the reason for attractive forces is surface tension: the smaller the particles, the bigger the surface area, and as surface energy equals "surface tension x area", the bigger the area, the bigger the surface energy which acts on smaller and smaller particles ==> agglomeration.

Surfaces with high surface energy tend to get together because that way they can reduce their interfacial energy.

The formation of self-assembled (SA) is determined by several factors: particle-particle and particle–substrates interaction. In addition, SA of NPs formation is driven by van der Waals interaction, dipole-diploe interaction, coulomb interaction and maximization of entropy. For SA formation, each nanoparticles should behave as a molecule and compete against other forces. As explained by Bernhard Wessling, the surface tension is responsible for bringing them closer.

Even though self-assembly SA may be a more refined terminology, meant especially for the nanoparticles for examples, it may mean also agglomeration in general. This happens in any uncontrolled system of NPs, by randomly agglomerating at varying sizes as has been normally so, so far. However, to enable a lattice of NPs, i.e. nanolattice, they have to be monodispersed (same size) like those of atomic lattice, wherein sizes are identical or some groups of them say in YBa2Cu3O7-d. The problems  in monodispersed NPs not forming their own nanolattices were that they should form so only when some external forces such as extra surfactant/s need/s to be used to glue them together to enable a nanolattice/ superstructure. Nevertheless, earlier materials scientists were not able understand that such a forced nanolattice is due to electrostatic interactions EIs of the nanoparticles due to the extra surfacant/s. Such EIs should have been there in these non-extra cases also but not sufficient to ENABLE FORMATION OF NANOLATTICE/S. If one chooses an appropriate surfactant such that it can act as dual purposes: enabling mondispersed NPs plus EIs that enable also NPs' own EIs to be able to make their own nanolattice/s. We recently prove this in the case of Ni NPs using TOP; TPP and oleylamine do not help either making the NPs monodispersed or (hence) no nanolattice. Even though Bernhard Wessling may be correct (however still then the origin of the force will be EIs; see Solid State Physics by Kittel or Ashcroft and Mermin for example. Further, surface tension may have to deal with solid-liquid interface. Even if it is applicable in solid state also, force shall have to be created by EIs?), since the hcp nanolattices we revealed had their c/a ratios very identical to those of atomic lattices, wherein EIs enable their formation which is why we preliminarily attempted to explain this way. Similar to the extra-nuclear electrons and nuclei in atomic lattices, negative and positive charges from the ligands of the surfactants/ imppurities might prevail to enable the nanolattice formation: cations of the ion pairs (Fig. 7, inset  of our JMCC paper) of the dissociated surfactant molecules or impurities attached on a nanoparticle may attract the anions of the surrounding nanoparticles until they sense the presence of other cations of the latter, leading to a repulsion.  Hence........