The wetting characteristics were assessed by measuring the equilibrium contact angles of the silica sol-gel on flat PVP-coated gold and amidine-functionalized PS substrates . The prehydrolyzed silica sol-gel solution has a very low contact angle of ≈10° and 30° on a gold surface and PS surface, respectively. These contact angles and the sizes of the NPs and PS colloids were used as input for numerical calculations of the equilibrium solid-liquid-vapor contact line shapes that form a stable sol-gel interface between the NP and the PS colloid for different volumes of air trapped around NPs during the infiltration step . Specifically, the calculations allowed accessing the maximum possible air volume that gives a stable silica-air interface . The upper limit is an air volume of ≈1.6 × 102 nm3, yielding a minimum theoretical value for NP immersion into the sol-gel solution of at least 30% of its surface . Full NP immersion occurs when no air is trapped around the NP upon infiltration, and all the intermediate immersion values are possible, as a function of the volume of air trapped around the individual NPs. Upon calcination, the immersed part of the particle will become embedded into the solidified silica matrix , and hence, the NPs in the RCT catalysts are expected to be significantly entrenched, with a broad distribution of embedding values.To verify the theoretical predictions, hydroponic channel the embedding of more than 200 individual Pd4Au96 NPs in silica-supported RCT catalyst was quantified using dual-axis electron tomography .
Previously, electron tomography has successfully been applied to obtain the distribution, size, and shape of metal particles in catalytic materials. Yet, an important obstacle in using electron tomography to investigate metal-support interfaces is so-called missing wedge artifacts. These artifacts stem from the reduced range of orientations in which the sample can be imaged. Ideally, one would rotate the sample over the full 180° range, but in practice, only tilt angles in the range of 140° can be probed. The missing range of angles, called the missing wedge, causes artifacts in the final 3D reconstruction. These artifacts are particularly severe around the metal particle due to diffraction contrast, thereby obscuring the metal-support interface and the local NP environment. Dual-axis tomography allows a significant reduction of missing wedge artifacts and therefore the assessment of the NP-support interfaces . As the direction of the artifacts around the metal particles depends on their orientation with respect to the tilt axis, combining the information of two tilt series recorded over perpendicular axes, renders a more complete picture of the local NP environment. The analysis procedure to achieve a quantitative and statistically relevant characterization of the local environment of the embedded NPs consisted of five steps : i) acquisition and alignment of the tilt series, ii) separate reconstruction of both data sets, iii) recombination of the reconstructions, iv) segmentation into particle, support and pore, and v) quantification of the embedding.
Full details are provided in the Experimental Section and Supporting Information. Both tilt series can be viewed in Movies S1 and S2, Supporting Information. No beam damage during the tomography acquisition occurred, which is crucial when combining the information from both tilt series into one . Each series was individually processed by aligning the images using 12 fiducial markers in the 141 tilt images per data set and reconstructing the 2 tilt series separately . Subsequently, the two 3D models were combined into one reconstruction using the known positions of 8 overlapping fiducial markers in both data sets . The combined reconstruction at full resolution consisted of more than 3 × 1010 voxels, making further analysis computationally challenging. Therefore, regions of interest of 150 × 150 × 150 voxels around the metal particles were carved out, allowing a reduction in the number of voxels by 3 orders of magnitude. To identify the ROIs, the images were downsized twice and a simple thresholding algorithm was applied to locate the NPs’ positions. The coordinates obtained from the lower-resolution images were mapped back to the high-resolution images and the ROIs were extracted around the estimated particles centers . Successively, each ROI was segmented, that is, every voxel was classified as particle, support, or void . The segmentation procedure consisted of i) denoising the raw images, by employing a total variation denoising algorithm; ii) identifying the voxels belonging to the particle, by applying contrast thresholding; iii) identifying the voxels belonging to the support, by employing a watershed algorithm . From the segmented ROI, the particle size and the fraction of particle surface exposed to the pore were calculated. After visual inspection, accurate quantification of the local environment of 204 particles was confirmed . The quantitative analysis revealed that the NPs were indeed largely embedded in the silica support .
The accuracy of our quantification procedure is supported by the excellent agreement between the average particle size as obtained via automatic segmentation and the manually measured particle size distribution based on a set of 2D images taken at different locations in the sample . Both approaches resulted in an average particle size of 7.4 nm. The relative amount of NPs surface area exposed to the macropores as a function of particle size for 204 particles present in the catalysts fragment, showed that 200 particles were more than 50% embedded in the support. The average and median exposed surface areas were 15% and 12%, respectively. However, the distribution of the exposed surface per particle in the right panel reveals that a substantial number of particles have less than 10% of the surface exposed. Furthermore, there are only 2 particles that have 60% of their surface exposed, which is the minimum embedding observed. As an example, two NPs, one moderately embedded and the other significantly embedded , are shown in Figure 3c, where the green and yellow borders indicate the exposed and embedded NP surfaces, respectively. Overall, the distribution in exposed surface area is broad, and no obvious correlation between particle size and the amount of exposed NP surface was found. Both, the broad distribution of the particles exposed surfaces and the smallest observed embedding of 40% are in excellent agreement with the numerical modeling results shown in Figure 1g.The Au96Pd4 NPs exhibited a characteristic penta-twinned crystal structure, irrespective of their size and degree of embedding . Three representative particles of 9, 11, and 14 nm in diameter at different tilt angles are shown in Figure S9, Supporting Information. The diffraction contrast in the 2D images acquired at different tilt angles clearly shows the fivefold symmetry in the crystal structure of the particles. It is likely that this stems from the citrate ligands used in the NP synthesis, hydroponic dutch buckets which are known to give rise to penta-twinned crystal structures in Au NPs. The fivefold symmetry of the free-standing AuNPs remained intact after introducing a small amount of Pd, attaching the NPs to PS and thermally treating the RCT catalyst at 500 °C. This is in line with previous work in which the crystal structure of free-standing AuNPs was studied as a function of temperature. The decahedral shape of 2–10 nm AuNPs was stable up to 550–600 °C, above which surface roughening and ultimately melting occurred. Another indication that the penta-twinned crystal structure was preserved during the RCT synthesis, is the random orientation of the facets with respect to the pore, which is expected, based on the random attachment of the metal NPs to the PS colloids. No evidence for differences in faceting of the exposed and embedded side of NPs was found.The electron tomography data provided quantitative information on the embedding of the metal NPs in the silica matrix and insight into the particle shape, but raised the question of whether the heavily embedded interfaces of the NPs were still chemically accessible. To assess this, a novel and complementary characterization approach was employed, relying on epitaxial overgrowth of the embedded NPs with a second metal, in this case, silver. Ag was chosen as it is known to grow epitaxially on Au due to the negligible lattice mismatch between the two metals . Hence, the epitaxial growth of Ag on Au enabled visualization of the crystal orientation of the Au surface facets exposed to the macropores. Furthermore, Ag has a significantly lower Z-contrast compared to Au, ensuring a clear difference between the Ag-shell and AuNPs in the electron microscopy visualization.
The high-resolution high-angle annular dark-field scanning transmission electron microscopy image shows a smooth, epitaxial spherical Ag-shell around the free-standing AuNPs with roughly the same 1.5 nm shell thickness at all sides, whereas clear anisotropic Ag-shell growth was observed on the embedded AuNPs . The difference in Ag-shell morphology is also evident from the UV–vis spectra . The localized surface plasmon resonance of the free-standing AuNPs blue-shifted from 517 to 500 nm and increased in intensity upon Ag growth, which is in line with previously reported theoretical and experimental work, where the increase in intensity of the LSPR peak can be ascribed to both the stronger plasmonic properties of Ag andincrease in particle volume. Contrarily, the LSPR peak of the embedded AuNPs split up into 2 separate peaks upon Ag-shell growth: a transverse and longitudinal LSPR peak, at high and low energy, respectively, similar to the plasmonic properties of Ag nanorods. The red-shift of the longitudinal LSPR peak from 521 to 565 nm indicates the growth of a more elongated particle, and hence matches well with the particle Janus-like shape observed in the HAADF-STEM images. In both cases, the shell thickness was controlled via the AgNO3 concentration in the reaction mixture. Although the Ag-shell preferentially grew on the NP surface exposed to the macropores, Energy-dispersive X-ray spectroscopy analysis revealed that Ag also grew on the embedded side of the NPs. This is further confirmed by the high-resolution images in Figure 5, where a continuous shell of approximately three atomic Ag layers was observed on the embedded side of the NPs . Furthermore, the crystal structure of the Ag-shell provided insight into the type and number of surface facets of the AuNPs facing the macropores. By “freezing in” the surface structure of the NPs in the Ag-shell, imaging at high resolution was possible without beam-induced restructuring. The high-resolution images in Figure 5 reveal that depending on the type of facets exposed to the pore, either single-crystalline or twinned Ag-shell formed. In line with our findings shown in Figure 4, no evidence for preferential orientation of specific surface facets towards the macropores was observed. The embedded metal surface area could be available to the reactants through the 0.7 nm gap, but its accessibility might also be related to the porosity of the silica support itself. As the micropores cannot easily be characterized using electron microscopy, nitrogen physisorption was employed to investigate the effect of the latter. The physisorption data confirm that the silica support is indeed microporous , thereby allowing mass-transport to the embedded part of the AuNPs. Furthermore, the physisorption data show that the porosity of the matrix can be tuned via thermal treatment, where repetitive thermal treatment at elevated temperature reduces the microporosity of the silica matrix.The numerical calculation, tomography, and overgrowth results concur in showing that the NPs in RCT catalysts are substantially embedded in the silica support. 200 out of 204 particles were more than 50% embedded in the silica support, and the majority of particles even more than 90%. NP embedding is known to prevent both Oswald ripening and particle sintering, the two mechanisms for particle growth. The large degree of embedding, and in particular the fact that the particles are embedded up to or above their waste line, prevents particle dislodgement and migration, explaining the absence of NP growth even upon thermal treatment at 800 and 950 °C, and during catalysis at elevated temperatures. The thermal stability is substantially enhanced compared to gold based catalysts with NPs lying on top of the support surface, which start to grow to larger particle sizes below 800 °C, and is comparable to fully encapsulated NPs. Hence, the RCT preparation approach endows a robust, porous 3D architecture with all catalytic NPs residing on the pore surfaces and available to the reactants, but significantly embedded into and thus stabilized within the support. The macropores ensure facile mass transport throughout the catalysts, preventing mass-transport limitations in the catalyst bed, whereas the micropores of the matrix enable local diffusion to the embedded side of the NP surface.