TERCERA ETAPA EVALUACIÓN DEL PROCESO DE EDUCACIÓN PARA LA PAZ.

In the previous section we considered a possible mesocrystallization pathway where crystal alignment was based on the fact that crystal partitions can easily rotate within the Wigner glass, and can align even before they are in close contact. A subsequent compression of the aligned system then ensures low density and the integrity of the desired mesocrystalline structure.

While parts of this mechanism are realistic both experimentally and in a simulation, the possibility of alignment within the Wigner glass is newly discussed here and is only hypothetical. Let us now address the role of collective rotational motion in another mechanism of mesocrystal formation which, as has been described previously by Leite and Ribeiro [2012], is based on the orientation dependent particle collisions.

In this model, partitions are represented by a system of equally sized and shaped nanocrystals with a crystal lattice defining their orientation. The system behaves like a gas of nanocrystals undergoing phase separation or gelation depend- ing on the strength of interaction. The interaction between the partitions is special in that its strength is highly dependent on whether the colliding partitions are crys- tallographically aligned or misoriented. If the collision between the nanocrystals is such that their lattice orientations match at contact (epitaxial alignment), the overall attractive interaction between the partitions is expected to be strong, irre- versible, and the collision is referred to aseffective. If the collision is such that the lattices are not aligned (nonepitaxial attachment), the energy associated with the creation of the interface is not supposed to be high, the collisions may be reversible and termednon-effective.

We will distinguish between reversible and irreversible collisions, and will consider two possible relative rotations. First, partitions collide and split in a non- effective fashion until an effective collision occurs. In this case rotations happen while partitions are isolated, i.e. do not form a single partition through any short- range contact between the two partitions. Second, partitions remain attached after a non-effective collision, and relative rotations of the two partitions (more precisely two subpartitions within a single partition) happen until they align.

In systems where long-range repulsion is strong enough (dispersed colloidal state), the crystalline partitions are expected to collide until the collision is such that the respective crystalline surfaces are crystallographically aligned. We know that the binding and unbinding probability is exponentially dependent on the energy of attachment [Grant et al., 2011], and that even a small energy difference between the non-aligned and aligned attachment may lead to the former being able to break much easier than the latter.

Irreversibility of the effective collisions and reversibility of the non-effective collisions are thus plausible assumptions. Nevertheless, the probability of an ideal collision where the crystal surfaces and their edges are aligned is low. The nanocrys- tals seem to be more likely to from an interface when they are part of a single par- tition. The driving force for alignment can again result from favourable energetics, interfacial tension [Leite and Ribeiro, 2012], or from the internal strains [Chen et al., 2014].

Our simulations do not lead to a perfect collection of uniformly sized and shaped nanocrystals, but we are able to obtain a system composed of isolated ag- gregates with a high internal crystallinity and with a well developed crystal surface. We propose improvements to the model, and to the simulation protocol, such that higher crystallinity and lower polydispersity may be achieved. This allows us to assume an equilibrium fluid composed of similarly shaped crystals with random lat- tice orientations, and point out the overlaps in mechanisms observed both in our model and in the collision model. We discuss in Sec. 3.6 that the renormalization of the partitions into larger spheres leads not only to shorter attraction range relative to the spheres, but also to the increase of its strength due to the larger number of microscopic contacts between particles constituting the two colliding partitions. The number of those contacts is generally highest if the attachment is epitaxial. The epitaxial alignment is thus energetically and not entropically favoured.

Our VMMC simulations showed that the intrapartition collective rearrange- ments are unlikely, especially if the partitions have high crystallinity. This suggests that the simulations do not capture the large collective rearrangements known as grain-rotation-induced grain growth mechanisms. This may be due to the simu- lation parameters, but also due to other properties of the system. The centres of rotations may be chosen in a more realistic or in a more efficient way. The geometry of the pairwise potential also plays a role. For example, Whitelam et al. [2009a] reported that better assembly by collective motion can be achieved in viral capsids than in spherical systems. It is also clear that flat surfaces composed of cubical nanoparticles [Smallenburg et al., 2012] can glide in contact more easily than rough

surfaces composed of spheres. The temperature choice is also worth specifying. It should not be higher than the melting temperature, but must be high enough such that self-recrystallization of the grain boundaries can happen after the oriented at- tachment [Leite and Ribeiro, 2012]. In the recent simulation study of Bjerre et al. [2013], the optimum temperature for grain-rotation-induced grain mechanism has been identified in the vicinity of the melting temperature. It would be interesting to compare this temperature with the temperature region discussed in this thesis.

Finally, precise control over the rotations in the generalized VMMC may also affect the non-classical crystallization process. If the crystalline partitions are represented by anisotropic particles, with specific directional interactions defined by their lattice orientation, the results of Hedges and Whitelam [2011] suggest that a cluster liquid is more likely to precede the cluster solid, if clusters rotate sluggishly. Hedges and Whitelam [2011] also claim that partitions must be aligned in order to crystallize, and that the specific inter-partition interactions must have an optimum value, in analogy with what has been discussed above.