An additional advantage of the RCT methodology over direct encapsulation in porous shells is the possibility for facile tuning of the composition of the supporting oxide, for example, to a mixed titania-silica, titania, alumina, and zirconia support, and therefore control over the metalsupport interactions. The high degree of embedding originates from the preparation method, in particular from the infiltration stage when favorable wetting of the PS colloids and gold particles by the silica sol-gel solution necessitates the formation of the equilibrium solid-liquid-vapor contact line at high immersion values of the NPs into the silica precursor, as confirmed by the contact angle measurements and numerical modeling in Figure 1f,g. The favorable wetting stems from the high ethanol content in the silica sol-gel, which significantly lowers the surface tension of the sol-gel. As a result, the NPs are close to fully wetted irrespective of their size. The large spread in NP embedding likely originates from the varying volume of air that can be trapped around the NPs during infiltration. Indeed, the numerical calculations of the equilibrium contact line show that exposed surface area can vary between 0 and 70% depending on the volume of the air trapped around the NPs. This is in line with the quantitative analysis in Figure 3b, bato bucket where the exposed surface areas of all 204 particles lie between 0 to 60% exposed surface area.
Future work can be directed to tune the degree of embedding via the composition of the sol-gel precursor and via the NP surface functionalization, allowing structural control over the metal-support interface geometry. The crystal structure of the NPs in the RCT catalysts was penta-twinned which matches that of free-standing Au NPs. Furthermore, the orientation of their facets was random with respect to the pore, which is consistent with the random attachment of the NPs to PS colloids during theraspberry colloid preparation . The absence of asymmetries in the particle structure due to the embedding is expected given that the silica support is porous , meaning that both the embedded and exposed particle sides are subjected to the same type of gas atmosphere upon treatment. Finally, we want to note that identifying the 3D shape for small NPs is not trivial as restructuring during electron beam illumination can occur. This was avoided in this work by recording the data set at a relatively low magnification , thereby decreasing the electron dose and increasing the number of NPs in the field of view. The epitaxial overgrowth experiments show that both the exposed as well as the embedded side of the NPs are chemically accessible. The chemical accessibility of the embedded NP surface is likely due to mass transport through the micropores of the silica support. Additionally, the chemical accessibility could be explained by a small gap between the embedded surface and silica support resulting in the observed 2–3 atom thick Ag shell growth.
Whether and how the embedded side contributes to the catalytic performance is an open question. Yet, it is worth mentioning that the embedding doesn’t necessarily lead to reduced reactivity compared to non-embedded NP catalysts. In fact, Shirman et al. showed that Pd on alumina RCT catalysts were significantly more active compared to the traditionally supported catalysts, allowing a 90% reduction of the Pd loading compared to the conventional catalysts with the same catalytic performance. Hence, the contribution of the embedded interfaces, that is, their chemical structure and accessibility, should be carefully considered in interpreting the reactivity of these materials. Likely, the contribution of the embedded NP surfaces to the catalytic performance is an interplay between mass-transport limitations and nano-confinement effects. B. Dong et al. demonstrated that mass-transport to embedded NP surfaces is suppressed, but that the catalytic activity at the embedded surface itself is significantly increased due to nanoconfinement effects. Such nano-confinement effects can arise due to local enhancement of the reactant concentrations, altered adsorption-desorption equilibria, and/or stabilization of reactant intermediate species or states, yielding enhanced catalytic activity and potentially altered product selectivity. Our combined tomography and epitaxial overgrowth approach is not limited to the RCT catalysts studied here, but can be extended to catalytic and nanostructured materials in general. Visualization of metal-support interfaces in 3D materials using only 2D imaging techniques can give deceiving information as the imaging orientation influences the appearance of the metal-support interfaces, which is why electron tomography is preferred, even for qualitative studies.
Important parameters to consider in applying tomography are the Z-contrast between the individual components and the stability of the sample under electron beam illumination. Ideally, the difference between the metal, support, and pores of the catalyst is large, allowing facile segmentation. Alternatively, the metal overgrowth approach is simpler and quicker compared to electron tomography, allowing the direct assessment of the NP accessibility and a qualitative understanding of the metal-support geometry. The metal overgrowth methodology can easily be adapted to match a specific catalyst system by choosing a suitable metal for the overgrowth. Herein, the most important considerations are the difference in Z-contrast between the core and shell, and the growth behavior of the metal shell. The choice of metal is critical in ensuring epitaxial growth and a well-defined shell morphology. Epitaxial shell growth is achieved by choosing a metal with a lattice constant closely matching the underlying core . Smooth and continuous shell growth is ensured by a favorable interaction between the core and the shell, meaning that the second metal preferentially deposits on the cores. Furthermore, sufficiently slow overgrowth kinetics are important to ensure homogeneous shell growth throughout the sample, which can be achieved by using mild reducing agents, and by controlling the growth rate via, for example, pH and reaction temperature. We hypothesized that the segregation of polymer ligands into surface-pinned micelles having a footprint area comparable with the surface area of the nanoparticle could be used as a thermodynamically mediated strategy for the patterning of the high-curvature surface of nanocolloids. The formation of pinned micelles on planar surfaces has been studied for polymer molecules strongly grafted to a macroscopic planar surface. When a polymertethered substrate was transferred from a good to a poor solvent, a smooth layer segregated into micelles composed of a dense core and stretched surface-tethered ‘legs’ . The proposed approach to patchy nanoparticles is illustrated in Fig. 1a . On the nanoparticle surface, following the reduction in solvent quality, a uniformly thick polymer brush layer breaks up into a discrete number of pinned micelles . The process is driven by attractive polymer–polymer interactions and the competition between the polymer grafting constraints and the reduction in its interfacial free energy. Here we validate this approach for nanoparticles with different dimensions, shapes and chemical compositions, which were capped with various types of polymer and copolymer ligands and subjected to different external stimuli. We show experimentally and theoretically that the size of patches is governed by the polymer dimensions and grafting density, whereas the number of patches per nanoparticle is determined by the ratio between the nanoparticle diameter and polymer size. The patches could be permanently vitrified by polymer photocrosslinking. The resulting patchy nanoparticles acted as in situ colloidal surfactants and their self-assembly exhibited new binding modalities. We note that in addition to the generation of patchy nanoparticles, polymer segregation on the surface of nanoparticles has other far-reaching implications. Polymer-tethered nanoparticles have a broad range of applications in imaging and medical diagnostics, therapeutics, dutch bucket hydroponic and chemical sensing. The change in morphology of the polymer layerunder varying ambient conditions is of fundamental importance and can be used for the efficient design of nanoparticles aimed at specific applications. To explore the proposed approach, we synthesized gold spherical nanoparticles with a mean diameter D in the range from 20 ± 1.0 nm to 80 ± 1.5 nm, which were stabilized with cetyltrimethylammonium bromide or cetylpyridinium chloride. These low-molecular-mass ligands were replaced with thiol-terminated polystyrene molecules with a molecular mass of 29,000 g mol−1 or 50,000 g mol−1 .
The polymer-capped nanospheres were dispersed in dimethylformamide , a good solvent for polystyrene molecules. Figure 1b shows a transmission electron microscopy image of 20-nm-diameter nanospheres functionalized with polystyrene-50K. When cast on the grid from the solution in DMF, the nanospheres were engulfed by a uniformly thick polymer shell. Following the reduction in solvent quality for the polystyrene ligands—by adding water to the nanosphere solution in DMF—the polymer layer transformed into a surface patch . Since the surface mobility of thiol-terminated molecules is suppressed for multi-facet gold nanospheres and for high-molecular-mass ligands, and since their lateral motion is generally slow, we expected that in a poor solvent, stretched polystyrene-50K molecules would be grafted to the nanosphere surface, as is shown in Fig. 1a, bottom. Upon polymer surface segregation, the yield of patchy nanospheres was 65%; other species included small self assembled nanosphere clusters and nanospheres with a smooth shell . After removal of the clusters by centrifugation, the fraction of patchy nanospheres was about 98%. Patch formation was reversible: upon dilution of the solution with DMF to a water concentration of Cw < 1 vol% the core–shell nanosphere morphology was recovered. The formation of multi-patch nanospheres was explored for nanospheres with larger dimensions. Figure 1d shows a three-dimensional electron tomography reconstruction image of the 60-nm-diameter patchy gold nanosphere capped with polystyrene-50K . The nanosphere carried three polymer patches, each shown with a different arbitrary colour for clarity. The side view, obtained from tomographic reconstruction, revealed an elongated patch shape, which could be induced by the partial wetting of the substrate with the polymer solution. Some accumulation of the polymer at the nanosphere–substrate interface , supports this assumption. To ensure that polymer surface segregation occurs in solution, patchy gold nanospheres tethered with thiol-terminated random copolymer polystyrene-co-polyisoprene were introduced into a 0.05 wt% solution of photoinitiator azobisisobutyronitrile in the DMF/ water mixture and exposed to ultraviolet irradiation for 5 min. Partitioning of the photoinitiator into the patches and copolymer photocrosslinking yielded a permanent patchy structure on the nanosphere surface, which was preserved in tetrahydrofuran, a good solvent for the copolymer . Without crosslinking, the patches transformed into a smoothshell . Below, we refer to the non-crosslinked patchy nanoparticles, which were characterized by analysing their two-dimensional projections in TEM images. Patch formation and their structure were governed by polymer length, nanosphere diameter, and polymer grafting density. In the first series of experiments, we examined transitions between the nanospheres with a smooth polymer shell and patchy nanospheres at varying ratios between the nanosphere and polymer size. The trends shown in Fig. 2c were captured in the theoretical state diagram in Fig. 2d . The structure of the polymer layer on the nanosphere surface was governed by the polymer–solvent interfacial energy and the energy of stretching of end-tethered polymer molecules. In Fig. 2d, at high σ values , extended polymer chains minimized their interfacial and stretching energies by forming a smooth layer. At lower values of σ , the layer became thinner than the unperturbed molecular size of the polymer and the interfacial polymer–solvent energy was lowered by polymer segregation in pinnedmicelles. The elastic energy of stretched micellar ‘legs’ was comparable to the polymer– solvent interfacial energy. For large nanospheres, the transition between the two regions is shown as a blue line approaching the grafting density τ/, where τ accounts for the solvent quality and b is the monomer length.The effect of nanosphere size on patch formation was revealed by the position and incline of the boundary between the core–shell and patchy nanosphere states. The balance between the interfacial energy of the polymer and the free energy of stretching of the micellar ‘legs’ led to a higher stability of micelles on small nanospheres and hence a negative slope of the boundary line. Thus overall, the experimental and theoretical results were in excellent agreement. The versatility of the nanopatterning method was explored for nanoparticles with different shapes and compositions, capped with different polymer ligands strongly binding to the nanoparticle surface and subjected to different solvents . Following the prediction of the theoretical model on a stronger tendency for patch formation on surfaces with a high curvature, we examined polymer segregation on nanorods, nanocubes and triangular nanoprisms.