Multiscale Nanoparticle Assembly: From Particulate Precise Manufacturing to Colloidal Processing

Transcription

1 FEATURE ARTICLE Particle Assembly Multiscale Nanoparticle Assembly: From Particulate Precise Manufacturing to Colloidal Processing Markus Niederberger Nanoparticle assembly and colloidal processing are two techniques with the goal to fabricate materials and devices from preformed particles. While colloidal processing has become an integral part of ceramic processing, nanoparticle assembly is still mainly limited to academic interests. It typically starts with the precise synthesis of building blocks, which are generally not only considerably smaller than those used for colloidal processing, but also better defined in terms of size, shape, and size distribution. Their arrangement into 1D, 2D, and 3D architectures is performed with great accuracy well beyond what is achieved by colloidal processing. At the same time, the final assembly is not sintered such that the intrinsic, nanospecific properties of the initial building blocks are preserved or even lead to collective behavior. However, in contrast to colloidal processing the structures accessible by nanoparticle assembly are often limited to a small length scale. The review presents selected examples of nanoparticle assembly and colloidal processing with the goal to reveal the capabilities of these two techniques to fabricate novel materials from preformed building blocks, and also to demonstrate the immense opportunities that would arise if the two methods could be combined with each other. 1. Introduction Powder processing has a long and successful history in industry, e.g., in ceramic engineering and in powder metallurgy. Through compaction and densification by heat treatment powders are manufactured into bulk materials, components, or products usually at low costs and with high productivity. One of the big challenges of powder processing is the minimization of heterogeneities, which might be introduced into the material at any stage of the fabrication process. Once present, the heterogeneities can hardly be removed anymore, leading to strength limiting flaws in the final materials. A major source of heterogeneity is agglomeration in powders. [1] However, this issue can be tackled by manipulating interparticle forces as it is done in colloid science. [1,2] As a matter of fact, colloidal processing of Prof. M. Niederberger Laboratory for Multifunctional Materials Department of Materials ETH Zürich Vladimir-Prelog-Weg 5, 8093 Zürich, Switzerland markus.niederberger@mat.ethz.ch The ORCID identification number(s) for the author(s) of this article can be found under DOI: /adfm ceramics, which puts a strong focus on interparticle forces, suspension rheology, consolidation techniques, and drying behavior, became a powerful tool for the preparation of complex shaped ceramics with multiscale structures. [3] Many of the fundamental problems faced in colloidal processing are also relevant in nanoparticle assembly. However, in contrast to ceramic processing, where the particle size and shape are important to understand and manipulate the interparticle forces and the colloidal stability, [2,4] in nanoparticle assembly the importance of size and shape goes well beyond optimizing processing parameters. As a matter of fact, the size and shape dependent properties of nanoparticles offer the unique opportunity to get a combination of the intrinsic properties from the building blocks and collective or synergistic effects from their assembly in the final material. [5] However, these effects are only accessible, if the nanoparticles preserve their initial properties in the end product, which is in strong contrast to ceramic processing, where the final piece is densified and the particles are sintered together. Accordingly, the experimental procedure has to provide full control over the assembly of the building blocks including their spatial arrangement, orientation, and their interaction during the whole fabrication process. [6] The big challenge is to place every nanoparticle precisely at the right spot over several length scales, basically involving nanoparticulate precise manufacturing, in analogy to atomically precise manufacturing. [7] Research on nanoparticle assembly produced fascinating results in the last years, which cannot all be summarized and mentioned herein. However, a symposium at the ACS Spring meeting on the occasion of the 2017 ACS Award in Colloid Chemistry presented to Prof. Nicholas A. Kotov under the theme of Frontiers of Nanoscale Assemblies of Particles gave a nice overview of the current activities and results in the field. Many aspects of nanoparticle assembly were presented, including the synthesis of precisely defined building blocks such as gold nanoparticles with high shape uniformity and monodispersity, [8] the influence of shape on the assembly behavior, [9] DNA-directed assembly, [10] responsive nanoparticle assemblies, [11] surface patterning of nanoparticles, [12] or chiral nanoparticle assemblies mimicking biological structures. [13] In spite of all the undeniable progress, it was obvious that the fabrication of materials and structures with macroscopic (1 of 18) 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

2 dimensions by nanoparticle assembly remains a difficult task. Assembly methods, which make it possible to bridge several orders of length scales from the nano to the macroscale without losing the size-specific properties of the nanoparticulate building blocks, include colloidal nanolithography/ surface patterning, [14] layer-by-layer (LbL) approaches, [15] Langmuir Blodgett (LB) technique, [16] assembly at interfaces, [17] and more and more additive manufacturing as well. Although these methods enable the assembly of nanoparticles over areas large enough to be employed as electronic and optoelectronic devices, [18] these large-scale structures are typically far away from the highly ordered, precisely defined, and structurally complex architectures of their nano and micrometersized counterparts. There is still a big gap between the preciseness achieved on the nanoscale and the accuracy obtained on the macroscale. The goal to transfer the preciseness of nanoscale assembly to colloidal processing to produce macroscopic materials with the same complexity as it is nowadays achieved on a smaller scale still remains. In addition to size, shape also plays an outstanding role in nanoparticle research, and it turned out to be extremely useful to categorize the different shapes of nanoparticles according to their dimensionality, [19] i.e., 0D for spherical nanoparticles, 1D for nanotubes, nanowires, and nanofibers, and 2D for planar structures such as nanosheets, nanoplatelets, or nanolayers. For quite a while, the synthesis of low-dimensional nanostructures and their shape-dependent properties was the main focus. Meanwhile, the importance of shape is also recognized for nano - particle assembly, [9b,c,20] and one takes advantage of the shape to produce architectures with a specific particle orientation. In colloidal processing, the shape of nanoparticles is important too; however, until recently one did rarely use it in the sense to get specifically textured ceramics. In nature, the orientation of particles in the structure is absolutely essential to get the desired properties. Examples include nacre and tooth enamel, where the orientation of the inorganic particles determines the mechanical properties such as fracture toughness or hardness. If we compare colloidal processing and nanoparticle assembly, then we find considerable similarities, but also strong differences. As both approaches make use of colloidal dispersions as starting materials, interparticle forces play a dominant role, although on different size levels. Interparticle forces include not only van der Waals forces as the most ubiquitous form of (primarily attractive) interactions, electrostatic forces (attractive between oppositely charged and repulsive between similarly charged particles), and entropic effects such as steric repulsion and depletion forces, but also magnetic and electric dipole interactions and molecular surface forces stemming from molecules bound to the surface of the particles. [21] In principle, all these forces can be used for nanoparticle assembly. At the same time, it is essential to prevent any uncontrolled agglomeration at any stage of the process to ensure the assembly of individual nanoparticles and to minimize the heterogeneities in ceramics. The most common methods to stabilize colloidal dispersions include electrostatic stabilization, e.g., by tuning the ph or the ionic strength of the aqueous medium and steric stabilization involving the adsorption of polymers or surfactants on the particle surface. [4,22] In nanoparticle synthesis, the use of organic surfactants to suppress agglomeration Prof. Markus Niederberger is chair of the Laboratory for Multifunctional Materials and full professor in the Department of Materials at ETH Zurich. He studied chemistry at ETH Zurich, where he also received his PhD degree. After a postdoctoral stay at the University of California at Santa Barbara, he became a group leader at the Max Planck Institute of Colloids and Interfaces in Potsdam. In 2007, he was appointed as an assistant professor at ETH Zurich. and to ensure easy dispersibility became routine. In addition, the surfactants have a key role in determining not only the surface chemistry [23] but also the size and the shape of the particles by affecting the nucleation and growth kinetics. [24] Typically, they remain adsorbed on the surface of the nanocrystals and thus influence their optical and electronic properties. [25] The resulting effects can be favorable as well as detrimental. In general, surfactants are hydrophobic and electrically insulating, such that they significantly hamper charge transport, which is a big issue for their application in optoelectronic devices. On the other hand, ligand exchange reactions make it possible to replace the surfactants from the synthesis by ligands optimized for a specific application. Furthermore and somewhat counterintuitive, the surface-bound species can be utilized to promote the catalytic performance of nanoparticles. [26] In colloidal processing, organic residues are less of a problem, because they are typically burnt off during the sintering of the ceramics. Regarding the size of the produced structures, colloidal processing and nanoparticle assembly are rather complementary, i.e., nanoscale assembly offers the possibility to arrange nanoparticles very precisely on a relatively small size scale, while colloidal processing operates on a larger size scale, however with less control. For future fabrication of complex architectures with structural control from the nanometer to the centimeter scale, in large quantities, and with macroscopic dimensions and intricate shapes, the concepts of both techniques have to be combined. However, before such a merger is possible, we have to get an idea of the capabilities of both processing techniques, which is the main goal of this review. The major part of the review is dedicated to nanoparticle assembly. As this is an extremely broad field, the examples are structured according to the dimensionality of the building blocks and final assembled structures. After a short and general overview on the use of nanoparticles as building blocks for a diverse range of architectures, the assembly of 0D, 1D, and 2D building blocks is presented. Finally, the review ends with a short section on colloidal processing. It is not the goal to provide an exhaustive summary of the whole field, but to discuss selected examples to illustrate the general developments. The field of colloidal processing has been extensively reviewed in the ceramics literature. [3,27] We restrict the discussion to recent (2 of 18) 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

3 examples, where specific efforts were undertaken to position and orient platelets as anisotropic building blocks within a macroscopic body. These examples are believed to have a direct link to nanoparticle assembly and thus present a connecting point to combine concepts of nanoparticle assembly with colloidal processing. 2. (Nano-) Particles as Building Blocks The synthesis of nanoparticles has always been and is still at the heart of nanoscience. For decades the progress in nanoscience was strongly connected to the discovery of new synthesis methods such as the citrate route for gold nanoparticles, [28] the oleic acid assisted preparation of monodisperse iron oxide nano - particles, [29] hot-injection method for semiconductor quantum dots, [30] shape-controlled synthesis of CdSe nanocrystals, [31] or atomically precise size controlled synthesis of iron oxide nanoparticles. [32] In the most cases, the nanoparticles are obtained as powders, which resulted in the formation of a new research branch in nanoscience, namely, the processing of nanopowders into structures that are suitable for a targeted application. The concept of using nanoparticles as building blocks, which can be assembled in a modular way to larger entities, turned out to be extremely successful. The fascination but also the technological potential of such an approach lies in the fact that starting from a limited number of building blocks, which can even be compositionally well-known compounds, an indefinite number of new architectures are accessible, only limited by imagination, through arrangement or patterning of single or multiple building blocks. Figure 1 displays some of the possible structures that can be produced from spherical nanoparticles as building blocks. On the smallest size scale, i.e., on the size scale of the individual building blocks, single nanoparticles can be patterned on a substrate (Figure 1a). Assembly of just a few nanoparticles might lead to colloidal molecules (Figure 1b) or nanocrystal clusters (Figure 1d). Nanoparticle assembly along one direction leads to pearl-necklace-like nanostructures (Figure 1c) and along two directions to monolayer films (Figure 1e). A step further is the assembly of nanoparticles into extended superlattices (Figure 1f), which are characterized by close packed structures. However, nanoparticles can also be assembled into porous structures such as mesoporous films (Figure 1g) or inverse opals (Figure 1h). In these two cases, the arrangement of the pores is ordered. While it is possible to make mesoporous materials as films with macroscopic dimensions (i.e., on the cm scale), it is a great challenge to reach similar length scales with inverse opals. If a macroscopically sized bulk material is desired, then typically the high order of the pores and also the narrow pore size distribution get lost, like in the case of foams (Figure 1i) or aerogels (Figure 1j). In Figure 1, the individual building blocks are still clearly distinguishable in the final structure. However, depending on the desired degree of connectivity of the nanoparticles, and thus the density and mechanical stability of the architecture, a sintering step can be added to the processing. In such a case, the properties of the initial building blocks will partly or fully be replaced by bulk properties. Accordingly, annealing represents another option to further tune the properties. In addition to the local placement of the nanoparticles, determining the overall geometry of the structure, and heat treatment affecting density/porosity and mechanical stability, a third parameter is represented by the crystallographic orientation of the nanoparticles with respect to each other (Figure 2). [33] Some nanoparticle assemblies can be produced in such a way Figure 1. Assembly of nanoparticles into a) single-particle arrays, b) colloidal molecules, c) 1D pearl-necklace structures, d) nanoparticle clusters, e) 2D monolayer arrays or thin films, f) 3D superlattices, g) (meso-) porous films, h) inverse opals, i) foams, and j) aerogels (3 of 18) 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

4 Figure 2. Schematic of the different possibilities to assemble nanocrystals. Left branch: Formation of ordered arrays with or without crystallographic alignment and with or without coalescence. Right branch: Analogous assembly options in disordered arrays. Adapted with permission. [33] Copyright 2013, Wiley-VCH Verlag GmbH & Co. KGaA. that the crystallographic orientation of the nanoparticles is not random. For example, a superlattice turns into a mesocrystal, where the nanoparticles within the assembly are crystallographically aligned, but the nanoparticles remain stabilized by organic ligands to prevent coalescence (Figure 2, left branch). If the organic surface coating is removed, then superlattices form polycrystals and mesocrystals transform into single crystals. The right branch in Figure 2 displays disordered arrangements of nanoparticles, which basically exhibit the analogous behavior like the ordered ones, i.e., arrays with or without aligned atomic lattice and coalescence. Although in the literature the disordered structures often attract less attention, simply because their appearance is less appealing, for many applications a high order of the nanoparticle assembly is not required. On the contrary, disorder in the system might lead to additional porosity, which can be beneficial. From a practical point of view, it is usually much easier to fabricate macroscopic materials with less order. Figure 1 already summarized the immense possibilities one has to arrange spherical nanoparticles into surface patterns, colloidal molecules, clusters, 1, 2, and 3D structures, densely packed or porous, with ordered or disordered porosity, with or without crystallographic orientation, and finally with coalesced or noncoalesced particles. One important parameter is still missing: the shape of the nanoparticles. Of course, all these structures can also be built up by using anisotropic or chiral nanoparticles, which expands the variety of available structures even more. Chiral nanoparticles and nanoparticle assemblies will not be discussed here (the interested reader is referred to ref. [34]). In the next three sections, we will direct our attention toward the use of spherical (0D) and anisotropic (1D and 2D) nanoparticles for arrays and structures covering all dimensionalities from 1D to 2D and 3D. We will start our discussion with spherical nanoparticles. 3. 0D Building Blocks As shown illustratively in Figure 1, there are many possibilities to arrange spherical nanoparticles. 2D nanoparticle patterns on surfaces in a non-close-packed arrangement (Figure 1a) and over large areas typically require the combination of lithography techniques with self-assembly, where the nanoparticles are patterned on prestructured surfaces. [35] 0D nanoparticles can also be used as building blocks for colloidal molecules, Janus particles, or heteronanostructures (Figure 1b). [36] As an example, Figure 3a,b shows transmission electron microscopy (TEM) and high resolution transmission electron microscopy (HRTEM) images images of different heterostructures formed by induced fusion of preformed iron oxide and gold nanoparticles. [37] The synthesis of chain-like, 1D nanostructures from spherical building blocks (Figure 1c) is a very active research area. [38] Examples include nanoparticle chains of Ag, [39] Au (Figure 3c), [40] CdTe, [41] or Co. [42] As long as the nanoparticles remain stabilized by organic ligands attached to their surface, they do not coalesce, but form pearl-necklace-like structures, whose formation has some analogy to the polymerization of organic monomers. [33,43,44] However, if the stabilizing molecules are removed, then the nanoparticles undergo oriented attachment, i.e., they coalesce in a crystallographically oriented fashion, forming single-crystalline-like nanorods or nanowires. Originally discovered during the coarsening of titania nanoparticles under hydrothermal conditions, [45] oriented attachment became a powerful tool to fabricate single-crystalline 1D nanostructures such as nanowires and nanorods from preformed spherical nanoparticles. [46] Just as one selected example, Figure 3d shows the formation of ZnO nanorods from the corresponding spherical building blocks. [47] If we go a step higher in dimensions, we come to 2D structures like monolayers, nanosheets, thin films, and coatings. A particularly suitable method for the arrangement of spherical nanoparticles into 2D geometries is based on self-assembly processes at interfaces. [17] An example here is the preparation of cobalt-platinum nanoparticle monolayers deposited by the Langmuir Blodgett technique (Figure 3e). [48] In this case, the monolayer consisted of densely packed, ordered nanoparticles, which, even though they were still surface-stabilized and thus separated from each other, showed thermal activated charge transport. According to the authors, the monolayers once transferred to silicon wafers extended over vast areas up to the millimeter range. On a smaller size scale CdTe nanocrystals were (4 of 18) 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

5 Figure 3. Collection of structures assembled from 0D nanoparticulate building blocks. a) TEM overview image of different Fe 3 O 4 -Au heterostructures. b) HRTEM image of a dumbbell-like Fe 3 O 4 -Au-Fe 3 O 4 heterostructure. Images (a) and (b) reproduced with permission. [37] Copyright 2006, American Chemical Society. c) TEM image of gold nanoparticle chains. Reproduced with permission. [40] Copyright 2011, Wiley-VCH Verlag GmbH & Co. KGaA. d) TEM image of spherical ZnO nanoparticles (left), which assembled into ZnO nanorods (right) by oriented attachment. Reproduced with permission. [47] Copyright 2002, Wiley-VCH Verlag GmbH & Co. KGaA. e) Scanning electron microscopy (SEM) images of a cobalt-platinum nanoparticle monolayer on a silicon wafer. Reproduced with permission. [48] Copyright 2008, American Chemical Society. f) TEM image of free-floating sheets of CdTe nanoparticles. Reproduced with permission. [49] Copyright 2006, American Association for the Advancement of Science. g) Illustration of the oriented attachment process of PbS quantum dots into a sheet. Reproduced with permission. [50] Copyright 2010, American Association for the Advancement of Science. found to self-assemble into free-floating sheets with sizes of up to µm (Figure 3f). [49] The sheets showed some mechanical robustness, although they were composed of just one monolayer of nanocrystals of 3.4 nm in diameter. As the nanocrystals were not fused together, but remain discrete units, they showed the typical photoluminescence of quantum dots, however, with the peak maximum being redshifted. The sheets were formed in the absence of any dimension-restrictive interface, raising the question, why the nanocrystals assembled into 2D structures rather than into simple agglomerates. The authors proposed a combination of anisotropic electrostatic interactions and directional hydrophobic attraction as a result of a slightly anisotropic particle shape. [49] If the nanocrystals within a 2D assembly undergo oriented attachment (Figure 3g), ultrathin single-crystalline sheets with dimensions on the micrometer scale are formed, as observed for PbS nanocrystals. [50] The driving force for oriented attachment in 2D was found not only in the minimization of high energy surfaces, but also in the formation of a highly ordered bilayer of surfactant between the nanosheets. The fusion of the particles under removal of the surfactants at these specific crystal facets resulted in pronounced photoconductivity in the in-plane direction. [50] Further examples of macroscopically sized, supported, and free-standing films with unique functionalities will be discussed in Section 6. If a dispersion of nanoparticles with a narrow size distribution is slowly evaporated or gently destabilized (e.g., by slow diffusion of a nonsolvent into the colloidal dispersion), 3D ensembles, (5 of 18) 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

6 Figure 4. a) TEM image of a binary Fe 2 O 3 -Au superlattice isostructural with NaCl. Reproduced with permission. [54] Copyright 2006, American Chemical Society. b) TEM image of crystallographically oriented maghemite nanocubes with long-range order. c) Electron diffraction pattern obtained from (b). Reproduced with permission. [55] Copyright 2007, National Academy of Sciences. d) SEM image of a macroporous CdSe nanocrystal solid. Reproduced with permission. [57] Copyright 1999, Wiley-VCH Verlag GmbH & Co. KGaA. e) SEM image of mesoporous SnO 2. Reproduced with permission. [58] Copyright 2005, Wiley-VCH Verlag GmbH & Co. KGaA. so-called superlattices, form. [51] These colloidal crystals or nanocrystal solids are composed of ordered, close-packed nanoparticles, which are regarded as artificial atoms in the next level of hierarchy. [52] In the simplest case, the superlattices are built up by only one type of nanoparticles, leading to face-centered cubic or hexagonal close-packed structures with maximum packing densities of [53] If binary superlattices from two different types of nanocrystals are formed, the obtained structures are strongly (but not exclusively) influenced by the relative sizes of the building blocks, in analogy to the anion-to-cation ratio in ionic crystals. [52b] Figure 4a shows as an illustrative example a TEM micrograph of a superlattice isostructural with NaCl. [54] In most of the superlattices, the orientation of the lattice planes of the spherical nanocrystals with respect to each other is random (cf. Figure 2). However, as soon as the nanocrystals start to develop more pronounced crystal facets, preferred orientation is observed in some cases. [52a] If the nanoparticles are superparamagnetic, like in the case of oleate-capped maghemite nanocubes, it is possible to produce highly oriented superlattices with long range order by applying a magnetic field during drying (Figure 4b). [55] The obtained mesocrystals (cf. Figure 2) reached dimensions of up to 10 µm. This is also the size range, which is typically obtained with other superlattices. [56] Because the degree of ordering in the superlattice is strongly connected to the evaporation rate of the solvent or the destabilization speed of the dispersion, fast evaporation, e.g., by using a low boiling solvent, resulted in the formation of glassy-like structures. [52a] These examples all describe densely packed 3D structures. Of course, nanocrystals can also be arranged into porous, ordered or disordered, 3D structures. Ordered porous materials are typically produced by making use of hard or soft templates. Photonic crystals from semiconductor quantum dots were prepared by infiltrating a network of ordered, close-packed silica spheres with CdSe quantum dots. [57] After removal of the silica spheres, a macroporous quantum dot solid with a regular arrangement of pores was obtained (Figure 4d). Such inverse opal structures (cf. Figure 1h) with pores ordered on an optical wave length scale offer interesting optical properties. While the nonannealed samples still showed the properties of the individual nanocrystals, sintering resulted in the formation of macroporous bulk semiconductor with the appropriate symmetry and high refractive index required for a complete photonic bandgap at optical wavelengths. [57] For the assembly of nanocrystals into materials with smaller pore sizes in the mesoscale range (cf. Figure 1g), the silica spheres are replaced by block copolymers as templates. For example, the evaporation-induced selfassembly of tin oxide nanocrystals of 3 6 nm in size in the presence of polybutadieneblock-poly(ethylene oxide) in tetrahydrofuran yielded porous tin oxide mesostructures with high order and with pore sizes in the range of nm (Figure 4e). [57] Although nanoparticle-based mesoporous structures are typically sintered to remove the organic templates, there is still some interparticle porosity left (in addition to the mesopores). Such a bimodal nanoporous architecture was found to be particularly beneficial for the development of nextgeneration electrochemical capacitors. [59] Based on a limited number of illustrative examples, this section highlighted the many possibilities one has to arrange spherical building blocks into 1D, 2D, and 3D assemblies with or without crystallographic orientation, with or without long-range order, densely packed or with porosity, and with or without fusion of the nanocrystals. But what is common for all these structures is the limitation in size. They are rarely larger than a few tens of micrometers. In the next section we will switch from 0D to 1D nanoparticles as building blocks and we will present selected examples of structures built up by nanowires, nanorods, and nanofibers. 4. 1D Building Blocks 1D nanostructures, such as wires, tubes, belts, and rods, are attractive building blocks due to their interesting and (6 of 18) 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

7 Figure 5. Assembly of 1D building blocks into ordered and disordered 1D, 2D, or 3D structures. unique electronic, optical, thermal, mechanical, and magnetic properties. [60] It is obvious that in comparison to spherical building blocks, anisotropic particles not only open up new possibilities for shape-induced assembly experiments in general, but also pose additional difficulties, because they require control over the orientation of the particles within the ensemble in addition to positional ordering to take full advantage of the anisotropic properties. Figure 5 shows an overview of the different possibilities to assemble 1D building blocks over several dimensions, either in an ordered and possibly close-packed fashion such that the obtained structure is also highly anisotropic, or in a disordered, porous configuration leading to (more) isotropic arrangements. Many strategies have been reported for the assembly of 1D nanocrystals, [61] including evaporation, field, and templateassisted approaches, assembly at interfaces, and chemical linkage. [33,44,62] One way to organize nanorods into disordered 1D ensembles involves the selective functionalization of specific regions of the particles with surfactants and polymers of varying hydrophilicity/hydrophobicity. This strategy was pioneered by the group of Kumacheva for amphiphilic gold nanorods. [63] Hydrophilic gold nanorods tethered with a hydrophobic homopolymer at both ends resemble the segments of ABA block copolymers. Taking advantage of the segregation properties of the constituent blocks, the nanorods can be organized into predictable structures either by selectively changing the solvent [63a] or by varying the molecular weight of the polymer molecules and their relative location on the rod. [63b] Figure 6a shows the effect of the molecular weight M n of the hydrophobic polymer on the self-assembly behavior of the gold nanorods in a dimethylformamide (DMF)/water mixture. The left column of Figure 6a shows a schematic of the different polymer structures at the tips of the nanorods, while the middle and the right columns, respectively, illustrate the association modes and the experimentally observed structures. [63b] If the polymer molecule is very short, then no self-assembly occurred (Figure 6b). With increasing molecular weight, the gold nanorods started to organize into chains (Figure 6c), sometimes forming rings (Figure 6c, inset). If the polymer coverage expands to the long side of the nanorods, both end-to-end and side-by-side assembly happened (Figure 6d). But there is clearly no long range order between the different nanorod chains and rings. To get highly ordered 1D structures of gold nanorods, the assembly can be performed on wrinkled surfaces. [64] Nearly perfect uniaxial alignment and close-packing of gold nanorods over macroscopic areas of several cm 2 was achieved. [64] Depending on the ratio of the wrinkle amplitude to nanorod diameter, triple, double, or single lines of nanorods were produced (Figure 6e (A, B, and C), respectively). The vast majority of assembly work using nanowires and nanorods as building blocks has been carried out for 2D arrangements, in particular for films. [61a] Also here, we can distinguish between close-packed ordered and porous disordered structures (cf. Figure 5). The preparation of disordered nanowire films is straightforward and typically involves the wet-chemical deposition of nanowire dispersions on a substrate by dip or spin-coating or doctor blading. These processing techniques offer two big advantages: The films easily reach macroscopic dimensions as often required for device fabrication and the porosity can be tuned over a broad range by applying different types and amounts of porogens or templates. The porosity increases the accessible surface area in the films, which is beneficial for applications in, e.g., gas sensing, [65] in catalysis, [65b] photocatalysis, [66] and in energy storage and conversion. [60b,67] Moreover, an appropriate porosity increases the transparency of the films, which makes nanowire-based films highly attractive as transparent electrodes in electronic and optoelectronic applications. [68] Nevertheless, it is obvious that if one wants to take full advantage of the highly anisotropic properties of the building blocks, the nanowires and nanorods have to be oriented within the 2D architecture. Such liquid-crystal-like structures were found to exhibit anisotropic photocatalytic properties [69] or good performance for UV light detection. [70] Gas liquid interfacial assembly in combination with the LB technique is particularly suitable for the oriented arrangement of 1D building blocks over a large area (Figure 7a). [16,62] Generally, a dispersion of hydrophobically surface-functionalized nanowires in an organic solvent is spread on the water surface of an LB trough, forming a monolayer of randomly (7 of 18) 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

8 Figure 6. a) Left column: Sketch of the relative location of PS molecules with varying molecular weight on the gold nanorods. Middle column: Sketch of the self-assembled structures upon controlled destabilization of the gold nanorod dispersions. Right column: SEM images of the self-assembled structures composed of gold nanorods with polymers of varying number-average molecular weights: b) 5000, c) , and d) Reproduced with permission. [63b] Copyright 2008, American Chemical Society. e) Dependency of the geometrical arrangement of the nanorods on the wrinkle amplitude using dip-coating: A) triple, B) double, and C) single lines with decreasing wrinkle amplitude. Reproduced with permission. [64] Copyright 2015, The Royal Society of Chemistry. oriented nanowires at the air water interface (Figure 7b, left). The monolayer of the nanowires is then compressed with two barriers and the surface pressure is monitored. With increasing compression, the packing density increases (Figure 7b, middle), until a dense arrangement of aligned nanowires is reached (Figure 7b, right). The film is finally transferred to a substrate by dip-coating. With this technique, macroscopic films can be produced not only on flat but also on patterned substrates, which enables the deposition of the films on substrates equipped with electrodes. This geometry is required for many electronic devices and has been successfully implemented in chemoresistive gas sensors composed of oriented tungsten oxide nanowire films. [71] If the nanowires have a high aspect ratio combined with a small diameter, then it is particularly challenging to get long-range order, because the structural flexibility induces bending, curling, and entanglement of the nanowires. On the other hand, if the aspect ratio is small, like in the case of short nanorods, the 1D building blocks can also be vertically aligned with respect to the substrate. In principle, the LB technique can be extended into 3D by repeating the process to get multilayer films. It is even possible to rotate the nanowires by 90 after every step to get nanowire films with crisscross patterns. [72] Unfortunately, the experimental LB procedures are time consuming and labor intensive such that the number of layers is usually limited to just (8 of 18) 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

9 Examples of disordered 3D structures reaching macroscopic dimensions using 1D building blocks will be discussed in section 6. Figure 7. a) Sketch of the Langmuir Blodgett technique to assemble nanorods at the air water interface. Reproduced with permission. [62] Copyright 2014, The Royal Society of Chemistry. b) TEM images of tungsten oxide nanowires at the air water interface at different surface pressures (1 mn m 1 (left), 5 mn m 1 (middle), and 25 mn m 1 (right)). Reproduced with permission. [71] Copyright 2015, Wiley-VCH Verlag GmbH & Co. KGaA. TEM and HRSEM images of CdS nanorods after evaporation of dispersions with increasing concentrations from (c) (e): c) disordered nanorods with their c-axis parallel to the TEM grid, d) monolayer, and e) multilayer of vertically oriented nanorods. Images (c) (e) reproduced with permission. [74] Copyright 2012, The Royal Society of Chemistry. a few. As a matter of fact, 3D assembly of 1D building blocks, especially in an oriented fashion, is in general challenging and restricted to small sizes. For example, Cao and co-workers reported the self-assembly of CdSe-CdS nanorods about 30 nm long and 7 nm wide into colloidal spherical superparticles with sizes controllable from 180 to 1100 nm. [73] Ordered supercrystals were also obtained by evaporating dispersions of CdS and CdSe nanorods of different aspect ratios. [74] Depending on the concentration of the dispersions, surface charge of the nanorods, their dipole moment, and the polarity and volatility of the solvent, the nature of the assembly could be controlled. The concentration of the rods has to be high enough to enable sufficiently small interrod distances, where attractive interactions dominate. These forces are additionally influenced by the screening effect of high permittivity solvents. Moreover, the boiling point of the solvent dictates the degree of order. Depending on the experimental conditions, the CdS nanorods are either randomly distributed with their c-axis parallel to the plane of the TEM grid (Figure 7c), or form mono layers of vertically oriented hexagonal close-packed arrays (Figure 7d) or 3D multilayers (Figure 7e). [74] 5. 2D Building Blocks The exciting new physics of graphene [75] arouse tremendous interest in 2D materials such that nowadays almost all classes of functional materials are available as atomically thin layers, including insulators (e.g., hexagonal boron nitride), [76] semiconductors (e.g., transition metal dichalcogenides), [77] metal oxides/hydroxides, [78] and polymers. [79] The unique electronic and optical properties in combination with high mechanical flexibility and strength make these layers promising candidates for a wide range of applications in electronics and optoelectronics. [77b,c] However, there are many other 2D materials, which are thicker, but still have one dimension much smaller than the other two, and which can be used as anisotropic 2D building blocks. Figure 8 summarizes the different architectures accessible from 2D particles. Again, from 1D to 2D and 3D, from ordered to disordered assemblies everything is possible. However, from an application point of view, not all structures received the same attention. Starting from nanolayers, especially 2D arrangements (i.e., thin films) and 3D ensembles (e.g., Van der Waals heterostructures) are attractive for devices. An interesting assembly behavior in dependence of the synthesis conditions was reported for tungsten oxide nanoplatelets. [80] While the reaction of tungsten chloride with benzyl alcohol resulted in the formation of isolated tungsten oxide platelets, the presence of a small amount of 4-tert-butylcatechol yielded tungsten oxide nanoplatelet stacks (Figure 9a). The stacks are nm in width and hundreds of nanometer long. Although some of these rods indicate a certain preference for lateral aggregation, most of the stacks are randomly oriented Figure 8. Sketch of the different possibilities to assemble 2D building blocks into ordered and disordered 1D, 2D, or 3D structures (9 of 18) 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

10 Figure 9. a) TEM image of 1D tungsten oxide nanoplatelet stacks. b) TEM image of a part of such a stack, revealing the organic inorganic nature of the assembly. c) Bundles of parallel oriented nanocolumns composed of self-aligned nanoplatelet stacks. Reproduced with permission.[80] Copyright 2006, The Royal Society of Chemistry. TEM images of parallel 2D arrangements of d) rhombohedral DyF3 nanoplates, e) small aspect ratio hexagonal TbF3 nanoplates, and f) large aspect ratio hexagonal EuF3 nanoplates together with wide angle (upper right) and small angle (lower right) electron diffraction patterns. Reproduced with permission.[9a] Copyright 2013, Macmillan Publishers Ltd. g) Sketch of a layered van der Waals heterostructure composed of atomic planes with different composition similar to the assembly of Lego bricks. Reproduced with permission.[84] Copyright 2013, Macmillan Publishers Ltd (10 of 18) 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

11 with respect to each other (cf. Figure 8, upper right). At higher magnification (Figure 9b) it becomes obvious that the stacks comprise a highly ordered, lamellar organic inorganic hybrid nanostructure with an interlamellar distance of about 1.7 nm. The platelet stacks can be broken down into the single platelets after dispersing them in water, indicating weak interactions between the platelets. A much higher order of parallel oriented nanoparticle stacks (cf. Figure 8, upper left) was observed for tungsten oxide nanoplatelets synthesized in 4-tert-butylbenzyl alcohol. In this case, the nanoplatelets are much smaller, forming columns of uniform diameter of 4 nm and lengths of several micrometers (Figure 9c). Each nanocolumn consists of self-aligned nanoplatelets facing each other along the entire stack. The nanocolumns form bundles, within which they lie parallel to each other along the whole length. However, different bundles of nanocolumns are disordered with respect to each other. Examples of planar and ordered 2D arrangements of faceted nanoplatelets are shown on the TEM images in Figure 9d f. [9a] Assembled by an interfacial strategy, rhombohedral DyF 3 nanoplates (Figure 9d), low aspect ratio hexagonal TbF 3 nanoplates (Figure 9e), and high aspect ratio hexagonal EuF 3 nanoplates (Figure 9f) form 2D superstructures at the liquid air interface with high periodicity and impressive packing precision. Clearly, the high monodispersity of the nanoplates is the prerequisite to achieve such long-range ordered tilings. However, for most applications, it is not necessary that the 2D building blocks form highly ordered films. On the contrary, the films can be disordered and porous as long as the building blocks form a continuous, percolating network (cf. Figure 8, middle right). A typical example in this direction is graphene. For most of the applications in electronics and optoelectronics, graphene has to be processed into thin film electrodes. [81] The easiest way to do this is by solution phase processes like dip or spin-coating, drop-casting, or spraying of graphene oxide dispersions onto a substrate, followed by a reduction step. In comparison to graphene layers grown by chemical vapor deposition, the quality and thus also the performance of such films are significantly lower. Solution deposition makes use of graphene oxide sheets, which are usually rather polydisperse and which therefore cannot be packed densely into ordered structures. On the other hand, the films can be deposited on 300 mm wafer scale and the properties are still good enough to be used in electronic devices, where the extraordinary electrical properties of chemical vapor deposition (CVD) grown films are not required. [82] Graphene is also a versatile 2D building block for porous and disordered 3D structures (cf. Figure 8, lower right), mainly for applications in electrochemistry. [83] However, in the last part of this section we want to focus on ordered 3D heterostructures, which were produced by stacking extended 2D building blocks like Lego bricks (Figure 9g). [84] This research direction developed as a result of the exciting discoveries made in the field of graphene research. Scientists started to direct their attention toward other 2D materials such as metal dichalcogenides or layered oxides. In comparison to single component materials like graphite, the idea is now to use exfoliated monolayers of different types of materials and reassemble them into a heterostructure of defined layer sequence. While strong covalent bonds exist within one atomic plane, only van der Waals forces act between the individual layers. Therefore, these artificial materials are also called Van der Waals heterostructures. [84] It is obvious that such a layer-by-layer approach (and here we talk about atomic monolayers) opens up fascinating possibilities to combine different types of materials within one 3D crystal. For example, the combination of ferromagnetic Ti 0.8 Co 0.2 O 2 nanosheets with dielectric perovskite-structured Ca 2 Nb 3 O 10 nanosheets made it possible to engineer the interlayer coupling, giving access to artificial multiferroic superlattices. [85] 2D materials, especially graphene, have risen tremendous expectations to become the next disruptive technology. [86] But in spite of worldwide efforts in academia and industry, there is still a long way to go, especially when it comes to scalability of the synthesis methods and accuracy of the assembly techniques. Other low-dimensional materials like 0D quantum dots or 1D nanowires or nanotubes were discovered and investigated a long time before, but their potential is still not yet fully exploited mainly due to the fact that it is difficult to process or implement them into macroscopic architectures and devices. The same is true for 2D building blocks. Only if they can be precisely assembled over several length scales, they will be of practical use. [87] An additional challenge in the use of 2D materials as building blocks is their strong tendency to restack, which destroys the unique 2D properties. This problem is particularly pronounced in solution-based assembly methods. Another issue of wet-chemical nanoparticle assembly is the high probability of introducing defects. In this regard, gas phase routes like CVD produce materials of higher quality. Nevertheless, advantages like scalability, high yield, simple and nonvacuum based equipment, and thus an overall good cost-effectiveness, still make solution processing the method of choice. Therefore, we will discuss in the last section of this review selected examples, in which particulate precise manufacturing is merged with colloidal processing. The examples show that precise particle positioning and orientation is also possible in macroscopically sized materials and bodies. Accordingly, these examples illustrate how the subtleness of nanoparticle assembly can be combined with an industrially relevant processing technique to fabricate real materials with synergistic properties stemming not only from the building blocks themselves but also from their geometrical arrangement and orientation. 6. From Nano to Macro: Assembly and Colloidal Processing over Several Length Scales After discussing selected examples of nanoparticle assembly routes, in which nanoparticles are arranged and positioned with high precision over several dimensions, but typically only over restricted size scales, we direct our attention in this section toward colloidal processing routes. Generally, they allow less control over the assembly of the nanoparticles, but give access to larger structures on the macroscopic size level. In contrast to nanoparticle assembly, colloidal processing methods are an integral part of ceramics technology. If we want to implement nanoparticle precise manufacturing into industrially viable processes, then we have to be able to combine the precision of nano particle assembly with the scalability of colloidal processing (11 of 18) 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

12 We will start this section with the assembly of 0D (i.e., spherical) nanoparticles into supported and self-standing films. Here the focus is on films with macroscopic dimensions such that they can directly be implemented in devices. Then, we will present selected examples of how to process 0D, 1D, and 2D nanoparticles into centimeter scale monolithic bodies with disordered porosity as present in foams and aerogels. The section will conclude with selected examples of textured ceramics and composites prepared by colloidal processing routes involving the use of magnetic fields to align and orient the 2D building blocks within the materials over several length scales. Thin films are an integral part of many electronic and optoelectronic devices. Accordingly, considerable research efforts were dedicated to the processing of nanoparticles into films. Meanwhile it is routine to produce films and coatings over macroscopic areas from nanoparticles by colloidal processing. Of course, in these cases, the nanoparticles are randomly arranged, without any preferred orientation or local position. But even though these systems seem to be rather simple, quite a lot of engineering is required for their fabrication. While thin films are straightforwardly accessible as long as their thickness is below nm, drying and sintering conditions have to be carefully optimized for thicker films. [88] The removal of the organic surface stabilizers during compaction of the films causes stresses, which results in cracking. [88b] However, the avoidance of crack formation is just one aspect of thin film processing. The real moment of truth comes, when the films are either implemented in a device or when they have to exhibit a specific function. Only few examples in the literature show nanoparticle films as part of a device, because in such a case not only the quality of the film is essential but also its stability during the next device fabrication steps and its compatibility with the other components of the device. Successful examples include Sb:SnO 2 nanoparticle films as anode layer in organic light emitting diodes, [89] functional layers for chalcogenide solar cells, [90] or nanoparticle multilayers in optical filters, which reflect a specific range of the electromagnetic spectra. [91] In comparison to films produced by processes like pulsed laser deposition or molecular beam epitaxy, nanoparticle-based films are of considerably lower quality. They are polycrystalline, disordered, and contain residual porosity even after sintering. Nevertheless, they can still fulfill rather complex functions like simultaneous ferroelectric and ferromagnetic order, and the coupling between these two orders even enables voltage control of the magnetism as reported for BaTiO 3 -CoFe 2 O 4 nanoparticle films. [92] To produce macroscopically sized, but free-standing films, the layer-by-layer (LbL) assembly approach is particularly promising. [15b,93] It includes the sequential deposition of oppositely charged building blocks. The problem here is that the increase in layer thickness per deposition cycle is rather small and therefore many cycles are required to fabricate mechanically stable free standing films. Kotov and co-workers reported the synthesis of self-standing films of gold nanoparticles in polyurethane. [15a] Even after 500 LbL deposition cycles the film is just 2 µm thick (Figure 10a, left). To get fully macroscopic materials, which can be used as stretchable conductors, several as-prepared films were laminated and consolidated into stacks by hot-pressing (Figure 10a, right). The most surprising aspect of these films is that they could be stretched up to 110% without losing their electrical conductivity. The reason, why the composite films preserved their conductivity in spite of the spherical shape of the gold nanoparticles was found to be in stress-induced nanoparticle organization into cellular networks along the stretching direction (Figure 10b). [15a] When it comes to the fabrication of macroscopic 3D bodies from preformed spherical nanoparticles, most of the efforts are dedicated to porous structures like foams or aerogels. Foams have the advantage that they are mechanically more stable due to a sintering step included in the processing, [94] while aerogels are fragile, but offer immense surface areas, exceptionally high porosity, and most importantly, full preservation of the nanoscale properties of the initial building blocks. [95] Different types of nanoparticles can be foamed and shaped into complex structures, which do not lose their macroscopic and monolithic form during drying and sintering of the wet foams. [96] Figure 10c shows a sintered alumina foam with complex shape. The preservation of the shape during heat treatment was only possible, because the surface of the nanoparticles was carefully lyophilized with short-chain amphiphilic molecules such that the nanoparticles strongly adsorbed at the air water interface and thus stabilized the air bubbles against coalescence. [97] A different strategy was pursued for the fabrication of copper foam monoliths (Figure 10d). To gain control over the porosity, spherical ZnO particles as templates were coated by a liquid phase process with copper. [98] After removal of the template, the hollow copper capsules were processed into different shapes by using standard techniques like slip-casting. [99] While the inner pore size of the capsules is defined by the diameter of the spherical template, the arrangement of the copper capsules within the body is random and disordered. The density of the foam can be precisely controlled by the degree of compaction with a minimum density of as low as 7% relative to bulk copper. However, there are many other ways to fabricate porous metal pieces of macroscopic sizes. [100] In contrast to the ceramic foams discussed above, where sintering leads to coalescence of the particles, nanoparticlebased aerogels are macroscopic bodies, in which the properties of the initial building blocks are still preserved. But similar to foaming, during gelation the surface chemistry of the nanoparticles plays a decisive role. In the fabrication of titania aerogel monoliths (Figure 10f) from preformed anatase nanoparticles, the selective removal of surface adsorbed organic ligands induced the formation of a finely structured, interwoven 3D network (Figure 10g). [101] The surface areas of the aerogels are considerably larger than those of the initial nanopowders, ranging from 300 up to 500 m 2 g 1. [102] However, the most fascinating aspect of using nanoparticle dispersions rather than molecular precursors for the preparation of aerogels is the possibility to combine different types of nanoparticles within the same monolith. Starting from a mixture of Fe 3 O 4 and TiO 2 nanoparticles, it is possible to produce binary aerogels, which are magnetically textured (if a magnetic field is applied during gelation such that the Fe 3 O 4 nanoparticles get aligned, Figure 10h), or exhibit a magnetic gradient (Figure 10i). [102] Interestingly, the porosity and the surface area of such aerogels can be subtly tuned by drying of the wet gels under ambient conditions up to a desired point, followed by supercritical drying. [103] While (12 of 18) 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

13 Figure 10. a) Photographs of free-standing, polyurethane gold nanoparticle composite films after 500 LbL cycles (1 LbL, left) and after laminating five films of 500 LbL cycles into one stack (5 LbL, right). b) SEM images of a focused-ion-beam milled 5 LbL film at various strains. Reproduced with permission. [15a] Copyright 2013, Macmillan Publishers Ltd. c) Photograph of a sintered alumina foam. d) Photograph of a cylindrical copper foam monolith. d) SEM image of the porous copper capsules as part of the monolith displayed in (d). Reproduced with permission. [99] Copyright 2013, Wiley-VCH Verlag GmbH & Co. KGaA. f) Photograph of a titania-nanoparticle-based aerogel monolith. g) SEM image of the microstructure of the aerogel displayed in (f). h) Photograph of a magnetically textured anatase magnetite aerogel monolith. i) Photograph of a layered anatase magnetite aerogel monolith with increasing magnetite concentration from top to bottom. Reproduced with permission. [102] Copyright 2014, The Royal Society of Chemistry. j) Photographs of BaTiO 3 gels at different drying stages (from left to right): as-gelled, after several hours of ambient drying, dried xerogel, xerogel sintered at 700 C. Reproduced with permission. [103] Copyright 2017, American Chemical Society. supercritical drying preserves the macroscopic shape and the microstructure of the wet gel, ambient drying results in extensive shrinkage (Figure 10j from left to right). The dried xerogel (second from right) lost over 97% of the original volume of the wet gel, and also the pore volume shrunk by 94%. After an additional heat treatment, the final body (Figure 10j, right) reached 60% of the bulk density of BaTiO 3. [103] Accordingly, the assembly of preformed nanoparticles into aero- and xerogels presents a flexible pathway to produce macroscopic materials with different types of porosities and densities. However, one of the big problems of such monoliths remains to be their mechanical fragility. InTaO 4 aerogels showed a compressive modulus of 0.04 and 1.4 MPa before and after sintering, respectively. [104] The latter value corresponds well to silica aerogels. [105] One way to introduce mechanical flexibility to aerogels to apply them, e.g., as anodes in lithium ion batteries with excellent electrochemical performance is to co-gel the inorganic nanoparticles with graphene oxide. [106] The preparation of aerogels is not limited to the use of spherical nanoparticles as building blocks. As a matter of fact, the gelation of nanowires or nanosheets represents a unique approach to transform lowdimensional nanoparticles into 3D bodies under full preservation of the anisotropy of the building blocks in the final network. Figure 11a c shows a photograph of a tungsten oxide aerogel made from tungsten oxide nanowires, [107] and Figure 11d f displays a Eu-doped yttrium oxide monolith composed of yttrium oxide nanosheets. [108] In the last part of this review, we will now discuss a few examples, which are rather complementary to aerogels, namely, the colloidal processing of dense composites and ceramics with a textured microstructure from suspended particles with anisotropic morphology. The final bodies are of macroscopic size and show a unique multiscale structure and complex shapes obtained by colloidal shape forming and additive manufacturing processes. [3] The preparation of macroscopic aerogel monoliths and the processing of textured ceramics basically represent efforts of two typically separated communities, which use different methods, however with comparable goals. Both communities start from preformed particles as building blocks, and both communities are well aware of the importance of having uniform and controlled particle sizes, morphologies, and purities (13 of 18) 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

14 Figure 11. a) Photograph of a tungsten oxide nanowire aerogel monolith. b) SEM and c) TEM images of the inner structure of the aerogel revealing the nanowire morphology. Reproduced with permission. [107] Copyright 2016, The Royal Society of Chemistry. d) Photograph of a Eu-doped yttrium oxide nanosheet aerogel monolith. e) SEM and f) TEM images of the microstructure of the aerogel revealing the nanosheet morphology. Reproduced with permission. [108] Copyright 2016, American Chemical Society. The big difference is that the size differs by a few orders of magnitude. While the nanocommunity faces the challenge to produce macroscopic materials from their nanosized building blocks, the ceramist deals with the issue to control the arrangement of the particles from the macro down to the micro or even nanoscale. In the end, the goal of both communities is to have a macroscopic material, in which the arrangement of the particulate building blocks is controlled over all length scales. The following examples are intended to highlight the state-of-the-art in colloidal processing with the focus on how the orientation of the particles can be controlled to produce textured materials. Introduction of texture is an effective means to improve the mechanical properties of ceramics. The application of an external field is one way to align anisotropic particles in a suspension. Interestingly, even if the magnetic susceptibility is low, like in the case of diamagnetic materials such as alumina, TiO 2, or ZnO, the energy of crystal anisotropy becomes comparable or larger than the energy of thermal motion in a high magnetic field, enabling the orientation of noncubic ceramic particles. [109] To make nonmagnetic particles also responsive to low magnetic fields, the particle size has to be optimized and their magnetic susceptibility has to be increased. The response of anisotropic particles like platelets and rods to an external magnetic field is the strongest, when the particle size is in the range of 5 20 µm, optimally balancing thermal motion and gravitational forces. [3] Surface coatings with small concentrations of superparamagnetic iron oxide nanoparticles enables orientation of diamagnetic particles in magnetic fields of less than 1 mt. Figure 12 shows the different orientations of iron oxide coated alumina platelets (20 vol%) embedded in an thermoplastic polyurethane elastomer. [110] The orientation of the platelets corresponds to the direction of the magnetic field applied. Figure 12a,b shows the arrangement of the platelets aligned by gravity without any magnetic field. In a static magnetic field, the platelets orient in the out-of-plane direction, however, without any in-plane ordering (Figure 12c,d). Such orientation can be achieved by a rotating magnetic field in the yz-plane, which produces a biaxial alignment of densely packed platelets (Figure 12e,f). Laminating together slabs with different platelet orientations gave access to multilayer composites, e.g., with a surface layer with out-of-plane oriented platelets for maximized hardness and wear resistance and internal layer with in-plane oriented platelets for high strength and toughness. Accordingly, the magnetic alignment of ceramic particles within a matrix offers the possibility to control the position and orientation of the reinforcing particles to locally tailor the mechanical properties including stiffness, strength, hardness, and wear resistance. Furthermore, the dimensions of the slabs were truly macroscopic with a thickness of 1 cm and width and length of 3 and 9 cm, respectively. [110] Combining this approach with slip-casting, a common processing route for the fabrication of complex shaped components from particle suspensions made it possible to design bioinspired composites with intricate macroscopic shapes and locally programmed texture. [111] The additive nature of the slip-casting process allows for great flexibility in terms of materials compositions and local texture control by simply changing the constituents in the fluid phase and by adjusting the angle of the rotating magnetic field. Similar to nanoparticle assembly, building blocks of different sizes, shapes, and compositions can be processed into a wide variety of different architectures. As a particularly intriguing example, Figure 12g shows the reproduction of the intricate microstructure and geometry of a biological tooth. [111] A bilayer structure mimicking the dentin enamel layers (Figure 12h j) was prepared by sequentially casting two different suspensions into a tooth-shaped porous gypsum mould. SEM investigations were performed to study the orientation of the reinforcing platelets in the two layers and it was found that the alignment agreed with the direction of the external magnetic field applied during (14 of 18) 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

15 Figure 12. Schematic (left) and top-view SEM images (right) of in-plane and out-of-plane alignment of iron oxide coated alumina platelets in a polyurethane elastomer: a,b) without magnetic field; c,d) with static magnetic field; and e,f) with rotating magnetic field. Reproduced with permission. [110] Copyright 2012, The American Association for the Advancement of Science. g) Optical microscopy image of a replicated natural tooth with h) schematic of its bilayer structure. i,j) Schematic of the interface with different platelet orientations mimicking the dentin enamel layers. j) SEM image of the different platelets orientations. Reproduced with permission. [111] Copyright 2015, Macmillan Publishers Ltd. k) Photograph of a 3D-printed object of macroscopic dimension with convex and concave curvatures and locally concentrated platelets arranged into a spiral staircase structure. Reproduced with permission. [112] Copyright 2015, Macmillan Publishers Ltd. the casting process (Figure 12j). Most importantly, the local mechanical properties were indeed influenced by the site-specific texture. The hardness was the highest at the outermost layer and continuously decreased toward the interior of the tooth. [111] Nowadays, such magnetic anisotropic particles can even be 3D printed. [112] Multimaterial dispenser, a two-component mixing unit, and the presence of a magnetic field enabled local control over particle orientation and composite composition. As an example, Figure 12k shows a photograph of a 3D printed object of macroscopic dimension with convex and concave curvatures. The interior of the object contains locally concentrated platelets, which form a helicoidal staircase arrangement from the bottom to the top. Printing of this elaborate heterogeneous structure was realized by combined deposition of a so-called shaping and texturing ink, exhibiting distinct rheological behaviors. The shaping ink prevented shape distortions, while the texturing ink enabled alignment of the suspended magnetic platelets in a low magnetic field. [112] The concept of dimensionality is well established in nanoparticle research, it is also extremely helpful for nanoparticle assembly. In colloidal processing, on the other hand, the effect of particle shape is much less explored. However, first examples clearly demonstrate the beneficial effect of particle alignment and orientation within bulk materials especially on the mechanical properties. These results should be motivating to intensify the efforts to combine the two worlds of ceramic processing and nanoscale assembly such that in the end materials of macroscopic dimensions with structural control over all length scales are accessible. 7. Conclusion and Outlook In materials science it is essential to address all length scales from the atoms to the macroscopic world. [6b] While in nanoparticle assembly basically the range from atoms to the micron scale is covered, the typical size regime of colloidal processing is in the micrometer to centimeter scale. Accordingly, if both processing techniques could be combined, all length scales of materials fabrication would be covered. But it is not only the size range that matters but also the conceptual priorities and complementarities. The first step in nanoparticle assembly is the controlled synthesis of the building blocks. Then, the building blocks are arranged and assembled with nearly particulate precision under full consideration of all the properties provided by the nanoparticles including composition, crystal structure, surface chemistry, and morphology (size, size distribution, and shape/dimensionality). These efforts typically result in highly defined structures, however, typically restricted to a size scale of a few tens of micrometers. The strength of colloidal processing, on the other hand, is the fabrication of bulk pieces with complex macroscopic shapes. Although also in colloidal processing the uniformity of the starting powders is recognized, the morphological aspects play a much smaller role than in nanoparticle assembly. However, first examples of taking advantage of the shape specific features of the building blocks in colloidal processing are promising and indicate the high potential. If the size and shape specific aspects of nanoparticle assembly can be combined with the ability to make complex shaped macroscopic bodies, then nanoparticle assembly (15 of 18) 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim