The sol-gel entrapment of noble metals in hybrid silicas: a molecular insight

Background Why are metal nanoparticles sol-gel entrapped in ORMOSIL so active and stable? In other words, why ORMOSIL-entrapped metal nanoparticles are more active and selective than many heterogenized counterparts, including silica-entrapped noble metals? Results Unveiling specific interactions between MNPs and the molecular structure of ORMOSIL, this work investigates subtle structural aspects through DRIFT spectroscopy. Conclusions The results point to interactions between entrapped Pd and Pt nanocrystallites with the organosilica sol-gel cages similar to those taking place in enzymes.


Introduction
Supported metal nanoparticles (MNPs) catalyze a number of reactions of enormous relevance in the petrochemical industry such as hydrogenation, epoxidation and monomer synthesis [1]. In the last two decades, the use of supported metal nanoparticles in synthetic organic chemistry has been extensively investigated in light of their ability to catalyze a range of chemical reactions [2], including asymmetric syntheses [3]. The aim was, and is, to apply heterogeneous catalysis to the synthesis of pharmaceutical and fine chemical products, for which the amount of effluent per tonne of product is orders of magnitude higher than that for a commodity chemical [4]. As a results of these efforts a number of new solid catalysts and green chemical processes are slowly being adopted by industry [5]; despite many remarkable research achievements like, as representative example, Cu ions immobilized on functionalized silica as recyclable and truly heterogeneous catalyst for the homocoupling of terminal alkynes in the presence of oxygen only [6].
In general, the atomic structure of the exposed surfaces of the active "naked" nanoparticles is made of plentiful unsaturated sites capable to adsorb and catalyze conversion of the reactants [7]. For example, Pd nanoparticles, which are well known for their catalytic activities, can easily aggregate to form Pd-black because of the very high surface energy of palladium [8]. Many efforts have been devoted to develop sinter-proof catalysts using, for example, pre-prepared colloidal metal nanoparticles with tuned size, shape and composition that are then "embedded" by porous support shells [9].
In principle, the heterogenization of MNPs should prevent the tendency of atoms of "naked" MNPs to aggregate into a bulk material due to their high surface energies, which results in rapid decrease in their intrinsic catalytic activity and selectivity over time [10]. Unfortunately, however, most heterogeneous catalysts reported in the literature, and especially palladium-based catalysts [11], act as reservoir for MNPs that are leached in solution where they catalyze reaction, but also rapidly aggregate resulting in spent catalyst of poor residual activity.
Sintering is caused by mobility of the metal particles on the support surfaces. Hence, to solve the sintering problem, the encapsulation of the metal nanoparticles within oxide architectures would minimise agglomeration and ensure catalyst recyclability. The validity of this approach was shown, for example, by McFarland and co-workers, comparing the catalytic performance of Pd particles deposited on the outer surface silica (Pd/ SiO 2 ), or encapsulated within the silica inner porosity (Pd@SiO 2 ) [12].
In this context, we have recently introduced a new catalyst series made of Pd and Pt nanocrystals encapsulated in one-step within the sol-gel cages of mesoporous organosilica xerogels. These materials are highly selective mediators in a number of important reactions including carbon-carbon coupling [13], debenzylation [14], highly selective hydrogenation of functionalized nitroarenes [15] and vegetable oils [16], and hydrosilylation of olefins [17]. Applications are not limited to this broad class of reactions and we are continuing to investigate new reactions and synthetic applications.
Work reported in this account investigates the structural origins of the enhanced performance of these new entrapped metal catalysts by Diffuse Reflectance Infrared Fourier Transform (DRIFT) spectroscopy, which is a powerful spectroscopic technique to investigate the molecular structure of these materials, revealing the subtle structural factors affecting their performance. Surface methods, indeed, are not suitable to investigate these materials due to the sol-gel encapsulation of the active species within the pores of the matrix.

Results and discussion
These novel catalytic materials typically contain Pd(0) or Pt(0) as the nanophase in hybrid alkyl-modified silica materials. For example, two catalysts made from methyl-triethoxysilane (MTEOS, Figure 1) doped with Pd(0) and Pt(0) were prepared according to the hydrolytic polycondensation process for the precursor MTEOS: Efficient entrapment of Pd or Pt nanoparticles is achieved by removing the released alcohol, using rotavapor distillation. Classical sol-gel encapsulation based on hydrolytic polycondensation of Si alkoxides releases large amounts of alcohol that rapidly reduce the Pd 2+ /Pt 2+ ions to bulk Pd(0)/Pt(0). The latter metal species are catalytically inactive, as only nanostructured MNPs are able to catalyze reactions, such as Pd nanoparticles mediating C-C coupling reactions [8].
Here instead, the resulting alcohol-free sol is doped with a Pd(II) or Pt(II) species, and undergoes further basic or acid catalyzed polycondensation to yield a mesoporous hydrogel that is dried under mild conditions to afford a xerogel doped with Pd 2+ /Pt 2+ . The latter material is chemically reduced under mild conditions to yield a mesostructured encapsulated catalyst in which the MNPs are physically and chemically stabilized, affording materials that can safely be used, without the need to exclude air or moisture. The TEM images clearly show the amorphous structure of the ORMOSIL embedding matrix structure in either catalyst. For SiliaCat Pt 0 (Figures 2A and Figure 2B), there is a considerable density of metallic Pt 0 nanocrystallites with average diameter of~16 nm, whereas for SiliaCat Pd 0 both the density and dimensions of the Pd 0 crystallites are smaller ( Figure 2C), pointing to a size of~8 nm.
The N 2 adsorption-desorption isotherms at 77 K of the two catalysts are shown and compared in Figure 3. For pore size evaluation we used the equation of Harkins and Jura [18]. Hence, the graphs next to the BET reports are the desorption dV/dD pore volume: Harkins and Jura plot with FAAS correction.
According to the IUPAC classification, the isotherms are type IV with hysteresis loops close to type H1, characteristic of capillary condensation in open cylindrical mesopores between spheroidal particles of fairly uniform array [19]. The mesopore size distribution using the BJH algorithm (that assumes a cylindrical pore shape) has a maximum population at~9 nm for SiliaCat Pt 0 and 6 nm for SiliaCat Pd 0 . The values of parameter c in the BET equation (99 for SiliaCat Pd 0 and 79.3 for SiliaCat Pt 0 ) and the t-plot analysis (Harkins and Jura) both indicate that there is no contribution from micropores in either sample. The textural parameters are summarized in Table 1.
Both structures have a very large surface area, exceeding 700 m 2 /g. SiliaCat Pd 0 , however, has significantly lower total pore volume. Accordingly, the average mesopore size is smaller for SiliaCat Pd 0 , but the size distribution is in either case peaked, pointing to negligible populations of smaller and larger pores.
The (DRIFT) spectra for both catalysts are shown in Figure 4.
The spectra of the two catalysts are very similar, with only slight changes in the relative intensities of some bands. The proposed band assignments are summarized in Table 2.
Given the high level of methylation of these matrices, it does not come as a surprise that both catalysts are hydrophobic. This is shown by the very weak ν(O-H) band and absence of the δ(HOH) mode, expected at~1640 cm -1 if any water molecules were adsorbed. The ν(O-H) band is therefore assigned to residual silanol (Si-OH) groups. The very low intensities of this band and of the νSi-O(H) or Si-O -(at~924 cm -1 ) bands show that the condensation reactions were extremely efficient. In addition, the high wavenumber of the ν(O-H) band maximum (~3500 cm -1 ) indicates that the very few silanol groups are not strongly interacting by hydrogen bonds [22].
The two bands at 2974 and 2914 cm -1 are assigned to the stretching modes (antisymmetric and symmetric, respectively) of the methyl groups bonded to a silicon atom. The corresponding deformation modes appear at 1410 and 1273 cm -1 , respectively. The absence of CH 2 related bands certifies that hydrolysis of methyltriethoxysilane was complete.
The two catalysts are thus hydrophobic and lipophilic. The visible bands in the spectra are assigned in Table 2, but the detailed assignment may only be confidently proposed after band deconvolution in two regions: i) 950-1250 cm -1 , which corresponds to the silica structural fingerprint, the ν as Si-O-Si mode, split in two observable components, with maxima at~1130 and~1036 cm -1 , cm -1 ; ii) 700-900 cm -1 , which apparently corresponds to two bands with maxima at 781 and 852 cm -1 , but in fact contains a number of overlapping components. The deconvolution of these spectral regions in a sum of Gaussian components was made by a non-linear least squares fitting method, previously described [23]. The deconvolution profiles are shown in Figure 5.
The results concerning the 950-1250 cm -1 region are summarized in Table 3, while those regarding the 700 and 900 cm -1 region are given in Table 4.
The ν as Si-O-Si mode was decomposed in longitudinal and transverse optical components (LO and TO, respectively) of different siloxane rings: six-member [(SiO) 6 ] and four-member [(SiO) 4 ]. The relative intensities of the components related to the four-member rings are much smaller indicating that the primary structure of the catalysts is predominantly formed by six-member siloxane rings: both ORMOSIL structures contain~80% of six-member siloxane rings ( Figure 6A) and~20% of four-member siloxane rings ( Figure 6B). These rings indeed are less tensioned, and more suitable to accommodate the extremely high content in methyl groups bonded to the Si atoms. An arrangement of four-member units would yield a much less three dimensional network, as only the terminal Si atoms are able to continue condensation. For undoped ORMOSILs produced from a mixture of trimethoxysilane (TMOS) and methyl-trimethoxysilane (MeTMOS), a similar result was obtained when the methylation degree was higher than 75% [24].
Accordingly, previous preliminary investigation of the SiliaCat Pd 0 structure by solid state NMR [25] has shown that the degree of cross-linking does not correlate with the catalytic activity.   Therefore, a predominantly six-member network is perfectly compatible with the porous structure found. Given the low content in four-member siloxane rings, the band at~550 cm -1 , which is usually assigned to a coupled mode in these units, must have some contribution from defective structures [26], possibly associated with Pd(II) and Pt(II) species that were not reduced.
Elimination of ethanol from the alcogel mixture contributes to the very large porosity observed in these hydrogel-derived catalysts. Indeed, the capillary tension at the solid-water cage interface is greatly reduced preventing collapse of the gel during drying [27]. Also, elimination of EtOH favours the Si alkoxide monomers hydrolysis and slows down condensation, so that rapid aggregation of the early sol particles is prevented and MTEOS can fully hydrolyse to CH 3 -Si(OH) 3 , which undergoes polycondensation in an open, amorphous structure made predominantly of 6-membered siloxane rings, entrapping Pd or Pt metallic nanophases with the well-known sol-gel stabilization of the nanoparticles.
The spectral region between 700 and 900 cm -1 is more difficult to decompose, given the number of overlapping components with close frequencies. The assignment in this region is not straightforward, because Si-C stretching modes are expected, with different frequencies depending on the local structure [28].
An interesting feature of these decompositions is that the ν s Si-O-Si band is in fact present, although not resolved. The similarity between the two decompositions confirms the resemblance between the molecular structure of the two catalysts.
The atomic dimension and electronegativity of the metal do not influence the main characteristics of the matrix structure: Pt belongs to the same group and to the following period as Pd (it is much larger and with higher electronegativity). Nevertheless, the ORMOSIL morphology and the dispersion of the metal nanophase are quite different. We emphasize herein the relevance of the support embedding structure in guiding and dictating the access of the reactants to the entrapped nanophase. In other words, encapsulation of the metal nanoparticles within the ORMOSIL structures results in materials that are remarkably more active than traditional catalysts; and this generally allows use of an ultralow amount of valued catalyst under conditions that are milder than those of state-of-the-art processes.
For instance, a 0.05 mol% amount of SiliaCat Pd(0) entrapped catalyst can be used to mediate the complete hydrogenation of a wide variety of vegetable oils under hydrogen balloon conditions without cis/ trans isomerisation; 16 whereas the best Pd catalyst previously known, made of Pd nanoparticles entrapped in the hexagonal porosity of SBA-15 mesoporous silica, mediates less selectively the same conversion at 100°C under 5 atm H 2 [29]. Similar findings have been reported for most of the catalytic processes catalyzed by the SiliaCat Pd 0 and SiliaCat Pt 0 mentioned above.

Conclusions
The DRIFT investigation of the molecular structure of ORMOSIL-entrapped metal nanoparticles suggests that the mechanism of action of nanoparticles encapsulated in organosilica is similar to that of enzymes. Once the metal nanoparticles are encapsulated and stabilized  within the sol-gel cages, it is the hydrophilic-lipophilic balance (HLB) of the matrix that dictates access and optimal catalysis with unprecedented performance, by promoting preferential adsorption of the lipohilic functional group moieties in reacting substrates adsorbed at the surface of the Pd and Pt nanoparticles in reactions as different as hydrogenation of fats, hydrogenation alkenes and nitroarenes, hydrosilylation of olefins and C-C coupling. The above mentioned reactions concern very large sectors of the chemical industry, many of which continue to use obsolete catalysts such as Ni Raney (hydrogenation of fats), or Pt/C (hydrogenation of olefins). Sol-gel entrapped metal nanophases will give an immense contribute to simplify those processes, whereas new catalysts are being developed capable to target other relevant reactions.

Materials synthesis
In a typical preparation, a mixture of methyltriethoxysilane (27 g, 30 mL, 151.4 mmol) and 10 mL of 0.042 M HCl was stirred for 15 minutes. The resulting solution was concentrated with a rotavapor under reduced pressure at 30°C   until complete ethanol removal (about 15 minutes). The alcohol-free sol thereby obtained was added with K 2 PdCl 4 (from 0.004 to 0.02 equivalent) dissolved in H 2 O (from 5 to 10 mL) and 60 mL acetonitrile. This mixture was added with 1 M NaOH (from 0.023 to 0.053 equivalent) to favour gelation that indeed rapidly occurred. The resulting transparent gel was left to dry in air for 4 days after which the xerogel was reduced at room temperature under inert conditions with a solution of sodium triacetoxyborohydride (Pd:Na(AcO) 3 BH = 1:6 molar ratio) in 80 mL THF, washed with THF and H 2 O and dried in air to afford a SiliaCat Pd 0 catalyst. The metal load in each catalyst was measured using the CAMECA SX100 instrument equipped with EPMA analyzer, a fully qualitative and quantitative method of non-destructive elemental analysis of micron-sized volumes at the surface of materials, with sensitivity at ppm level.

DRIFT analysis
Fourier transform spectroscopy in diffuse reflectance mode was performed in a Mattson RS1 FTIR spectrometer with a Graseby Specac Selector, in the range 400-4000 cm -1 , at 4 cm -1 resolution. The analyses were carried out at ambient temperature and pressure, using a powder catalyst sample as received by the catalysts manufacturer (SiliCycle, Inc.). No further treatment of the catalyst was undertaken prior to measurement.

TEM analysis
The TEM pictures were obtained in an electron microscope Hitachi H-8100, operated at 200 kV, with a LaB6 filament. The samples were dispersed in ethanol and then dropped onto a Formvar®-coated Cu grid and left to evaporate.

BET analysis
Nitrogen adsorption and desorption isotherms at 77 K were measured using a an ASAP 2020 system from Micromeritics, analyzing the resulting data with the Tristar 3000 software (version 4,01). The desorption branch was used to calculate the pore size distribution.