Synthesis of heterogeneous catalysts with well shaped platinum particles to control reaction selectivity

Abstract

Colloidal and sol-gel procedures have been used to prepare heterogeneous catalysts consisting of platinum metal particles with narrow size distributions and well defined shapes dispersed on high-surface-area silica supports. The overall procedure was developed in three stages. First, tetrahedral and cubic colloidal metal particles were prepared in solution by using a procedure derived from that reported by El-Sayed and coworkers [Ahmadi TS, Wang ZL, Green TC, Henglein A, El-Sayed MA (1996) Science 272:1924–1926]. This method allowed size and shape to be controlled independently. Next, the colloidal particles were dispersed onto high-surface-area solids. Three approaches were attempted: (i) in situ reduction of the colloidal mixture in the presence of the support, (ii) in situ sol-gel synthesis of the support in the presence of the colloidal particles, and (iii) direct impregnation of the particles onto the support. Finally, the resulting catalysts were activated and tested for the promotion of carbon–carbon double-bond cis-trans isomerization reactions in olefins. Our results indicate that the selectivity of the reaction may be controlled by using supported catalysts with appropriate metal particle shapes.

On the basis of their kinetic behavior, catalytic reactions are often classified as either mild or demanding (1–3). Demanding reactions—such as the oxidation of CO, NO, or hydrocarbons; the synthesis of ammonia; and most oil processing conversions—usually require high temperatures and pressures, and involve small concentrations of intermediates similar to those identified under vacuum. The performance of these reactions often depends strongly on the structure of the catalyst used (4, 5). In contrast, mild reactions—in particular, hydrogenations and isomerizations of unsaturated hydrocarbons—take place under less-demanding temperature and pressure conditions. Mild reactions have historically been considered structure-insensitive (6–8), but that conclusion has been drawn from studies on reactivity vs. metal dispersion that used ill-defined supported catalysts (9, 10) and has been questioned by more recent studies using better catalytic models (11). For instance, both experimental (12–14) and theoretical (15) studies on the selective catalytic hydrogenation of CO bonds in unsaturated aldehydes have suggested that such reactions may be promoted by close-packed (111) surfaces. In another example, the dehydrogenation of cyclohexene was found to be faster on Pt (111) than on Pt (100) single-crystal surfaces (16). Our recent surface-science investigations on the isomerization of unsaturated olefins (17–19) strongly suggest that selectivity toward the formation of the cis isomer may be favored by Pt (111) facets. Additional surface-science reports on the conversion of alkyl and alkene adsorbates under vacuum conditions (20–25), as well as studies with more realistic model systems (26, 27), point to a potential structure sensitivity in the conversion of other olefins and unsaturated hydrocarbons.

These results not only suggest that mild reactions are structure-sensitive, but also highlight the separate roles that particle size and particle shape may play in defining activity and selectivity in those catalytic processes. Accordingly, the independent control of particle size and shape during catalyst manufacture may offer avenues for the design of highly selective catalysts. The self-assembly and nano technologies developed during the last few years may now place such catalysts within reach. Indeed, several colloidal (28–30) and dendrimer (31) methodologies are already available for making metal particles with well defined sizes in solution. The control of particle shape is more challenging but has also been achieved recently, largely by controlling the concentration ratio of the capping polymer material to the metal cation used during synthesis (32–35). Tetrahedral, cubic, irregular-prismatic, icosahedral, and cubo-octahedral particles have been produced in this way. Alternative routes have also been proposed for controlling both size and shape in metal colloidal particles (36–38).

The next challenge is to extend the liquid-phase chemistry mentioned above to the preparation of well defined supported catalysts. This requires the resolution of a number of issues, including the proper dispersion of the colloidal particles throughout the pores of suitable high-surface-area supports and the removal of the organic material used to stabilize the particles in solution to expose the clean metal surface. A particularly important issue in connection with the latter step is the need to avoid the loss of the well defined size and shape of the particles or the occurrence of particle agglomeration (sintering). Once proper particle dispersion and activation are accomplished, the performance of the resulting catalysts needs to be tested to identify clear correlations between activity and/or selectivity and the size and/or shape of the particles. We in our laboratory have been working to resolve these issues and here report on key results indicating the feasibility of this approach. Specifically, the selective promotion of trans-to-cis carbon–carbon double-bond isomerizations in olefins was achieved by using catalysts with tetrahedral Pt particles exposing high fractions of (111) facets. The loss of those facets upon high-temperature treatment leads to a reversal to the thermodynamically expected preference for cis-to-trans conversion.

Results

Synthesis of Colloidal Platinum Particles.

Supported platinum catalysts with well defined shapes were prepared and tested by using the strategy shown in Scheme 1. The initial step focused on the synthesis of colloidal platinum particles with different, well defined sizes and shapes. The starting point for this synthetic effort was the protocol developed by El-Sayed and coworkers (32, 39, 40), who identified a shape-controlled growth mechanism based on differences between the rate of the catalytic reduction process of Pt²⁺ on (111) vs. (100) facets, a competition between the Pt²⁺ reduction and the capping process on the different nanoparticle surfaces, and a concentration-dependent buffer action of the polymer itself. These authors have reported the selective production of cubic, tetrahedral, and truncated octahedral platinum nanoparticles by changing the ratio of the concentration of the capping polymer to that of the platinum cation in the solution used for the reductive synthesis of the colloidal nanoparticles.

Scheme 1.

Methodology used for the preparation of supported platinum catalysts with well defined particle shapes and sizes.

We conducted a series of experiments to determine the optimum parameters to be used in preparing our tetrahedral and cubic Pt particles. The best tetrahedral Pt particles were obtained by using H₂PtCl₆ and poly(N-vinyl-2-pyrrolidone) (PVP, average M_w = 360,000), and the best cubic particles employed K₂PtCl₄ and sodium polyacrylate (SPA, average M_w = 2,100). The specifics of this synthetic approach are provided in the supporting information (SI) Materials and Methods. Our own experience in these studies indicated that, indeed, the most critical factor in controlling both particle shape and size is the concentration ratio between the transition metal precursor and the capping agent. Use of low concentrations of the polymer led to narrower size distributions, within 10% of the mean, but not necessarily to well defined particle shapes. The concentration ratio is more critical with H₂PtCl₆, but smaller amounts of the polymer are required for the preparation of tetrahedral Pt nanoparticles. Also, in the case of the cubic particles, using an excess of the polymer leads instead to the growth of octahedral particles. The lighter the capping polymer, the higher the concentration needed: 10 times higher concentration ratios of PVP of average M_w = 40,000 than of PVP of average M_w = 360,000 are needed to prepare similar tetrahedral Pt nanoparticles. In our experiments, the optimum ratios for making tetrahedral and cubic particles were determined to be [H₂PtCl₆]:[PVP, average M_w = 360,000] = 1:0.0005 and [K₂PtCl₄]:[SPA, average M_w = 2,100] = 1:0.75, respectively. These ratios work in only a limited range of precursor concentrations—somewhere between 8 × 10⁻⁵ and 4 × 10⁻⁴ M—but within that range the concentration can be used to control the size of the particles (between ≈3 and 10 nm). Other factors, such as pH and reducing time, also must be carefully controlled to obtain high-quality shapes and highly monodispersed nanoparticles. Slower reduction rates appear to lead to narrower size distributions.

Fig. 1 shows representative transmission electron microscopy (TEM) images of Pt colloidal particles with cubic, tetrahedral, and star-like (tetrapod) shapes obtained by this colloidal synthetic procedure. The average sizes of tetrahedral and cubic Pt nanoparticles from TEM images were estimated at 4.8 ± 0.3 and 5.0 ± 0.3 nm, respectively, and the purity of the shapes (fraction of tetrahedral or cubic particles over total) at 83% and 87%, respectively. A more detailed analysis is provided in Fig. S2. The nature of the exposed surfaces—(111) and (100) facets for the tetrahedral and cubic particles, respectively—was confirmed by electron diffraction (see Fig. S3).

Fig. 1.

TEM images of different platinum colloidal particles obtained by tuning the conditions used for their assembly. Both particle size and shape can be controlled in this way. Note the high-quality cubic (A), tetrahedral (B), and star-like (tetrapod) (C) shapes obtained in the three cases reported here [with preferential exposed (100) and (111) facets in the first two cases].

We placed particular emphasis on lowering the concentration of the capping polymers as much as possible, to minimize the problems associated with their removal after dispersion of the particles on the high-surface-area solids (see below). The final ratios of polymer to metal precursor used here are significantly lower than those reported in the literature [i.e., 1:0.0005 vs. 1:0.035 for the tetrahedral particles (41)]. In further addressing this issue, it will be important to test other capping polymers.

Particle Dispersion on High-Surface-Area Supports.

The procedures developed for the deposition and activation of the platinum nanoparticles onto high-surface-area solid supports are perhaps the most critical aspect of the preparation of our catalysts because the stabilizing capping polymers must be removed with no significant loss of particle size or shape. Three basic approaches were explored, as indicated in Scheme 1. The first was to grow the Pt particles in situ in the presence of a porous xerogel by carrying the reduction step after the addition of the high-surface-area support. Unfortunately, all our attempts to prepare catalytic samples by following this procedure failed because significant metal particle aggregation was observed and because particle penetration inside the porous structure of the support was quite limited (Fig. 2A).

Fig. 2.

TEM of 0.5 wt % tetrahedral colloidal Pt particles dispersed on porous silica supports, prepared by three different procedures. (A) Precipitation in situ by reduction in the presence of a high-surface-area silica xerogel powder. (B) Growing of the silica xerogel in situ using a sol-gel method. (C) Impregnation of a porous silica xerogel powder with the Pt colloidal particles, followed by filtration. Of the three methods, C appears to be the best for obtaining well dispersed catalysts.

Our second approach was to synthesize the solid support in situ in the presence of the colloidal particles, to encapsulate the latter within the pores of the former during the growth process (42). The results obtained from preliminary experiments using this methodology are encouraging. In one test, ammonia and ammonium fluoride were first mixed with a colloidal solution obtained by leaving the metal precursor plus capping polymer mixture in the dark for several hours and then adding an appropriate amount of tetraethyl orthosilicate (TEOS) to initiate the gelling of a silica solid (43, 44). The resulting xerogels were dried under vacuum without aging until gray powders were obtained. A TEM image of the resulting catalyst, with its tetrahedral particles embedded in the in situ-grown sol-gel silica, is shown in Fig. 2B. Sol-gel chemistry is typically carried out in ethanol; however, because ethanol interferes with the stability of the colloidal particles, water was chosen here as the solvent. We also learned that careful control of the concentrations and total amounts of all of the chemicals is necessary to match the colloidal and gelling chemistries. More studies are needed in this area. A similar approach has been used successfully by Somorjai and coworkers (37, 45) to incorporate platinum nanoparticles synthesized by alcohol reduction into mesoporous SBA-15 silica.

The main shortcoming of the second procedure is that the resulting xerogels typically display inferior pore size and shape distributions. We estimate that in our synthesis >50% of the pore structure shrunk during the drying process. Consequently, we explored a third synthetic route in which the Pt particles were dispersed within the porous material by simple impregnation, followed by either evaporation or filtration (46, 47). The resulting dry catalysts displayed a gray color, indicating metal deposition, and showed good dispersion in the TEM images (Fig. 2C). This approach was also briefly tested for the dispersion of Pt particles on a homemade SBA-15 mesoporous sample. That solid is crystalline, with one-dimensional pores of ≈5–7 nm in diameter (as indicated by TEM), and only allows the incorporation of nanoparticles with sizes of up to ≈5 nm; the larger particles were deposited only on the outside surfaces (see Fig. S4).

Catalyst Activation.

After preparation, the catalysts were washed with either water or alcohol, dried, and annealed and/or calcined in air for 1 h to produce the desired dispersed metal catalysts. As mentioned before, gentle calcination treatments are needed to minimize the potential changes in size or shape of the Pt particles. Interestingly, past TEM experiments with unsupported platinum colloidal particles have indicated that although heating to 550 K is sufficient to remove the capping polymers, the metal particles retain their shape even after annealing at 625 K (48). In the present study, the thermal chemistry of Pt colloidal particles deposited on xerogel supports was explored by using transmission IR absorption spectroscopy, X-ray photoelectron spectroscopy (XPS), and TEM. Representative IR spectra are shown in Fig. 3. Those data, which correspond to cubic Pt particles dispersed on silica gel, indicate that most (>95%) of the SPA capping polymer can be removed by heating to <625 K, even after deposition of the colloidal particles on the solid; the key vibrational peak for the asymmetric OCO stretch, ν_a(OCO), of SPA at 1,558 cm⁻¹ disappears completely by 675 K. However, some polymer decomposition does occur, as manifested by the persistence of a few of the CH stretching bands in the IR traces (data not shown), and subsequent oxidation–reduction (OR) cycles are needed to remove the residual carbonaceous deposits (1 h for each half cycle under 1 ml/s flow of either O₂ or H₂ at the same temperature of calcination). This chemistry is confirmed by XPS, which also indicates the extent to which the hydrogen treatment is able to reduce the Pt particles (Fig. 4): the Pt 4f_7/2 peak at 75.1-eV binding energy due to the oxidized platinum seen in the fresh sample is significantly diminished (from 30% to 10% of the total signal) after the reduction treatment.

Fig. 3.

Transmission IR spectra for 0.5 wt % cubic colloidal Pt particles dispersed on a xerogel silica solid at different stages of deposition and activation. Reference spectra for the SPA capping polymer and the silica support are provided to aid in the peak assignment. (Left) Evidence for the deposition of the colloidal particles on the silica; (Right) Evidence for the successful removal of all hydrocarbon residues after calcination and reduction at 625 K.

Fig. 4.

Pt 4f XPS data for samples prepared by deposition of 0.5 wt % tetrahedral colloidal Pt particles on a silica xerogel. Some Pt signal is lost upon calcination (middle trace), but regained with almost complete reduction to the metal state after reduction at 625 K.

The evolution of the shape of the supported particles as a function of the severity of the cleaning procedure was also tested by TEM. Fig. 5 shows a typical sequence of TEM images for 1.0 wt % tetrahedral Pt nanoparticles deposited by impregnation on a silica xerogel, for the fresh sample and after calcination to various temperatures between 375 and 575 K. Minor shape changes already take place after calcination at 375 K, but not until 575 K do the particles lose their original shape and become rounded; thus, the calcination temperature must be kept below 575 K to preserve the desired particle shape. Lowering the concentration of the polymer used during the colloidal preparation step, as mentioned in the previous section, could be tested as a way to minimize the severity of the conditions needed to clean the supported particles. Displacement of the original capping polymers by other easier-to-remove ligands once the particles have been grown may also help minimize this contamination problem.

Fig. 5.

TEM of 1.0 wt % tetrahedral colloidal Pt particles dispersed on a porous silica xerogel, freshly prepared (A) and after calcination to 375 K (B), 425 K (C), 475 K (D), 525 K (E), and 575 K (F). The platinum particles retain their initial tetrahedral shape after the low-temperature treatments but are clearly transformed into larger and rounder shapes upon calcination at 575 K. These high temperatures are, therefore, to be avoided in the catalyst activation stage.

Catalytic Performance.

The catalytic performance of the materials made by using the procedures described above was evaluated. Specifically, the selectivity for the isomerization of olefins was probed by measuring the kinetics of isomerization of both cis- and trans-2-butene in the presence of a small amount of hydrogen. The ability to control selectivity in these reactions may have significant practical implications because, for instance, the trans fatty acids that are produced during the partial hydrogenation of natural oils to edible fats are believed to have serious negative health effects (49, 50). Although the conversion of alkenes by transition metals is one of the oldest and most studied systems in catalysis (51–53), key issues such as the dependence of activity on structure remain unresolved (54). As mentioned in our Introduction, results from recent surface-science work in our laboratory have indicated that, at least on Pt (111) single-crystal surfaces, the isomerization of trans-2-butene to its cis isomer is easier than the opposite cis-to-trans conversion (17–19). A kinetic study on the catalytic conversion of butenes on single-crystal surfaces (55) has also suggested different selectivities on Pt (111) vs. Pt (100) surfaces.

Fig. 6 displays the ratios of the rates for trans-to-cis vs. cis-to-trans conversions obtained with both tetrahedral and cubic supported particles as a function of the temperature used for calcination during the catalyst pretreatment. For the cubic particles, comparable rates are seen for both isomerizations—the conversion of cis-2-butene to its trans isomer and the reverse trans-to-cis transformation—with all catalysts, regardless of the calcination temperature used. With the tetrahedral particles, on the other hand, not only are the rates of reaction quite different, but their relative values change dramatically as the calcination temperature is increased. On the catalysts treated at low temperatures (<500 K), the trans-to-cis conversion occurs approximately twice as fast as the opposite cis-to-trans conversion. In contrast, the trans-to-cis reaction dominates after catalyst calcination at temperatures >500 K, being as much as three to four times faster than the cis-to-trans conversion. These changes correlate well with the changes seen in the TEM images in Fig. 5 and indicate that the preference for trans-to-cis conversion is lost once the particles are calcined to temperatures high enough to affect their shape and reduce the fraction of (111) facets exposed. The inference is that it is the (111) facets of the platinum tetrahedral particles that promote the nonthermodynamic isomerization of trans olefins to their cis counterparts, as previously hypothesized.

Fig. 6.

Kinetic data for the conversion of cis- and trans-2-butenes with hydrogen on catalysts prepared by impregnation of 1.0 wt % Pt nanoparticles on silica xerogel as a function of calcination temperature. Shown are the ratios of the initial rates for the conversion of the trans to the cis isomer vs. those of the cis-to-trans isomerization for catalysts made with tetrahedral and with cubic particles. The rates of both cis-to-trans and trans-to-cis conversions are comparable for the cubic particles over the calcination temperature range studied here, but with the tetrahedral particles, a dramatic switch in selectivity is seen at ≈500 K.

Finally, the integrity of the catalysts after their exposure to the catalytic conditions was checked with TEM. It is well known that exposure of supported metal particles to reactive environments can induce surface reconstruction (56) and could in this case reshape the original particles, reduce the high fraction of (111) facets exposed, and remove their selectivity. However, this does not seem to be the case here. The similarity of the TEM images obtained before and after the catalytic reaction indicates minimal changes in the shape of the particles (Fig. 7). It appears that, in this case, the catalytic conditions are sufficiently mild to avoid major surface reconstruction on the platinum-supported particles.

Fig. 7.

TEM images from a catalyst prepared by using tetrahedral colloidal particles, before and after being used for the catalytic conversion of 2-butenes in experiments similar to those reported in Fig. 6. The initial catalyst corresponds to a 1.0 wt % load of Pt nanoparticles on a silica xerogel after calcination to 475 K. The similarity of the two images suggests minimal deterioration of the shape of the supported Pt particles upon their exposure to the catalytic conditions.

Discussion

In this study, we have shown that colloidal chemistry can be used to prepare catalysts consisting of dispersed metal particles with well defined sizes and shapes. Some operational optimization in terms of catalyst deposition and pretreatment may still be needed, but the results summarized above clearly prove the viability of this approach. Perhaps more interestingly, it was also found that particles with different shapes may promote different reactions selectively. The example provided here refers to double-bond isomerizations, but similar behavior is conceivable with other important catalytic reactions. The one caveat is that, because of the limited stability of the deposited particles, only mild reactions can be expected to benefit from the use of these shape-controlled catalysts. Nevertheless, this includes a large number of processes of potential interest to industry, including the hydrogenation and isomerization of many unsaturated hydrocarbons (olefins, aldehydes, ketones, organic acids, imines, etc.), as well as electrocatalysis such as that involved in fuel cells. It is also worth mentioning that these colloidal synthetic methods, in conjunction with other self-assembly procedures, offer an array of options for the preparation of specific catalysts. For example, the use of crystalline mesoporous materials such as SBA-15 or MCM-21 could add shape selectivity, and bimetallic colloidal particles could be used to control the nature of the active sites in these catalysts.

The concept of using colloidal methods to control size and shape in supported catalysts has been explored by others (57). The usefulness of the resulting catalysts has also been tested, mostly for electrocatalysis (35, 58–60), but in a few instances also for gas-phase catalytic conversions (61–64). In fact, recent designs of shape-selective catalysts have been implemented industrially for reforming and hydrogen peroxide synthesis (65). However, to the best of our knowledge, ours is the first systematic and comprehensive study of the evolution of the shape of the particles throughout the various preparation and activation steps needed to make and use these supported heterogeneous catalysts. It is also, we believe, the first time that particle shape has been shown to control selectivity in a catalytic process.

Materials and Methods

The following chemicals were used in the synthesis of the nanoparticles: dihydrogen hexachloroplatinate hexahydrate (H₂PtCl₆·6H₂O, 99.9% purity; Alfa Aesar); potassium tetrachloroplatinate (K₂PtCl₄, 99.99% purity), poly(N-vinyl-2-pyrrolidone) (PVP, [CH₂CH(NC₃H₆CO)]_n, average M_w = 40,000 and 360,000), sodium polyacrylate (SPA, [CH₂CH(CO₂Na)]_n, average M_w = 2,100), 2-methyl-2-propanethiol [(CH₃)₃CSH, 99% purity], toluene (C₆H₅CH₃, 99.5% purity), 1,3,5-trimethylbenzene [(C₆H₃(CH₃)₃, 99.2% purity], all from Aldrich; Pluronic P123 triblock copolymer (BASF); and ammonium hydroxide (30% ammonia), ammonium fluoride (NH₄F, 98% purity), and tetraethyl orthosilicate (TEOS, 98% purity), all from Acros. In addition, the following gases were used: argon (ultrahigh purity), hydrogen (ultrahigh purity), oxygen (ultrahigh purity), air (high purity), and nitrogen (high purity), all from Airgas; and cis-2-butene (>95% purity) and trans-2-butene (>95% purity), from Mathson Tri-Gas. All compounds were used as received.

UV-visible (UV-vis) absorption spectra between 190 and 1,000 nm were recorded by using a Varian Cary 50 UV-vis spectrometer. Transmission IR absorption spectra of the catalysts (pressed into pellets) were acquired with a Bruker Tensor 27 and a deuterated triglycine sulfate (DTGS) detector. For the colloidal samples, a commercial ZnSe attenuated total reflection (ATR) accessory (PIKE Technologies) was used to minimize the contribution from background water. A spectral resolution of 4 cm⁻¹ was used in all cases.

The TEM experiments were performed on a Philips TECNAI 12 TEM (20–120 kV accelerating voltage). The colloidal samples were deposited directly onto Cu grids covered with a holey carbon support film (Ted Pella), and the catalysts were dispersed in Milli-Q water by ultrasonication and deposited from the resulting suspension onto the Cu grids.

The XPS data were collected in an ultrahigh vacuum chamber equipped with an Al K_α (hν = 1,486.6 eV) x-ray excitation source and a Leybold EA11 electron energy analyzer with multichannel detection (66). A constant band pass energy of 100.8 eV was used, corresponding to a spectral resolution of ≈2.0 eV.

The catalytic activity of the Pt-supported catalysts was investigated by using a 150-ml stainless-loop batch reactor (67, 68). The composition of the gas mixture was determined periodically by gas chromatography, using a 23% SP-1700 on 80/100 Chromosorb PAW column and a flame ionization detector. Additional details on the operation of the batch reactor are provided in SI Materials and Methods.