in turn, more sensitive to these characteristics. In summary, under the adopted preparation conditions, Pd precipitation occurs in the liquid phase (i.e., independently of the presence and nature of the support (section 3.1). Besides the very first steps in the precipitation process (0.5 and 1.0 wt % Pd), this results in samples having a Pd2þ-supported phase that has the same Pd local environment (Figure 5b) and the same H2 reducibility (Figure 5d).
5. H2-Reduced Pd Nanoparticles from 0.5 to 5.0 wt % From the TPR experiments reported in section 4, we learned that 393 K (in an H2 atmosphere) is by far a safe temperature to allow complete Pd2þ f Pd0 conversion for all as-precipitated samples. This hypothesis is verified by the total absence of hydrogen consumption in TPR experiments made after in situ H2 reduction at 393 K (curves not reported). This result is confirmed by the XANES spectra reported in Figure 6a for the highest loaded samples. Pd(5.0)/Cw_H2, Pd(5.0)/Cp_H2, and Pd(5.0)/Al2O3_H2 samples show virtually indistinguishable XANES spectra that are red shifted by about 3 eV with respect to the edge of the as-precipitated samples. (See the spectrum of sample Pd(5.0)/Cw reported for comparison.) The XANES spectra of the reduced samples are perfectly aligned with that of the Pd foil but exhibit smaller oscillations in the near-edge region because of the nanometer size of the metal particles. The nanometric nature of the Pd metal particles is better appreciated in the k3-weighted, phase-uncorrected Fourier transform (FT) of the EXAFS spectra14,17,27-34 (Figure 6b). In all cases, H2 reduction completely modifies the FT curves of the asprecipitated samples, showing the disappearance of the first-shell (27) Frenkel, A. I.; Hills, C. W.; Nuzzo, R. G. J. Phys. Chem. B 2001, 105, 12689–12703. (28) Keresszegi, C.; Grunwaldt, J. D.; Mallat, T.; Baiker, A. J. Catal. 2004, 222, 268–280. (29) Caravati, M.; Meier, D. M.; Grunwaldt, J. D.; Baiker, A. J. Catal. 2006, 240, 126–136. (30) Grunwaldt, J. D.; Caravati, M.; Baiker, A. J. Phys. Chem. B 2006, 110, 25586–25589. (31) Sun, Y.; Frenkel, A. I.; Isseroff, R.; Shonbrun, C.; Forman, M.; Shin, K. W.; Koga, T.; White, H.; Zhang, L. H.; Zhu, Y. M.; Rafailovich, M. H.; Sokolov, J. C. Langmuir 2006, 22, 807–816. (32) Frenkel, A. Z. Kristallogr. 2007, 222, 605–611. (33) Knecht, M. R.; Weir, M. G.; Frenkel, A. I.; Crooks, R. M. Chem. Mater. 2008, 20, 1019–1028. (34) Groppo, E.; Liu, W.; Zavorotynska, O.; Agostini, G.; Spoto, G.; Bordiga, S.; Lamberti, C.; Zecchina, A. Chem. Mater. 2010, 22, 2297–2308.
DOI: 10.1021/la1005117
11209
Article
Agostini et al.
Figure 6. (a) Normalized XANES and (b) EXAFS (k3-weighted, phase-uncorrected FT) data of Pd(5.0)/Cw (blue lines), Pd(5.0)/Cp (red lines), and Pd(5.0)/Al2O3 (black lines) reduced in an H2 atmosphere at 393 K. The inset of (b) reports a magnification of the higher-shell contributions. For comparison, the data of the as-precipitated Pd(5.0)/Cw (dotted blue curves) and the Pd foil reference compound (gray curves) are also reported.
Figure 7. Pd dispersion of catalysts measured by CO chemisorption as a function of the palladium content: from 0.5 to 5%. Samples were reduced in situ with H2 at 393 K. Blue, red, and black symbols refer to Cw, Cp, and Al2O3 supports, respectively. Lines of the same color report the corresponding best linear fits.
Pd-O signal around 1.6 A˚ and of the higher-shell contributions of the oxidic phase around 2.7 and 3.2 A˚, accompanied by the rise in the Pd-Pd first-shell signal at 2.5 A˚ together with the typical higher shell peaks of the fcc structure at around 3.7, 4.5, 5.1, 5.8, and 7.0 A˚. All of these features are significantly less intense than the corresponding features of the Pd metal foil (gray curve), reflecting a decrease in the average coordination number, as expected in the case of nanometer-sized Pd particles. Among reduced supported samples, a trend is observed indicating that the average Pd particle size on the three different supports is slightly different and follows the Pd(5.0)/Al2O3_H2 < Pd(5.0)/ Cp_H2 < Pd(5.0)/Cw_H2 order. The whole set of reduced samples has been subjected to dispersion measurements by CO chemisorption. For the reasons outlined in section 2.3, we assume dispersion D to be the CO/Pd ratio. Figure 7 shows the trend in dispersion at increasing Pd concentration on the three supports. Two main features are evident. (i) The dispersion tends to decrease with increasing Pd concentration, less significantly on Al2O3-supported samples and more on carbon-supported ones. (ii) The dispersion depends on the carrier and follows the sequence Al2O3 . Cp > Cw, in qualitative agreement with the EXAFS data (Figure 6b). The dependence of the Pd dispersion on the Pd concentration agrees qualitatively with the literature, where some papers report the (35) Dodgson, I. L.; Webster, D. E. In The Effect of Thermal Ageing on Metal Crystallite Growth and Catalytic Activity of Supported Platinum Group Metal Catalysts; Delmon, B., Jacobs, P. A., Poncelet, G., Eds.; Studies in Surface Science and Catalysis; Elsevier: Amsterdam, 1976; Vol. 1, pp 279-292. (36) Heal, G. R.; Mkayula, L. L. Carbon 1988, 26, 815–823.
11210 DOI: 10.1021/la1005117
data obtained on a given support by varying the Pd concentration.13,35,36 Conversely, the dependence of the Pd dispersion on the type of support is less evident in the literature because reliable systematic work on different supports is not readily available and the comparison among different papers is difficult owing to different preparations, reductions, and measurement conditions.36-38 In summary, under the adopted preparation conditions, Pd(5.0)/Cw, Pd(5.0)/Cp, and Pd(5.0)/Al2O3 samples consist of very similar supported Pd2þ phases (section 4 and Figure 5b, d). Once the as-precipitated Pd2þ-supported phase is reduced in an H2 atmosphere at 393 K, the role of the support (in terms of the supported phase/support interaction strength) becomes dominant in determining the dispersion of the Pd metal nanoparticles. The stronger interaction involving the Pd2þsupported phase with γ-Al2O3 is efficient in anchoring the particles on the surface, preventing effective sintering from occurring.16 Conversely, the weaker interaction that Pd has on carbon surfaces implies a particle sliding on the surface with consequent sintering. We conclude that the Pd-support interaction follows the Al2O3 . Cp > Cw trend. This model explains both the EXAFS results obtained on 5.0 wt % Pd (Figure 6b) and the different slopes obtained for the whole set of data by CO chemisorption (Figure 7).
6. Conclusions and Perspectives In this work, we have investigated the first stages of the preparation of Pd-supported catalysts following the depositionprecipitation method12,13 using Na2PdCl4 as a palladium precursor and Na2CO3 as a basic agent. In situ UV-vis spectroscopy allowed us to demonstrate that the main Pd2þ species present in the precipitation solutions are [PdCl2(H2O)2] and [PdCl(H2O)3]þ, whose concentration rapidly decreases for pH >4; the deposition process is over at pH 5. The precipitated phase is an amorphous Pd(OH)2 phase that is progressively converted into a crystalline PdO phase upon heating at increasing temperature. A complete structural characterization of the amorphous, unsupported Pd(OH)2 phase has been achieved by EXAFS. The structural properties and the reducibility of the as-precipitated Pd2þ phase on three different (37) Didillon, B.; Merlen, E.; Pages, T.; Uzio, D. In From Colloidal Particles to Supported Catalysts: A Comprehensive Study of Palladium Oxide Hydrosols Deposited on Alumina; Delmon, B., Jacobs, P. A., Maggi, R., Martens, J. A., Grange, P., Poncelet, G., Eds.; Studies in Surface Science and Catalysis; Elsevier: Amsterdam, 1998; Vol. 118, pp 41-54. (38) Polizzi, S.; Riello, P.; Balerna, A.; Benedetti, A. Phys. Chem. Chem. Phys. 2001, 3, 4614–4619.
Langmuir 2010, 26(13), 11204–11211
Agostini et al.
supports (alumina, active carbons from wood and from peat) have been successively investigated by means of EXAFS and TPR as a function of palladium loading (from 0.5 to 5.0 wt % Pd). In all as-precipitated samples, no appreciable amount of the Pd metal phase has been detected by EXAFS, meaning that the upper limit of the reduced metal phase is 1 to 2%. In these samples, the average Pd local environment is intermediate between that observed for amorphous Pd(OH)2 and crystalline PdO phases. A clear evolution of the EXAFS signal along the loading is observed, which saturated at 2.0 wt % Pd. The effect of the support is appreciable at the lowest loading where the highest shell signals exhibits an intensity that increases upon moving from Cp through Cw to Al2O3, reflecting the increasing order of the amorphous supported phase. The same trend holds for the maximum investigated loading, but the differences become much less evident. Once the samples are reduced in an H2 atmosphere to metal nanoparticles, two trends are observed: (i) the dispersion tends to decrease with increasing Pd concentration, less significantly on Al2O3-supported samples and more significantly on carbonsupported ones and (ii) the dispersion depends on the carrier following the sequence Al2O3 . Cp > Cw. This trend has been
Langmuir 2010, 26(13), 11204–11211
Article
explained in terms of a different Pd0-support interaction strength. The differences among the supports (Al2O3, Cp, and Cw), barely appreciable in the as-precipitated phase, become dramatic upon H2 reduction at 393 K under dry conditions. This fact implies that the reduction process deeply affects the characteristics of the supported metal nanoparticles. In future work we will investigate the effect of different reduction condition in comparing dry and wet reductions and using different reducing molecules. Acknowledgment. We thank F. Rotunno, who performed the time/pH-dependent UV-vis study reported in Figure 1a. We acknowledge the whole staff of the BM26 beamline at the ESRF and S. Nikitenko in particular for important and competent support during Pd K-edge EXAFS data collection. We also thank Massimo Graziani (Chimet S.p.A.) for the CO chemisorption measurements. Supporting Information Available: FT of all EXAFS curves plotted up to R = 10 A˚. This material is available free of charge via the Internet at http://pubs.acs.org.
DOI: 10.1021/la1005117
11211
